1 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3 * lto-section-out.c (lto_begin_section): Do not print section
4 name for noaddr and unnumbered dumps.
6 2018-06-25 Richard Biener <rguenther@suse.de>
8 * tree-vectorizer.h (struct vec_info_shared): New structure
9 with parts split out from struct vec_info and loop_nest from
10 struct _loop_vec_info.
11 (struct vec_info): Adjust accordingly.
12 (struct _loop_vec_info): Likewise.
13 (LOOP_VINFO_LOOP_NEST): Adjust.
14 (LOOP_VINFO_DATAREFS): Likewise.
15 (LOOP_VINFO_DDRS): Likewise.
16 (struct _bb_vec_info): Likewise.
17 (BB_VINFO_DATAREFS): Likewise.
18 (BB_VINFO_DDRS): Likewise.
19 (struct _stmt_vec_info): Add dr_aux member.
20 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
21 (DR_MISALIGNMENT_UNINITIALIZED): New.
22 (set_dr_misalignment): Adjust.
23 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
24 (vect_analyze_loop): Adjust prototype.
25 (vect_analyze_loop_form): Likewise.
26 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
27 Compute dependences lazily.
28 (vect_record_base_alignments): Use shared datarefs/ddrs.
29 (vect_verify_datarefs_alignment): Likewise.
30 (vect_analyze_data_refs_alignment): Likewise.
31 (vect_analyze_data_ref_accesses): Likewise.
32 (vect_analyze_data_refs): Likewise.
33 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
34 constructor parameter for shared part.
35 (vect_analyze_loop_form): Pass in shared part and adjust.
36 (vect_analyze_loop_2): Pass in storage for the number of
37 stmts. Move loop nest finding to the caller. Compute
39 (vect_analyze_loop): Pass in shared part.
40 (vect_transform_loop): Verify shared datarefs are unchanged.
41 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
42 constructor parameter for shared part.
43 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
44 (vect_slp_bb): Verify shared datarefs are unchanged before
46 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
48 (new_stmt_vec_info): Initialize DR_AUX misalignment to
49 DR_MISALIGNMENT_UNINITIALIZED.
50 * tree-vectorizer.c (vec_info::vec_info): Add constructor
51 parameter for shared part.
52 (vec_info::~vec_info): Adjust.
53 (vec_info_shared::vec_info_shared): New.
54 (vec_info_shared::~vec_info_shared): Likewise.
55 (vec_info_shared::save_datarefs): Likewise.
56 (vec_info_shared::check_datarefs): Likewise.
57 (try_vectorize_loop_1): Construct shared part live for analyses
58 of a single loop for multiple vector sizes.
59 * tree-parloops.c (gather_scalar_reductions): Adjust.
61 2018-06-25 Richard Biener <rguenther@suse.de>
63 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
64 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
65 (vect_analyze_data_refs): Remove similar code from here and
68 2018-06-25 Richard Biener <rguenther@suse.de>
70 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
71 for reverse storage order accesses rather than asserting
72 they cannot happen here.
74 2018-06-25 Tom de Vries <tdevries@suse.de>
77 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
78 Use data16 instead of .byte for insn prefix.
80 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
83 * parser.c (make_char_string_pack): Pass this literal chars
84 through cpp_interpret_string.
85 (cp_parser_userdef_numeric_literal): Check the result of
86 make_char_string_pack.
88 2018-06-24 Maya Rashish <coypu@sdf.org>
90 * ginclude/stddef.h: Simplify conditions around avoiding
91 re-definition of __size_t.
93 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
95 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
96 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
98 2018-06-22 Maya Rashish <coypu@sdf.org>
100 * doc/invoke.texi (mno-fancy-math-387): Update for changes
101 made to OpenBSD and NetBSD through the years.
103 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
105 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
106 behavior of vec_pack (vector double, vector double) to match
107 behavior of vec_float2 (vector double, vector double).
109 2018-06-22 Olivier Hainque <hainque@adacore.com>
111 * gimplify.c (gimplify_function_tree): Prevent creation
112 of a trampoline for the address of the current function
113 passed to entry/exit instrumentation hooks.
115 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
118 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
121 2018-06-22 Martin Liska <mliska@suse.cz>
123 PR tree-optimization/86263
124 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
125 Bail out if is_enabled is false.
126 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
128 (jump_table_cluster::is_enabled): New function.
130 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
132 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
133 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
134 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
135 (lto_input_ts_binfo_tree_pointers): Likewise.
136 * tree-streamer-out.c (streamer_write_tree_bitfields,
137 write_ts_binfo_tree_pointers): Likewise.
138 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
140 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
142 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
144 2018-06-22 Martin Liska <mliska@suse.cz>
146 * symbol-summary.h (get): Make it pure and inline move
147 functionality from ::get function.
148 (get): Remove and inline into ::get and ::get_create.
149 (get_create): Move code from ::get function.
151 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
154 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
155 -x assembler-with-cpp.
157 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
159 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
160 _FILE_OFFSET_BITS=64 for C++.
162 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
164 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
165 conversion insn that shows up when pr85657-3.c is compiled using
166 IEEE 128-bit long double.
167 (neg<mode>2_internal): Use the correct mode to check whether the
168 mode is IBM extended.
169 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
170 multiply and divide external functions from being created more
173 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
175 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
177 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
178 the edge can be forwarded.
179 (cfg_layout_merge_blocks): Likewise.
181 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
183 * except.c (finish_eh_generation): Commit edge insertions only after
184 the EH edges have been redirected from post-landing to landing pads.
186 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
188 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
189 create_tmp_var_for to create the FRAME decl.
190 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
192 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
194 * tree-inline.c (copy_edges_for_bb): Minor tweak.
195 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
196 debug statement when resetting its value.
197 (expand_call_inline): Copy the locus of the call onto the assignment
198 of the return value, if any. Use local variable in more cases.
200 2018-06-21 Martin Liska <mliska@suse.cz>
202 * ipa-pure-const.c (propagate_nothrow): Use
203 funct_state_summaries->get.
204 (dump_malloc_lattice): Likewise.
205 (propagate_malloc): Likewise.
207 2018-06-21 Richard Biener <rguenther@suse.de>
209 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
210 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
211 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
213 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
214 BLOCK_ABSTRACT_ORIGIN unconditionally.
216 2018-06-21 David Malcolm <dmalcolm@redhat.com>
218 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
220 * ipa-reference.c (ipa_reference_c_finalize): Delete
221 ipa_ref_opt_sum_summaries and set it to NULL.
223 2018-06-21 Tom de Vries <tdevries@suse.de>
225 PR tree-optimization/85859
226 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
227 test with comment from bb_no_side_effects_p.
229 2018-06-21 Richard Biener <rguenther@suse.de>
231 PR tree-optimization/86232
232 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
233 max for constant niter.
235 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
237 * config/aarch64/aarch64-simd.md
238 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
240 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
242 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
243 Make opernads of the unspec commutative.
245 2018-06-21 Richard Biener <rguenther@suse.de>
247 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
248 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
249 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
250 (vect_analyze_data_ref_dependence): Re-order checks to deal with
252 (vect_record_base_alignments): Do not record base alignment
253 for gathers or scatters.
254 (vect_compute_data_ref_alignment): Drop return value that is always
255 true. Bail out early for gathers or scatters.
256 (vect_enhance_data_refs_alignment): Bail out early for gathers
258 (vect_find_same_alignment_drs): Likewise.
259 (vect_analyze_data_refs_alignment): Remove dead code.
260 (vect_slp_analyze_and_verify_node_alignment): Likewise.
261 (vect_analyze_data_refs): For possible gathers or scatters do
262 not create an alternate DR, just check their possible validity
263 and mark them. Adjust DECL_NONALIASED handling to not rely
265 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
266 update inits of gathers or scatters.
267 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
268 Also copy gather/scatter flag to pattern vinfo.
270 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
272 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
273 behavior of vec_packsu (vector unsigned long long, vector unsigned
274 long long) to match behavior of vec_packs with same signature.
276 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
277 Thomas Schwinge <thomas@codesourcery.com>
278 Cesar Philippidis <cesar@codesourcery.com>
280 * gimplify.c (gimplify_scan_omp_clauses): Add support for
281 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
282 (gimplify_adjust_omp_clauses): Likewise.
283 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
284 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
285 (gimplify_omp_target_update): Update handling of acc update and
287 * omp-low.c (install_var_field): Remove unused parameter
288 base_pointers_restrict.
289 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
290 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
292 (omp_target_base_pointers_restrict_p): Delete.
293 (scan_omp_target): Update call to scan_sharing_clauses.
294 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
296 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
297 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
298 (convert_local_omp_clauses): Likewise.
299 * tree-pretty-print.c (dump_omp_clause): Likewise.
300 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
302 (omp_clause_code_name): Likewise.
304 2018-06-20 Jakub Jelinek <jakub@redhat.com>
307 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
310 PR tree-optimization/86231
311 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
312 anti-range don't overwrite *vr0min before using it to compute *vr0max.
314 2018-06-20 Tom de Vries <tdevries@suse.de>
316 PR tree-optimization/86097
317 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
318 iv type if signedness of iv type is not the same as that of *nit.
320 2018-06-20 Jakub Jelinek <jakub@redhat.com>
322 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
323 EDGE_EH edges, verify they are all EDGE_EH.
325 2018-06-20 Maya Rashish <coypu@sdf.org>
327 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
329 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
331 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
332 * config/aarch64/aarch64.c (xgene1_tunings): Add
333 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
334 (aarch64_mode_valid_for_sched_fusion_p):
336 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
337 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
339 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
341 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
342 * config/aarch64/iterators.md (VQ2): New mode iterator.
344 2018-06-20 Martin Liska <mliska@suse.cz>
346 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
347 Change default ratio from 10 to 8.
349 2018-06-20 Martin Liska <mliska@suse.cz>
351 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
353 (bit_test_cluster::find_bit_tests): Likewise.
354 (switch_decision_tree::analyze_switch_statement): Find clusters.
355 * tree-switch-conversion.h (struct jump_table_cluster): Document
358 2018-06-20 Martin Liska <mliska@suse.cz>
360 * tree-switch-conversion.c (switch_conversion::collect):
361 Record m_uniq property.
362 (switch_conversion::expand): Bail out for special conditions.
363 (group_cluster::~group_cluster): New.
364 (group_cluster::group_cluster): Likewise.
365 (group_cluster::dump): Likewise.
366 (jump_table_cluster::emit): New.
367 (switch_decision_tree::fix_phi_operands_for_edges): New.
368 (struct case_node): Remove struct.
369 (jump_table_cluster::can_be_handled): New.
370 (case_values_threshold): Moved to header.
371 (reset_out_edges_aux): Likewise.
372 (jump_table_cluster::is_beneficial): New.
373 (bit_test_cluster::can_be_handled): Likewise.
374 (add_case_node): Remove.
375 (bit_test_cluster::is_beneficial): New.
376 (case_bit_test::cmp): New.
377 (bit_test_cluster::emit): New.
378 (expand_switch_as_decision_tree_p): Remove.
379 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
380 (fix_phi_operands_for_edge): Likewise.
381 (switch_decision_tree::analyze_switch_statement): New.
382 (compute_cases_per_edge): Move ...
383 (switch_decision_tree::compute_cases_per_edge): ... here.
384 (try_switch_expansion): Likewise.
385 (switch_decision_tree::try_switch_expansion): Likewise.
386 (record_phi_operand_mapping): Likewise.
387 (switch_decision_tree::record_phi_operand_mapping): Likewise.
388 (emit_case_decision_tree): Likewise.
389 (switch_decision_tree::emit): Likewise.
390 (balance_case_nodes): Likewise.
391 (switch_decision_tree::balance_case_nodes): Likewise.
392 (dump_case_nodes): Likewise.
393 (switch_decision_tree::dump_case_nodes): Likewise.
394 (emit_jump): Likewise.
395 (switch_decision_tree::emit_jump): Likewise.
396 (emit_cmp_and_jump_insns): Likewise.
397 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
398 (emit_case_nodes): Likewise.
399 (switch_decision_tree::emit_case_nodes): Likewise.
400 (conditional_probability): Remove.
401 * tree-switch-conversion.h (enum cluster_type): New.
403 (struct cluster): Likewise.
404 (cluster::cluster): Likewise.
405 (struct simple_cluster): Likewise.
406 (simple_cluster::simple_cluster): Likewise.
407 (struct group_cluster): Likewise.
408 (struct jump_table_cluster): Likewise.
409 (struct bit_test_cluster): Likewise.
410 (struct min_cluster_item): Likewise.
411 (struct case_tree_node): Likewise.
412 (case_tree_node::case_tree_node): Likewise.
413 (jump_table_cluster::case_values_threshold): Likewise.
414 (struct case_bit_test): Likewise.
415 (struct switch_decision_tree): Likewise.
416 (struct switch_conversion): Likewise.
417 (switch_decision_tree::reset_out_edges_aux): Likewise.
419 2018-06-20 Martin Liska <mliska@suse.cz>
421 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
422 (hoist_edge_and_branch_if_true): Likewise.
423 (expand_switch_using_bit_tests_p): Likewise.
424 (struct case_bit_test): Likewise.
425 (case_bit_test_cmp): Likewise.
426 (emit_case_bit_tests): Likewise.
427 (switch_conversion::switch_conversion): New class.
428 (struct switch_conv_info): Remove old struct.
429 (collect_switch_conv_info): More to ...
430 (switch_conversion::collect): ... this.
431 (check_range): Likewise.
432 (switch_conversion::check_range): Likewise.
433 (check_all_empty_except_final): Likewise.
434 (switch_conversion::check_all_empty_except_final): Likewise.
435 (check_final_bb): Likewise.
436 (switch_conversion::check_final_bb): Likewise.
437 (create_temp_arrays): Likewise.
438 (switch_conversion::create_temp_arrays): Likewise.
439 (free_temp_arrays): Likewise.
440 (gather_default_values): Likewise.
441 (switch_conversion::gather_default_values): Likewise.
442 (build_constructors): Likewise.
443 (switch_conversion::build_constructors): Likewise.
444 (constructor_contains_same_values_p): Likewise.
445 (switch_conversion::contains_same_values_p): Likewise.
446 (array_value_type): Likewise.
447 (switch_conversion::array_value_type): Likewise.
448 (build_one_array): Likewise.
449 (switch_conversion::build_one_array): Likewise.
450 (build_arrays): Likewise.
451 (switch_conversion::build_arrays): Likewise.
452 (gen_def_assigns): Likewise.
453 (switch_conversion::gen_def_assigns): Likewise.
454 (prune_bbs): Likewise.
455 (switch_conversion::prune_bbs): Likewise.
456 (fix_phi_nodes): Likewise.
457 (switch_conversion::fix_phi_nodes): Likewise.
458 (gen_inbound_check): Likewise.
459 (switch_conversion::gen_inbound_check): Likewise.
460 (process_switch): Use the newly created class.
461 (switch_conversion::expand): New.
462 (switch_conversion::~switch_conversion): New.
463 * tree-switch-conversion.h: New file.
465 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
467 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
468 tree-vect-patterns.c.
469 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
470 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
471 (vect_recog_sad_pattern): Likewise.
472 (vect_recog_widen_sum_pattern): Likewise.
473 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
474 (vect_recog_widen_shift_pattern): Remove the type_in argument.
475 (vect_recog_rotate_pattern): Likewise.
476 (vect_recog_mult_pattern): Likewise.
477 (vect_recog_vector_vector_shift_pattern): Likewise.
478 (vect_recog_divmod_pattern): Likewise.
479 (vect_recog_mixed_size_cond_pattern): Likewise.
480 (vect_recog_bool_pattern): Likewise.
481 (vect_recog_mask_conversion_pattern): Likewise.
482 (vect_try_gather_scatter_pattern): Likewise.
483 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
484 (vect_recog_over_widening_pattern): Likewise.
485 (vect_recog_gather_scatter_pattern): Likewise.
486 (vect_recog_func_ptr): Move from tree-vectorizer.h
487 (vect_vect_recog_func_ptrs): Move further down the file.
488 (vect_recog_func): Likewise. Remove the third argument.
489 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
490 (vect_pattern_recog_1): Expect the pattern function to do any
491 necessary target tests. Also expect it to provide a vector type.
492 Remove the type_in handling.
494 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
496 * tree-vect-patterns.c (vect_pattern_detected): New function.
497 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
498 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
499 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
500 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
501 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
502 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
503 (vect_recog_mask_conversion_pattern)
504 (vect_try_gather_scatter_pattern): Likewise.
506 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
508 * tree-vect-patterns.c (vect_get_internal_def): New function.
509 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
510 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
511 (search_type_for_mask_1): Use it.
513 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
515 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
516 redundant WIDEN_SUM_EXPR handling.
517 (vect_recog_sad_pattern): Likewise.
519 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
521 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
522 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
523 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
524 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
525 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
527 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
529 * tree-vect-stmts.c (vectorizable_call): Make sure that we
530 use the stmt_vec_info of the original bb statement for the
531 new zero assignment, even if the call is part of a pattern.
533 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
535 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
536 that the sequence is attached to the original statement rather
537 than the pattern statement.
538 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
539 PATTERN_DEF_SEQ from the original statement rather than
540 the main pattern statement.
541 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
542 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
543 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
545 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
547 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
548 definition statements before the early exit for statements that aren't
550 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
552 (vect_transform_loop): ...here. Process pattern definition
553 statements without first checking whether the main pattern
554 statement is live or relevant.
556 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
558 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
559 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
561 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
563 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
564 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
565 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
566 (expand_block_compare): Change select_block_compare_mode call.
567 (expand_strncmp_align_check): Use new functions, fix comment.
568 (emit_final_str_compare_gpr): New function.
569 (expand_strn_compare): Refactor and clean up code.
570 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
572 2018-06-19 Tony Reix <tony.reix@atos.com>
573 Damien Bergamini <damien.bergamini@atos.com>
574 David Edelsohn <dje.gcc@gmail.com>
576 * collect2.c (static_obj): New variable.
577 (static_libs): New variable.
578 (is_in_list): Uncomment declaration.
579 (main): Track AIX libraries linked statically.
580 (is_in_list): Uncomment definition.
581 (scan_prog_file): Don't add AIX shared libraries initializer
582 to constructor list if linking statically.
584 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
586 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
588 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
590 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
592 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
595 2018-06-19 Martin Liska <mliska@suse.cz>
597 * config/i386/i386.c (ix86_can_inline_p): Do not use
598 ipa_fn_summaries::get_create.
599 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
601 (devirtualization_time_bonus): Likewise.
602 (ipcp_propagate_stage): Likewise.
603 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
604 (edge_set_predicate): Likewise.
605 (evaluate_conditions_for_known_args): Likewise.
606 (evaluate_properties_for_edge): Likewise.
607 (ipa_call_summary::reset): Tranform to ...
608 (ipa_call_summary::~ipa_call_summary): ... this.
609 (ipa_fn_summary::reset): Transform to ...
610 (ipa_fn_summary::~ipa_fn_summary): ... this.
611 (ipa_fn_summary_t::remove): Rename to ...
612 (ipa_fn_summary_t::remove_callees): ... this.
613 (ipa_fn_summary_t::duplicate): Use placement new
614 instead of memory copy.
615 (ipa_call_summary_t::duplicate): Likewise.
616 (ipa_call_summary_t::remove): Remove.
617 (dump_ipa_call_summary): Change get_create to get.
618 (ipa_dump_fn_summary): Dump only when summary exists.
619 (analyze_function_body): Use symbol_summary::get instead
621 (compute_fn_summary): Likewise.
622 (estimate_edge_devirt_benefit): Likewise.
623 (estimate_edge_size_and_time): Likewise.
624 (inline_update_callee_summaries): Likewise.
625 (remap_edge_change_prob): Likewise.
626 (remap_edge_summaries): Likewise.
627 (ipa_merge_fn_summary_after_inlining): Likewise.
628 (write_ipa_call_summary): Likewise.
629 (ipa_fn_summary_write): Likewise.
630 (ipa_free_fn_summary): Likewise.
631 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
632 (struct ipa_call_summary): Likewise.
633 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
635 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
636 (estimate_size_after_inlining): Likewise.
637 (estimate_growth): Likewise.
638 (growth_likely_positive): Likewise.
639 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
640 (inline_call): Likewise.
641 * ipa-inline.c (caller_growth_limits): Likewise.
642 (can_inline_edge_p): Likewise.
643 (can_inline_edge_by_limits_p): Likewise.
644 (compute_uninlined_call_time): Likewise.
645 (compute_inlined_call_time): Likewise.
646 (want_inline_small_function_p): Likewise.
647 (edge_badness): Likewise.
648 (update_caller_keys): Likewise.
649 (update_callee_keys): Likewise.
650 (inline_small_functions): Likewise.
651 (inline_to_all_callers_1): Likewise.
652 (dump_overall_stats): Likewise.
653 (early_inline_small_functions): Likewise.
654 (early_inliner): Likewise.
655 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
656 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
657 * ipa-pure-const.c (malloc_candidate_p): Likewise.
658 * ipa-split.c (execute_split_functions): Likewise.
659 * symbol-summary.h: Likewise.
660 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
662 2018-06-19 Richard Biener <rguenther@suse.de>
664 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
665 (vectorize_loops): ... here. Fix dbgcnt handling.
666 (try_vectorize_loop): Wrap try_vectorize_loop_1.
668 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
671 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
672 ieee128 argument takes up only one (vector) register, not two (floating
675 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
677 * gimplify.c (gimplify_init_constructor): Really never clear for an
678 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
680 2018-06-19 Richard Biener <rguenther@suse.de>
682 PR tree-optimization/86179
683 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
684 after failed recognition.
686 2018-06-18 Martin Sebor <msebor@redhat.com>
689 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
690 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
691 Handle integer subtraction.
692 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
693 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
695 2018-06-18 David Malcolm <dmalcolm@redhat.com>
697 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
698 param from rtx to rtx_insn *.
699 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
701 (frv_ifcvt_modify_insn): Likwise.
702 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
703 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
704 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
705 as_a <rtx_insn *> cast to local "unprotected_region" once
706 it's been established that it's not NULL or pc_rtx.
707 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
708 param "sethi" from rtx to rtx_insn *.
709 (nds32_group_float_insns): Likewise for param "insn".
710 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
712 (vax_output_int_subtract): Likewise.
713 * config/vax/vax.c (vax_output_int_add): Likewise for param
715 (vax_output_int_subtract): Likewise.
716 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
717 (emit_pattern_after): Likewise for param "after".
718 (emit_insn_after): Likewise.
719 (emit_jump_insn_after): Likewise.
720 (emit_call_insn_after): Likewise.
721 (emit_debug_insn_after): Likewise.
722 (emit_pattern_before): Likewise for param "before".
723 (emit_insn_before): Likewise.
724 (emit_jump_insn_before): Likewise.
725 * final.c (get_insn_template): Likewise for param "insn", removing
727 * output.h (get_insn_template): Likewise for 2nd param.
728 * rtl.h (emit_insn_before): Likewise.
729 (emit_jump_insn_before): Likewise.
730 (emit_debug_insn_before_noloc): Likewise.
731 (emit_insn_after): Likewise.
732 (emit_jump_insn_after): Likewise.
733 (emit_call_insn_after): Likewise.
734 (emit_debug_insn_after): Likewise.
735 (set_insn_deleted): Likewise for param.
737 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
740 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
741 floating point modes, so that IFmode is numerically greater than
742 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
743 to declare the ordering. This prevents IFmode from being
744 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
745 machine. Include rs6000-modes.h to share the fractional values
746 between genmodes* and the rest of the compiler.
750 * config/rs6000/rs6000-modes.h: New file.
751 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
752 meaning of rs6000_long_double_size so that 126..128 selects an
753 appropriate 128-bit floating point type.
754 (rs6000_option_override_internal): Likewise.
755 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
756 (TARGET_LONG_DOUBLE_128): Change the meaning of
757 rs6000_long_double_size so that 126..128 selects an appropriate
758 128-bit floating point type.
759 (LONG_DOUBLE_TYPE_SIZE): Update comment.
760 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
761 source and destination to match the standard usage.
762 (truncifkf2): Likewise.
763 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
764 ISA 2.07 to use an explicit clobber, instead of passing in a
766 (copysign<mode>3_soft): Likewise.
768 2018-06-18 David Malcolm <dmalcolm@redhat.com>
770 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
771 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
772 (vect_slp_analyze_instance_dependence): Likewise.
773 (vect_enhance_data_refs_alignment): Likewise.
774 (vect_analyze_data_refs_alignment): Likewise.
775 (vect_slp_analyze_and_verify_instance_alignment
776 (vect_analyze_data_ref_accesses): Likewise.
777 (vect_prune_runtime_alias_test_list): Likewise.
778 (vect_analyze_data_refs): Likewise.
779 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
780 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
781 (vect_analyze_scalar_cycles_1): Likewise.
782 (vect_get_loop_niters): Likewise.
783 (vect_analyze_loop_form_1): Likewise.
784 (vect_update_vf_for_slp): Likewise.
785 (vect_analyze_loop_operations): Likewise.
786 (vect_analyze_loop): Likewise.
787 (vectorizable_induction): Likewise.
788 (vect_transform_loop): Likewise.
789 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
790 * tree-vect-slp.c (vect_analyze_slp): Likewise.
791 (vect_make_slp_decision): Likewise.
792 (vect_detect_hybrid_slp): Likewise.
793 (vect_slp_analyze_operations): Likewise.
794 (vect_slp_bb): Likewise.
795 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
796 (vectorizable_bswap): Likewise.
797 (vectorizable_call): Likewise.
798 (vectorizable_simd_clone_call): Likewise.
799 (vectorizable_conversion): Likewise.
800 (vectorizable_assignment): Likewise.
801 (vectorizable_shift): Likewise.
802 (vectorizable_operation): Likewise.
803 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
805 2018-06-18 Martin Sebor <msebor@redhat.com>
807 PR tree-optimization/81384
808 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
809 * builtins.c (expand_builtin_strnlen): New function.
810 (expand_builtin): Call it.
811 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
812 * builtins.def (BUILT_IN_STRNLEN): New.
813 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
814 Warn for bounds in excess of maximum object size.
815 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
816 single-value ranges. Handle strnlen.
817 (handle_builtin_strlen): Handle strnlen.
818 (strlen_check_and_optimize_stmt): Same.
819 * doc/extend.texi (Other Builtins): Document strnlen.
821 2018-06-18 Maya Rashish <coypu@sdf.org>
823 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
824 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
825 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
827 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
829 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
831 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
833 * tree.c (escaped_string::escape): Replace cast to char * by
834 const_cast<char *> (unescaped).
836 2018-06-18 Nick Clifton <nickc@redhat.com>
839 * tree.c (escaped_string): New class. Converts an unescaped
840 string into its escaped equivalent.
841 (warn_deprecated_use): Use the new class to convert the
842 deprecation message, if present.
843 (test_escaped_strings): New self test.
844 (test_c_tests): Add test_escaped_strings.
845 * doc/extend.texi (deprecated): Add a note that the
846 deprecation message is affected by the -fmessage-length
847 option, and that control characters will be escaped.
848 (#pragma GCC error): Document this pragma.
849 (#pragma GCC warning): Likewise.
850 * doc/invoke.texi (-fmessage-length): Document this option's
851 effect on the #warning and #error preprocessor directives and
852 the deprecated attribute.
854 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
856 * tree.c (decl_value_expr_lookup): Revert latest change.
857 (decl_value_expr_insert): Likewise.
859 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
861 * gimplify.c (nonlocal_vlas): Delete.
862 (nonlocal_vla_vars): Likewise.
863 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
865 (gimplify_body): Do not create and destroy nonlocal_vlas.
866 * tree-nested.c: Include diagnostic.h.
867 (use_pointer_in_frame): Tweak.
868 (lookup_field_for_decl): Add assertion and declare the transformation.
869 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
870 internal error when the reference is in a wrong context. Do not
871 create a debug decl by default.
872 (note_nonlocal_block_vlas): Delete.
873 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
874 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
875 create a debug decl by default.
876 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
877 call is in a wrong context.
878 (fixup_vla_decls): New function.
879 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
880 debug variables were created.
881 * tree.c (decl_value_expr_lookup): Add checking assertion.
882 (decl_value_expr_insert): Likewise.
884 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
887 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
888 * tree-scalar-evolution.c (interpret_expr): Likewise.
889 (expression_expensive_p): Likewise.
890 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
891 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
892 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
893 (ssa_defined_by_minus_one_stmt_p): New.
895 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
898 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
899 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
900 * dojump.c (do_jump): Likewise.
901 * expr.c (expand_expr_real_2): Check operand type's sign.
902 * fold-const.c (const_unop): Handle ABSU_EXPR.
903 (fold_abs_const): Likewise.
904 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
905 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
906 (strip_sign_op_1): Likesise.
907 * match.pd: Add new pattern to generate ABSU_EXPR.
908 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
909 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
910 * tree-eh.c (operation_could_trap_helper_p): Likewise.
911 * tree-inline.c (estimate_operator_cost): Likewise.
912 * tree-pretty-print.c (dump_generic_node): Likewise.
913 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
914 * tree.def (ABSU_EXPR): New.
916 2018-06-16 Jakub Jelinek <jakub@redhat.com>
919 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
920 documented as preserved for backward compatibility only.
921 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
923 PR rtl-optimization/86108
924 * bb-reorder.c (create_forwarder_block): Renamed to ...
925 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
926 jump from new landing pad to the second part.
927 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
930 2018-06-15 Jakub Jelinek <jakub@redhat.com>
933 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
934 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
935 Only call store_expr for halves if the mode is the same.
938 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
939 Fix up comment formatting.
941 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
943 * typed-splay-tree.h (typed_splay_tree::remove): New function.
944 (typed_splay_tree::closure,
945 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
946 (typed_splay_tree::typed_splay_tree,
947 typed_splay_tree::operator =): Declared private.
948 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
949 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
950 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
951 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
952 typed_splay_tree::splay_tree_splay,
953 typed_splay_tree::splay_tree_foreach_helper,
954 typed_splay_tree::splay_tree_insert,
955 typed_splay_tree::splay_tree_remove,
956 typed_splay_tree::splay_tree_lookup,
957 typed_splay_tree::splay_tree_predecessor,
958 typed_splay_tree::splay_tree_successor,
959 typed_splay_tree::splay_tree_min,
960 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
961 (typed_splay_tree::root, typed_splay_tree::comp,
962 typed_splay_tree::delete_key,
963 typed_splay_tree::delete_value): New data members.
964 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
965 typed_splay_tree::remove.
967 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
969 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
970 -mginv and -mno-ginv to the assembler.
971 * config/mips/mips.opt (-mcrc): New option.
973 * doc/invoke.text (-mcrc): Document.
976 2018-06-15 Nick Clifton <nickc@redhat.com>
979 * tree.c (escaped_string): New class. Converts an unescaped
980 string into its escaped equivalent.
981 (warn_deprecated_use): Use the new class to convert the
982 deprecation message, if present.
983 (test_escaped_strings): New self test.
984 (test_c_tests): Add test_escaped_strings.
985 * doc/extend.texi (deprecated): Add a note that the
986 deprecation message is affected by the -fmessage-length
987 option, and that control characters will be escaped.
988 (#pragma GCC error): Document this pragma.
989 (#pragma GCC warning): Likewise.
990 * doc/invoke.texi (-fmessage-length): Document this option's
991 effect on the #warning and #error preprocessor directives and
992 the deprecated attribute.
994 2018-06-15 Richard Biener <rguenther@suse.de>
996 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
997 here, also noting vector size used.
998 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
999 size used in MSG_OPTIMIZED_LOCATIONS dump.
1000 (pass_slp_vectorize::execute): Adjust.
1002 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
1005 * config/arc/arc.c (arc_return_address_register): Fix
1008 2018-06-15 Richard Biener <rguenther@suse.de>
1011 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
1012 leave useless conversion stripping to force_gimple_operand_gsi.
1013 (gimplify_build2): Likewise.
1014 (gimplify_build1): Likewise.
1016 2018-06-15 Richard Biener <rguenther@suse.de>
1019 * tree-cfg.c (move_stmt_op): unshare invariant addresses
1020 before adjusting their block.
1022 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
1024 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
1025 multilibs for *-*-rtems*.
1026 * config/riscv/t-rtems: New file.
1028 2018-06-14 Jakub Jelinek <jakub@redhat.com>
1031 * match.pd ((A +- CST1) +- CST2): Punt if last resort
1032 unsigned_type_for returns NULL.
1035 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
1036 subregs of multi-word pseudos unless the float mode has word size.
1038 2018-06-14 Richard Biener <rguenther@suse.de>
1041 * tree-vect-generic.c (build_word_mode_vector_type): Remove
1042 duplicate and harmful type_hash_canon.
1043 * tree.c (type_hash_canon): Assert we didn't find ourselves.
1045 2018-06-14 Richard Biener <rguenther@suse.de>
1048 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
1051 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
1053 * config/rtems.h (STDINT_LONG32): Define.
1055 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
1056 Prachi Godbole <prachi.godbole@imgtec.com>
1058 * config/mips/mips-cpus.def: Define P6600.
1059 * config/mips/mips-tables.opt: Regenerate.
1060 * config/mips/mips.c (mips_ucbranch_type): New enum.
1061 (mips_rtx_cost_data): Add support for P6600.
1062 (mips_issue_rate): Likewise.
1063 (mips_multipass_dfa_lookahead): Likewise.
1064 (mips_avoid_hazard): Likewise.
1065 (mips_reorg_process_insns): Likewise.
1066 (mips_classify_branch_p6600): New function.
1067 * config/mips/mips.h (TUNE_P6600): New define.
1068 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
1069 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
1070 * config/mips/mips.md: Include p6600.md.
1071 (processor): Add p6600.
1072 * config/mips/p6600.md: New file.
1073 * doc/invoke.texi: Add p6600 to supported architectures.
1075 2018-06-13 Martin Sebor <msebor@redhat.com>
1077 PR tree-optimization/86114
1078 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
1080 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
1082 2018-06-13 Richard Biener <rguenther@suse.de>
1084 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
1085 Properly set vector type of the intermediate stmt.
1086 * tree-vect-stmts.c (vectorizable_operation): The destination
1087 var always has vectype_out type.
1089 2018-06-13 Jeff Law <law@redhat.com>
1091 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
1092 integer 0 for argument to print_rtl_with_bb.
1093 (rl78_reorg): Likewise.
1095 2018-06-13 David Malcolm <dmalcolm@redhat.com>
1097 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
1098 from rtx to rtx_insn *.
1099 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
1101 (add_sched_insns_for_speculation): Likewise for local "target",
1102 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1103 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1104 from rtx_insn ** to rtx_code_label **.
1105 (reorg_emit_nops): Likewise.
1106 (c6x_reorg): Likewise for local "call_labels".
1107 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1109 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1110 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1111 the loops over LABEL_REFS.
1112 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1114 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1115 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1116 (split_branches): Strengthen local "olabel" from rtx to
1117 rtx_insn *, adding a safe_as_a cast.
1118 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1120 (add_insn_after): Likewise for first two params.
1121 (add_insn_before): Likewise.
1122 (remove_insn): Likewise for param.
1123 (emit_pattern_before_noloc): Likewise for second and third params.
1124 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1125 (emit_call_insn_before_noloc): Likewise.
1126 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1128 (emit_barrier_before): Likewise.
1129 (emit_label_before): Strengthen "label" param from "rtx" to
1130 "rtx_code_label *". Strengthen "before" param from "rtx" to
1132 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1134 (emit_pattern_after_noloc): Likewise.
1135 (emit_insn_after_noloc): Likewise.
1136 (emit_jump_insn_after_noloc): Likewise.
1137 (emit_call_insn_after_noloc): Likewise.
1138 (emit_debug_insn_after_noloc): Likewise.
1139 (emit_barrier_after): Likewise.
1140 (emit_label_after): Likewise for both params.
1141 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1142 "loc" param from "int" to "location_t".
1143 (emit_insn_after_setloc): Likewise.
1144 (emit_jump_insn_after_setloc): Likewise.
1145 (emit_call_insn_after_setloc): Likewise.
1146 (emit_debug_insn_after_setloc): Likewise.
1147 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1148 "loc" param from "int" to "location_t".
1149 (emit_pattern_before): Convert NULL_RTX to NULL.
1150 (emit_insn_before_setloc): Convert "loc" param from "int" to
1152 (emit_jump_insn_before_setloc): Likewise.
1153 (emit_call_insn_before_setloc): Likewise.
1154 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1155 rtx_insn *. Convert "loc" param from "int" to "location_t".
1156 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1157 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1158 Convert 3rd param from "int" to "location_t".
1159 (emit_barrier_before, emit_barrier_after, next_real_insn):
1160 Strengthen param from rtx to rtx_insn *.
1161 (emit_label_before): Strengthen 1st param from "rtx" to
1162 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1164 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1165 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1166 Strengthen 2nd param from "rtx" to "rtx_insn *".
1167 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1168 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1169 Likewise. Convert 3rd param from "int" to "location_t".
1170 (emit_label_after): Strengthen 1st param from "rtx" to
1172 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1174 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1175 from "rtx" to "rtx_insn *".
1177 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1179 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1180 bodies streamed in with -Q.
1181 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1182 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1183 * gimple-streamer-out.c: Include gimple-pretty-print.h
1184 (output_bb): Dump stmts streamed.
1185 * lto-section-out.c: Include print-tree.h
1186 (lto_begin_section): Dump sections created.
1187 (lto_output_decl_index): Dump decl encoded.
1188 * lto-streamer-out.c: Include print-tree.h
1189 (create_output_block): Dump output block created.
1190 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1191 (output_function): Dump function output.
1192 (output_constructor): Dump constructor streamed.
1193 (write_global_stream): Output indexes encoded.
1194 (produce_asm_for_decls): Dump streams encoded.
1195 * lto-streamer.c (streamer_dump_file): New global var.
1196 * lto-streamer.h (streamer_dump_file): Declare.
1197 * passes.c (ipa_write_summaries): Initialize streamer dump.
1198 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1201 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1204 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1205 offsets for register save directives. Emit a second batch of save
1206 directives, if need be, when the function accesses prior frames.
1208 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1210 * config/arc/fpu.md (fmasf4): Force operand to register.
1211 (fnmasf4): Likewise.
1213 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1215 * config/arc/arc-protos.h (arc_pad_return): Remove.
1216 * config/arc/arc.c (machine_function): Remove force_short_suffix
1218 (arc_print_operand): Adjust printing of '&'.
1219 (arc_verify_short): Remove conditional printing of short suffix.
1220 (arc_final_prescan_insn): Remove reference to size_reason.
1221 (pad_return): New function.
1222 (arc_reorg): Call pad_return.
1223 (arc_pad_return): Remove.
1224 (arc_init_machine_status): Remove reference to force_short_suffix.
1225 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1226 (attr length): When attribute iscompact is true force to 2
1227 regardless; in the case of maybe check if we want to force the
1228 instruction to have 4 bytes length.
1229 (nopv): Change it to generate 4 byte long nop as well.
1230 (blockage): New pattern.
1231 (simple_return): Remove call to arc_pad_return.
1232 (p_return_i): Likewise.
1234 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1236 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1238 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1240 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1243 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1245 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1246 for ARC700 and ARCv2.
1248 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1251 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1252 operands[2] instead of operands[1].
1255 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1257 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1258 case, check whether the outer register overlaps an unallocatable
1259 register, not just whether it fits the required class.
1261 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1263 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1265 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1266 (memrefs_conflict_p): Likewise.
1267 (init_alias_analysis): Likewise.
1268 * cfgexpand.c (expand_debug_expr): Likewise.
1269 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1270 * cse.c (fold_rtx): Likewise.
1271 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1272 * expr.c (emit_block_move_hints): Likewise.
1273 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1274 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1275 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1276 (emit_group_store): Likewise.
1277 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1278 to read the PRE/POST_MODIFY increment.
1279 * calls.c (store_one_arg): Use strip_offset.
1280 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1282 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1284 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1285 (simplify_binary_operation_1): Extend CONST_INT handling to
1287 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1288 than a HOST_WIDE_INT.
1289 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1291 (adjust_mems, add_stores): Update accodingly.
1292 (vt_canonicalize_addr): Track polynomial offsets.
1293 (emit_note_insn_var_location): Likewise.
1294 (vt_add_function_parameter): Likewise.
1295 (vt_initialize): Likewise.
1297 2018-06-12 Jeff Law <law@redhat.com>
1299 * config.gcc (alpha*-*-freebsd*): Remove.
1300 * config/alpha/freebsd.h: Remove.
1302 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1305 * spellcheck-tree.c (levenshtein_distance): Rename to...
1306 (get_edit_distance): ...this, and update for underlying renaming.
1307 * spellcheck-tree.h (levenshtein_distance): Rename to...
1308 (get_edit_distance): ...this.
1309 * spellcheck.c (levenshtein_distance): Rename to...
1310 (get_edit_distance): ...this. Convert from Levenshtein distance
1311 to Damerau-Levenshtein distance by supporting transpositions of
1312 adjacent characters. Rename "v1" to "v_next" and "v0" to
1314 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1315 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1316 update for underlying renaming.
1317 (selftest::levenshtein_distance_unit_test): Rename to...
1318 (selftest::test_get_edit_distance_unit): ...this, and update for
1319 underlying renaming.
1320 (selftest::test_find_closest_string): Add example from PR 69968
1321 where transposition helps
1322 (selftest::test_metric_conditions): Update for renaming.
1323 (selftest::test_metric_conditions): Likewise.
1324 (selftest::spellcheck_c_tests): Likewise.
1325 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1326 (get_edit_distance): ...this.
1327 (best_match::consider): Update for renaming.
1329 2018-06-12 Martin Sebor <msebor@redhat.com>
1331 PR tree-optimization/85259
1332 * builtins.c (compute_objsize): Handle constant offsets.
1333 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1334 true iff a warning has been issued.
1335 * gimple.h (gimple_nonartificial_location): New function.
1336 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1337 gimple_nonartificial_location and handle -Wno-system-headers.
1338 (handle_builtin_stxncpy): Same.
1340 2018-06-12 Martin Sebor <msebor@redhat.com>
1343 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1345 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1347 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1348 BUILTIN_VEC_XST entries for pointer to double and long long.
1350 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1353 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1355 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1358 2018-06-12 Martin Liska <mliska@suse.cz>
1360 * doc/options.texi: Document IntegerRange.
1362 2018-06-12 Martin Liska <mliska@suse.cz>
1364 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1365 * opt-functions.awk: Handle Deprecated flag.
1366 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1368 (read_cmdline_option): Report warning for a deprecated option.
1369 * opts.h (struct cl_option): Add new field cl_deprecated.
1370 (CL_ERR_DEPRECATED): New.
1372 2018-06-12 Martin Liska <mliska@suse.cz>
1374 * doc/options.texi: Document Deprecated option flag.
1376 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1378 * config/arc/arc-arch.h (arc_extras): New enum.
1379 (arc_cpu_t):Add field extra.
1380 (arc_cpu_types): Consider the extras.
1381 * config/arc/arc-cpus.def: Add extras info.
1382 * config/arc/arc-opts.h (processor_type): Consider extra field.
1383 * config/arc/arc.c (arc_override_options): Handle extra field.
1385 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1387 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1388 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1390 * config/arc/arc.c (arc_init): Update pic warning.
1391 (irq_range): Update irq range parsing warnings.
1392 (arc_override_options): Update various warning messages.
1393 (arc_handle_aux_attribute): Likewise.
1395 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1397 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1399 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1401 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1402 compiler messages directives.
1404 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1406 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1407 * config/mips/mips-tables.opt: Regenerate.
1408 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1410 * doc/invoke.texi: Document -march=i6500.
1412 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1414 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1415 (i6400_gpmul): Add cpu_unit.
1416 (i6400_gpdiv): Likewise.
1417 (i6400_msa_add_d): Update reservations.
1418 (i6400_msa_int_add) Likewise.
1419 (i6400_msa_short_logic3) Likewise.
1420 (i6400_msa_short_logic2) Likewise.
1421 (i6400_msa_short_logic) Likewise.
1422 (i6400_msa_move) Likewise.
1423 (i6400_msa_cmp) Likewise.
1424 (i6400_msa_short_float2) Likewise.
1425 (i6400_msa_div_d) Likewise.
1426 (i6400_msa_long_logic1) Likewise.
1427 (i6400_msa_long_logic2) Likewise.
1428 (i6400_msa_mult) Likewise.
1429 (i6400_msa_long_float2) Likewise.
1430 (i6400_msa_long_float4) Likewise.
1431 (i6400_msa_long_float5) Likewise.
1432 (i6400_msa_long_float8) Likewise.
1433 (i6400_fpu_fadd): Include frint type.
1434 (i6400_fpu_store): New define_insn_reservation.
1435 (i6400_fpu_load): Likewise.
1436 (i6400_fpu_move): Likewise.
1437 (i6400_fpu_fcmp): Likewise.
1438 (i6400_fpu_fmadd): Likewise.
1439 (i6400_int_mult): Include imul3nc type and update reservation.
1440 (i6400_int_div): Include idiv3 type and update reservation.
1441 (i6400_int_load): Update to check type not move_type.
1442 (i6400_int_store): Likewise.
1443 (i6400_int_prefetch): Set zero latency.
1445 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1447 * gcc.c: Document new %@{...} sequence.
1448 (LINK_COMMAND_SPEC): Use it for the -L switches.
1449 (cpp_unique_options): Use it for the -I switches.
1450 (at_file_argbuf): New global variable.
1451 (in_at_file): Likewise.
1452 (alloc_args): Create at_file_argbuf.
1453 (clear_args): Truncate at_file_argbuf.
1454 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1455 (open_at_file): New function.
1456 (close_at_file): Likewise.
1457 (create_at_file): Delete.
1458 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1461 (validate_switches_from_spec): Deal with %@{...} sequence.
1462 (validate_switches): Likewise.
1463 (driver::finalize): Call clear_args.
1465 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1467 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1469 2018-06-11 Martin Sebor <msebor@redhat.com>
1471 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1472 (Wno-class-memaccess): Add @opindex.
1473 (Wno-templates, Wno-multiple-inheritance): Same.
1474 (Wno-virtual-inheritance, Wno-namespaces): Same.
1475 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1476 (Wno-format-overflow, Wno-format-truncation): Same.
1477 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1478 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1479 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1480 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1481 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1482 (Wno-misspelled-isr): Same.
1484 2018-06-11 Martin Sebor <msebor@redhat.com>
1486 * PR tree-optimization/86083
1487 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1489 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1491 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1493 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1496 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1497 on the correct operand.
1498 (*movdi_internal64): Ditto.
1500 2018-06-11 Martin Liska <mliska@suse.cz>
1502 PR tree-optimization/86089
1503 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1505 2018-06-11 Julia Koval <julia.koval@intel.com>
1507 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1508 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1509 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1511 2018-06-11 Olivier Hainque <hainque@adacore.com>
1513 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1514 for Ada with strict dwarf2.
1516 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1519 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1522 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1524 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1526 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1528 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1531 2018-06-08 Martin Liska <mliska@suse.cz>
1533 * tree-cfg.h (debug_function): Fix argument type to match
1536 2018-06-08 Martin Liska <mliska@suse.cz>
1538 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1539 Remove usage of MPX-related (and removed) fields.
1540 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1542 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1544 * cfg.c (debug): Use TDF_NONE rather than 0.
1545 * cfghooks.c (debug): Likewise.
1546 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1547 (struct dump_option_value_info): Convert to...
1548 (struct kv_pair): ...this template type.
1549 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1551 (optinfo_verbosity_options): Likewise.
1552 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1554 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1555 than int for "optgroup_flags" param.
1556 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1558 (dump_dec): Likewise.
1559 (dump_finish): Use TDF_NONE rather than 0.
1560 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1561 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1562 than 0. Update for change to option_ptr.
1563 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1564 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1565 0. Update for changes to optinfo_verbosity_options and
1567 (opt_info_switch_p): Convert optgroup_flags from int to
1569 (dump_basic_block): Use dump_flags_t rather than int
1570 for "dump_kind" param.
1571 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1572 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1573 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1574 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1575 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1576 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1577 TDF_NONE): Convert from macros to...
1578 (enum dump_flag): ...this new enum.
1579 (dump_flags_t): Update to use enum.
1580 (operator|, operator&, operator~, operator|=, operator&=):
1581 Implement for dump_flags_t.
1582 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1583 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1584 Convert from macros to...
1585 (enum optgroup_flag): ...this new enum.
1586 (optgroup_flags_t): New typedef.
1587 (operator|, operator|=): Implement for optgroup_flags_t.
1588 (struct dump_file_info): Convert field "alt_flags" to
1589 dump_flags_t. Convert field "optgroup_flags" to
1591 (dump_basic_block): Use dump_flags_t rather than int for param.
1592 (dump_generic_expr_loc): Likewise.
1593 (dump_dec): Likewise.
1594 (dump_register): Convert param "optgroup_flags" to
1596 (opt_info_enable_passes): Likewise.
1597 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1599 * gimple-pretty-print.c (debug): Likewise.
1600 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1601 (merged_store_group::apply_stores): Likewise.
1602 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1603 * gimple.c (verify_gimple_pp): Likewise.
1604 * graphite-poly.c (print_pbb_body): Likewise.
1605 * passes.c (pass_manager::register_one_dump_file): Convert
1606 local "optgroup_flags" to optgroup_flags_t.
1607 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1609 (debug_body): Likewise.
1610 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1611 to optgroup_flags_t.
1612 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1614 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1615 (convert_mult_to_fma): Likewise.
1616 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1617 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1618 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1619 "dump_kind" to dump_flags_t.
1621 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1623 * config/rs6000/rs6000.c (min, max): Delete.
1625 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1627 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1628 -mabi=spe and -mabi=no-spe.
1630 2018-06-08 Martin Liska <mliska@suse.cz>
1632 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1633 where we expect an existing summary.
1635 2018-06-08 Martin Liska <mliska@suse.cz>
1637 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1638 * ipa-inline.h (estimate_edge_growth): Likewise.
1640 2018-06-08 Martin Liska <mliska@suse.cz>
1642 * cgraph.c (function_version_hasher::hash): Use
1643 cgraph_node::get_uid ().
1644 (function_version_hasher::equal):
1645 * cgraph.h (cgraph_node::get_uid): New method.
1646 * ipa-inline.c (update_caller_keys): Use
1647 cgraph_node::get_uid ().
1648 (update_callee_keys): Likewise.
1649 * ipa-utils.c (searchc): Likewise.
1650 (ipa_reduced_postorder): Likewise.
1651 * lto-cgraph.c (input_node): Likewise.
1652 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1653 * symbol-summary.h (symtab_insertion): Likewise.
1654 (symtab_removal): Likewise.
1655 (symtab_duplication): Likewise.
1656 * tree-pretty-print.c (dump_function_header): Likewise.
1657 * tree-sra.c (convert_callers_for_node): Likewise.
1659 2018-06-08 Martin Liska <mliska@suse.cz>
1661 * cgraph.c (symbol_table::create_edge): Always assign a new
1663 (symbol_table::free_edge): Do not recycle numbers.
1664 * cgraph.h (cgraph_edge::get): New method.
1665 * symbol-summary.h (symtab_removal): Use it.
1666 (symtab_duplication): Likewise.
1667 (call_summary::hashable_uid): Remove.
1669 2018-06-08 Martin Liska <mliska@suse.cz>
1671 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1672 (initialize_growth_caches): Remove.
1673 (free_growth_caches): Likewise.
1674 (do_estimate_edge_time): Use edge_growth_cache.
1675 (do_estimate_edge_size): Likewise.
1676 (do_estimate_edge_hints): Likewise.
1677 * ipa-inline.c (reset_edge_caches): Likewise.
1678 (recursive_inlining): Likewise.
1679 (inline_small_functions): Likewise.
1680 * ipa-inline.h (initialize_growth_caches): Remove.
1681 (estimate_edge_size): Likewise.
1682 (estimate_edge_time): Likewise.
1683 (estimate_edge_hints): Likewise.
1684 (reset_edge_growth_cache): Likewise.
1685 * symbol-summary.h (call_summary::remove): New method.
1687 2018-06-08 Martin Liska <mliska@suse.cz>
1689 * ipa-cp.c (class edge_clone_summary): New summary.
1690 (grow_edge_clone_vectors): Remove.
1691 (ipcp_edge_duplication_hook): Remove.
1692 (class edge_clone_summary_t): New call_summary class.
1693 (ipcp_edge_removal_hook): Remove.
1694 (edge_clone_summary_t::duplicate): New function.
1695 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1696 (create_specialized_node): Likewise.
1697 (ipcp_driver): Initialize edge_clone_summaries and do not
1700 2018-06-08 Martin Liska <mliska@suse.cz>
1702 * symbol-summary.h (get): New function.
1703 (call_summary::m_initialize_when_cloning): New class member.
1705 2018-06-08 Martin Liska <mliska@suse.cz>
1707 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1708 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1709 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1710 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1711 (remove_cgraph_node_from_order): Use the removed_nodes set.
1712 (do_per_function_toporder): Likwise.
1713 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1714 instead of summary_uid.
1715 (symtab_removal): Likewise.
1716 (symtab_duplication): Likewise.
1718 2018-06-08 Martin Liska <mliska@suse.cz>
1720 * ipa-cp.c (ipcp_store_bits_results): Use
1721 ipcp_transformation_sum.
1722 (ipcp_store_vr_results): Likewise.
1723 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1725 (ipcp_transformation_initialize): ... this.
1726 (ipa_set_node_agg_value_chain):
1727 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1728 (write_ipcp_transformation_info): Likewise.
1729 (read_ipcp_transformation_info): Likewise.
1730 (ipcp_update_bits): Likewise.
1731 (ipcp_update_vr): Likewise.
1732 (ipcp_transform_function): Likewise.
1733 * ipa-prop.h: Rename ipcp_transformation_summary to
1734 ipcp_transformation.
1735 (class ipcp_transformation_t): New function summary.
1736 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1737 (ipa_get_agg_replacements_for_node): Likewise.
1739 2018-06-08 Martin Liska <mliska@suse.cz>
1741 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1743 (class funct_state_summary_t): New function_summary class.
1744 (has_function_state): Remove.
1745 (get_function_state): Likewise.
1746 (set_function_state): Likewise.
1747 (add_new_function): Likewise.
1748 (funct_state_summary_t::insert): New function.
1749 (duplicate_node_data): Remove.
1750 (remove_node_data): Remove.
1751 (funct_state_summary_t::duplicate): New function.
1752 (register_hooks): Create new funct_state_summaries.
1753 (pure_const_generate_summary): Use it.
1754 (pure_const_write_summary): Likewise.
1755 (pure_const_read_summary): Likewise.
1756 (propagate_pure_const): Likewise.
1757 (propagate_nothrow): Likewise.
1758 (dump_malloc_lattice): Likewise.
1759 (propagate_malloc): Likewise.
1760 (execute): Do not register hooks, just remove summary
1762 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1765 2018-06-08 Martin Liska <mliska@suse.cz>
1767 * ipa-reference.c (remove_node_data): Remove.
1768 (duplicate_node_data): Likewise.
1769 (class ipa_ref_var_info_summary_t): New class.
1770 (class ipa_ref_opt_summary_t): Likewise.
1771 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1772 (get_reference_optimization_summary): Use
1773 ipa_ref_opt_sum_summaries.
1774 (set_reference_vars_info): Remove.
1775 (set_reference_optimization_summary): Likewise.
1776 (ipa_init): Create summaries.
1777 (init_function_info): Use function summary.
1778 (ipa_ref_opt_summary_t::duplicate): New function.
1779 (ipa_ref_opt_summary_t::remove): New function.
1780 (get_read_write_all_from_node): Fix GNU coding style.
1781 (propagate): Use function summary.
1782 (write_node_summary_p): Fix GNU coding style.
1783 (stream_out_bitmap): Likewise.
1784 (ipa_reference_read_optimization_summary): Use function summary.
1785 (ipa_reference_c_finalize): Do not release hooks.
1787 2018-06-08 Martin Liska <mliska@suse.cz>
1789 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1790 (analyze_function_body): Extract multiple calls of get_create.
1791 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1792 * ipa-inline.c (recursive_inlining): Use ::get method.
1793 * ipa-inline.h (estimate_edge_growth): Likewise.
1795 2018-06-08 Martin Liska <mliska@suse.cz>
1797 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1799 (hsa_function_summary::hsa_function_summary): Use the new enum
1801 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1802 * hsa-gen.c (hsa_get_host_function): Likewise.
1803 (get_brig_function_name): Likewise.
1804 * ipa-hsa.c (process_hsa_functions): Likewise.
1805 (ipa_hsa_write_summary): Likewise.
1806 * symbol-summary.h (symtab_duplication): Use ::get function/
1807 (get): New function.
1809 2018-06-08 Martin Liska <mliska@suse.cz>
1811 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1813 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1814 (hsa_register_kernel): Likewise.
1815 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1816 * hsa-gen.c (hsa_get_host_function): Likewise.
1817 (get_brig_function_name): Likewise.
1818 (generate_hsa): Likewise.
1819 (pass_gen_hsail::execute): Likewise.
1820 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1821 (devirtualization_time_bonus): Likewise.
1822 (ipcp_propagate_stage): Likewise.
1823 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1824 (edge_set_predicate): Likewise.
1825 (evaluate_conditions_for_known_args): Likewise.
1826 (evaluate_properties_for_edge): Likewise.
1827 (ipa_fn_summary::reset): Likewise.
1828 (ipa_fn_summary_t::duplicate): Likewise.
1829 (dump_ipa_call_summary): Likewise.
1830 (ipa_dump_fn_summary): Likewise.
1831 (analyze_function_body): Likewise.
1832 (compute_fn_summary): Likewise.
1833 (estimate_edge_devirt_benefit): Likewise.
1834 (estimate_edge_size_and_time): Likewise.
1835 (estimate_calls_size_and_time): Likewise.
1836 (estimate_node_size_and_time): Likewise.
1837 (inline_update_callee_summaries): Likewise.
1838 (remap_edge_change_prob): Likewise.
1839 (remap_edge_summaries): Likewise.
1840 (ipa_merge_fn_summary_after_inlining): Likewise.
1841 (ipa_update_overall_fn_summary): Likewise.
1842 (read_ipa_call_summary): Likewise.
1843 (inline_read_section): Likewise.
1844 (write_ipa_call_summary): Likewise.
1845 (ipa_fn_summary_write): Likewise.
1846 (ipa_free_fn_summary): Likewise.
1847 * ipa-hsa.c (process_hsa_functions): Likewise.
1848 (ipa_hsa_write_summary): Likewise.
1849 (ipa_hsa_read_section): Likewise.
1850 * ipa-icf.c (sem_function::merge): Likewise.
1851 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1852 (do_estimate_edge_time): Likewise.
1853 (estimate_size_after_inlining): Likewise.
1854 (estimate_growth): Likewise.
1855 (growth_likely_positive): Likewise.
1856 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1857 (inline_call): Likewise.
1858 * ipa-inline.c (caller_growth_limits): Likewise.
1859 (can_inline_edge_p): Likewise.
1860 (can_inline_edge_by_limits_p): Likewise.
1861 (compute_uninlined_call_time): Likewise.
1862 (compute_inlined_call_time): Likewise.
1863 (want_inline_small_function_p): Likewise.
1864 (edge_badness): Likewise.
1865 (update_caller_keys): Likewise.
1866 (update_callee_keys): Likewise.
1867 (recursive_inlining): Likewise.
1868 (inline_small_functions): Likewise.
1869 (inline_to_all_callers_1): Likewise.
1870 (dump_overall_stats): Likewise.
1871 (early_inline_small_functions): Likewise.
1872 (early_inliner): Likewise.
1873 * ipa-inline.h (estimate_edge_growth): Likewise.
1874 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1875 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1876 * ipa-prop.h (IPA_NODE_REF): Likewise.
1877 (IPA_EDGE_REF): Likewise.
1878 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1879 (propagate_malloc): Likewise.
1880 * ipa-split.c (execute_split_functions): Likewise.
1881 * symbol-summary.h: Rename get to get_create.
1883 (get_create): Likewise.
1884 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1886 2018-06-08 Martin Liska <mliska@suse.cz>
1888 * symbol-summary.h (release): Move definition out of class
1890 (symtab_removal): Likewise.
1891 (symtab_duplication): Likewise.
1893 2018-06-08 Martin Liska <mliska@suse.cz>
1895 * symbol-summary.h (function_summary): Move constructor
1896 implementation out of class declaration.
1897 (release): Likewise.
1898 (symtab_insertion): Likewise.
1899 (symtab_removal): Likewise.
1900 (symtab_duplication): Likewise.
1903 2018-06-08 Martin Liska <mliska@suse.cz>
1905 * Makefile.in: Remove support for MPX (macros, related functions,
1906 fields in cgraph_node, ...).
1907 * builtin-types.def (BT_BND): Likewise.
1908 (BT_FN_BND_CONST_PTR): Likewise.
1909 (BT_FN_CONST_PTR_BND): Likewise.
1910 (BT_FN_VOID_PTR_BND): Likewise.
1911 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1912 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1913 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1914 (expand_builtin_mempcpy_with_bounds): Likewise.
1915 (expand_builtin_memset_with_bounds): Likewise.
1916 (expand_builtin_memset_args): Likewise.
1917 (std_expand_builtin_va_start): Likewise.
1918 (expand_builtin): Likewise.
1919 (expand_builtin_with_bounds): Likewise.
1920 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1921 (DEF_LIB_BUILTIN_CHKP): Likewise.
1922 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1923 (DEF_CHKP_BUILTIN): Likewise.
1924 (BUILT_IN_MEMCPY): Likewise.
1925 (BUILT_IN_MEMMOVE): Likewise.
1926 (BUILT_IN_MEMPCPY): Likewise.
1927 (BUILT_IN_MEMSET): Likewise.
1928 (BUILT_IN_STPCPY): Likewise.
1929 (BUILT_IN_STRCAT): Likewise.
1930 (BUILT_IN_STRCHR): Likewise.
1931 (BUILT_IN_STRCPY): Likewise.
1932 (BUILT_IN_STRLEN): Likewise.
1933 (BUILT_IN_MEMCPY_CHK): Likewise.
1934 (BUILT_IN_MEMMOVE_CHK): Likewise.
1935 (BUILT_IN_MEMPCPY_CHK): Likewise.
1936 (BUILT_IN_MEMSET_CHK): Likewise.
1937 (BUILT_IN_STPCPY_CHK): Likewise.
1938 (BUILT_IN_STRCAT_CHK): Likewise.
1939 (BUILT_IN_STRCPY_CHK): Likewise.
1940 * calls.c (store_bounds): Likewise.
1941 (emit_call_1): Likewise.
1942 (special_function_p): Likewise.
1943 (maybe_warn_nonstring_arg): Likewise.
1944 (initialize_argument_information): Likewise.
1945 (finalize_must_preallocate): Likewise.
1946 (compute_argument_addresses): Likewise.
1947 (expand_call): Likewise.
1948 * cfgexpand.c (expand_call_stmt): Likewise.
1949 (expand_return): Likewise.
1950 (expand_gimple_stmt_1): Likewise.
1951 (pass_expand::execute): Likewise.
1952 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1953 (cgraph_node::remove): Likewise.
1954 (cgraph_node::dump): Likewise.
1955 (cgraph_node::verify_node): Likewise.
1956 * cgraph.h (chkp_function_instrumented_p): Likewise.
1957 (symtab_node::get_alias_target): Likewise.
1958 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1959 (cgraph_local_p): Likewise.
1960 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1961 (cgraph_edge::rebuild_references): Likewise.
1962 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1963 (walk_polymorphic_call_targets): Likewise.
1964 (cgraph_node::expand_thunk): Likewise.
1965 (symbol_table::output_weakrefs): Likewise.
1966 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1967 (ix86_handle_option): Likewise.
1968 * config/i386/constraints.md: Likewise.
1969 * config/i386/i386-builtin-types.def (BND): Likewise.
1973 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1974 (BDESC_FIRST): Likewise.
1976 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1977 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1978 * config/i386/i386.c (enum reg_class): Likewise.
1979 (ix86_target_string): Likewise.
1980 (ix86_option_override_internal): Likewise.
1981 (ix86_conditional_register_usage): Likewise.
1982 (ix86_valid_target_attribute_inner_p): Likewise.
1983 (ix86_set_indirect_branch_type): Likewise.
1984 (ix86_set_current_function): Likewise.
1985 (ix86_function_arg_regno_p): Likewise.
1986 (init_cumulative_args): Likewise.
1987 (ix86_function_arg_advance): Likewise.
1988 (ix86_function_arg): Likewise.
1989 (ix86_pass_by_reference): Likewise.
1990 (ix86_function_value_regno_p): Likewise.
1991 (ix86_function_value_1): Likewise.
1992 (ix86_function_value_bounds): Likewise.
1993 (ix86_return_in_memory): Likewise.
1994 (ix86_setup_incoming_vararg_bounds): Likewise.
1995 (ix86_va_start): Likewise.
1996 (indirect_thunk_need_prefix): Likewise.
1997 (print_reg): Likewise.
1998 (ix86_print_operand): Likewise.
1999 (ix86_expand_call): Likewise.
2000 (ix86_output_function_return): Likewise.
2001 (reg_encoded_number): Likewise.
2002 (BDESC_VERIFYS): Likewise.
2003 (ix86_init_mpx_builtins): Likewise.
2004 (ix86_init_builtins): Likewise.
2005 (ix86_emit_cmove): Likewise.
2006 (ix86_emit_move_max): Likewise.
2007 (ix86_expand_builtin): Likewise.
2008 (ix86_builtin_mpx_function): Likewise.
2009 (ix86_get_arg_address_for_bt): Likewise.
2010 (ix86_load_bounds): Likewise.
2011 (ix86_store_bounds): Likewise.
2012 (ix86_load_returned_bounds): Likewise.
2013 (ix86_store_returned_bounds): Likewise.
2014 (ix86_class_likely_spilled_p): Likewise.
2015 (ix86_hard_regno_mode_ok): Likewise.
2016 (x86_order_regs_for_local_alloc): Likewise.
2017 (ix86_mitigate_rop): Likewise.
2018 (ix86_bnd_prefixed_insn_p): Likewise.
2019 (ix86_mpx_bound_mode): Likewise.
2020 (ix86_make_bounds_constant): Likewise.
2021 (ix86_initialize_bounds): Likewise.
2022 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
2023 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
2024 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
2025 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
2026 (TARGET_CHKP_BOUND_MODE): Likewise.
2027 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
2028 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
2029 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
2030 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
2031 * config/i386/i386.h (TARGET_MPX): Likewise.
2032 (TARGET_MPX_P): Likewise.
2033 (VALID_BND_REG_MODE): Likewise.
2034 (FIRST_BND_REG): Likewise.
2035 (LAST_BND_REG): Likewise.
2036 (enum reg_class): Likewise.
2037 (BND_REG_P): Likewise.
2038 (BND_REGNO_P): Likewise.
2039 (BNDmode): Likewise.
2040 (ADJUST_INSN_LENGTH): Likewise.
2041 * config/i386/i386.md: Likewise.
2042 * config/i386/i386.opt: Likewise.
2043 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
2044 (defined): Likewise.
2045 (LINK_MPX): Likewise.
2046 (MPX_SPEC): Likewise.
2047 (LIBMPX_SPEC): Likewise.
2048 (LIBMPXWRAPPERS_SPEC): Likewise.
2049 (CHKP_SPEC): Likewise.
2050 * config/i386/predicates.md: Likewise.
2051 * dbxout.c (dbxout_type): Likewise.
2052 * doc/extend.texi: Likewise.
2053 * doc/invoke.texi: Likewise.
2054 * doc/md.texi: Likewise.
2055 * doc/tm.texi: Likewise.
2056 * doc/tm.texi.in: Likewise.
2057 * dwarf2out.c (is_base_type): Likewise.
2058 (gen_formal_types_die): Likewise.
2059 (gen_subprogram_die): Likewise.
2060 (gen_type_die_with_usage): Likewise.
2061 (gen_decl_die): Likewise.
2062 (dwarf2out_late_global_decl): Likewise.
2063 * expr.c (expand_assignment): Likewise.
2064 (emit_storent_insn): Likewise.
2065 (store_expr_with_bounds): Likewise.
2066 (store_expr): Likewise.
2067 (expand_expr_real_1): Likewise.
2068 * expr.h (store_expr_with_bounds): Likewise.
2069 * function.c (use_register_for_decl): Likewise.
2070 (struct bounds_parm_data): Likewise.
2071 (assign_parms_augmented_arg_list): Likewise.
2072 (assign_parm_find_entry_rtl): Likewise.
2073 (assign_parm_is_stack_parm): Likewise.
2074 (assign_parm_load_bounds): Likewise.
2075 (assign_bounds): Likewise.
2076 (assign_parms): Likewise.
2077 (expand_function_start): Likewise.
2078 * gcc.c (CHKP_SPEC): Likewise.
2079 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
2080 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
2081 (wrestrict_dom_walker::check_call): Likewise.
2082 * gimple.c (gimple_build_call_from_tree): Likewise.
2083 * gimple.h (enum gf_mask): Likewise.
2084 (gimple_call_with_bounds_p): Likewise.
2085 (gimple_call_set_with_bounds): Likewise.
2086 * gimplify.c (gimplify_init_constructor): Likewise.
2087 * ipa-cp.c (initialize_node_lattices): Likewise.
2088 (propagate_constants_across_call): Likewise.
2089 (find_more_scalar_values_for_callers_subset): Likewise.
2090 * ipa-hsa.c (process_hsa_functions): Likewise.
2091 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
2092 * ipa-icf.c (sem_function::merge): Likewise.
2093 * ipa-inline.c (early_inliner): Likewise.
2094 * ipa-pure-const.c (warn_function_noreturn): Likewise.
2095 (warn_function_cold): Likewise.
2096 (propagate_pure_const): Likewise.
2097 * ipa-ref.h (enum GTY): Likewise.
2098 * ipa-split.c (find_retbnd): Likewise.
2099 (consider_split): Likewise.
2100 (split_function): Likewise.
2101 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2102 * ipa.c (walk_polymorphic_call_targets): Likewise.
2103 (symbol_table::remove_unreachable_nodes): Likewise.
2104 (process_references): Likewise.
2105 (cgraph_build_static_cdtor_1): Likewise.
2106 * lto-cgraph.c (lto_output_node): Likewise.
2107 (output_refs): Likewise.
2108 (compute_ltrans_boundary): Likewise.
2109 (input_overwrite_node): Likewise.
2110 (input_node): Likewise.
2111 (input_cgraph_1): Likewise.
2112 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2113 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2114 (class pass_chkp_instrumentation_passes): Likewise.
2115 (make_pass_chkp_instrumentation_passes): Likewise.
2116 * passes.def: Likewise.
2117 * rtl.h (struct GTY): Likewise.
2118 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2119 * stor-layout.c (layout_type): Likewise.
2120 * symtab.c: Likewise.
2121 * target.def: Likewise.
2122 * targhooks.c (default_chkp_bound_type): Likewise.
2123 (default_chkp_bound_mode): Likewise.
2124 (default_builtin_chkp_function): Likewise.
2125 (default_chkp_function_value_bounds): Likewise.
2126 (default_chkp_make_bounds_constant): Likewise.
2127 (default_chkp_initialize_bounds): Likewise.
2128 * targhooks.h (default_chkp_bound_type): Likewise.
2129 (default_chkp_bound_mode): Likewise.
2130 (default_builtin_chkp_function): Likewise.
2131 (default_chkp_function_value_bounds): Likewise.
2132 (default_chkp_make_bounds_constant): Likewise.
2133 (default_chkp_initialize_bounds): Likewise.
2134 * toplev.c (compile_file): Likewise.
2135 (process_options): Likewise.
2136 * tree-core.h (DEF_BUILTIN): Likewise.
2137 (DEF_BUILTIN_CHKP): Likewise.
2138 * tree-inline.c (declare_return_variable): Likewise.
2139 (remap_gimple_stmt): Likewise.
2140 (copy_bb): Likewise.
2141 (initialize_inlined_parameters): Likewise.
2142 (expand_call_inline): Likewise.
2143 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2144 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2145 (make_pass_ipa_chkp_produce_thunks): Likewise.
2146 (make_pass_chkp): Likewise.
2147 (make_pass_chkp_opt): Likewise.
2148 (make_pass_chkp_instrumentation_passes): Likewise.
2149 * tree-pretty-print.c (dump_generic_node): Likewise.
2150 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2151 * tree-ssa-dce.c (propagate_necessity): Likewise.
2152 (eliminate_unnecessary_stmts): Likewise.
2153 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2154 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2155 * tree-ssa-sccvn.h: Likewise.
2156 * tree-ssa-strlen.c (get_string_length): Likewise.
2157 (valid_builtin_call): Likewise.
2158 (adjust_last_stmt): Likewise.
2159 (handle_builtin_strchr): Likewise.
2160 (handle_builtin_strcpy): Likewise.
2161 (handle_builtin_stxncpy): Likewise.
2162 (handle_builtin_memcpy): Likewise.
2163 (handle_builtin_strcat): Likewise.
2164 (strlen_check_and_optimize_stmt): Likewise.
2165 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2166 * tree-streamer-in.c: Likewise.
2167 * tree-streamer.c (record_common_node): Likewise.
2168 * tree.c (tree_code_size): Likewise.
2169 (wide_int_to_tree_1): Likewise.
2170 (type_contains_placeholder_1): Likewise.
2171 (build_common_tree_nodes): Likewise.
2172 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2173 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2174 (POINTER_BOUNDS_P): Likewise.
2175 (BOUNDED_TYPE_P): Likewise.
2176 (BOUNDED_P): Likewise.
2177 (CALL_WITH_BOUNDS_P): Likewise.
2178 (pointer_bounds_type_node): Likewise.
2179 * value-prof.c (gimple_ic): Likewise.
2180 * var-tracking.c (vt_add_function_parameters): Likewise.
2181 * varasm.c (make_decl_rtl): Likewise.
2182 (assemble_start_function): Likewise.
2183 (output_constant): Likewise.
2184 (maybe_assemble_visibility): Likewise.
2185 * varpool.c (ctor_for_folding): Likewise.
2186 * chkp-builtins.def: Remove.
2187 * ipa-chkp.c: Remove.
2188 * ipa-chkp.h: Remove.
2189 * rtl-chkp.c: Remove.
2190 * rtl-chkp.h: Remove.
2191 * tree-chkp-opt.c: Remove.
2192 * tree-chkp.c: Remove.
2193 * tree-chkp.h: Remove.
2195 2018-06-07 Carl Love <cel@us.ibm.com>
2197 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2198 vextract_fp_from_shortl): Add BE support.
2200 2018-06-07 Paul Koning <ni1d@arrl.net>
2202 * compare-elim.c (try_merge_compare): Don't merge compare if
2203 address contains a side effect.
2204 (try_eliminate_compare): Likewise.
2206 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2208 * config.gcc: Support "tremont".
2209 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2210 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2212 * config/i386/i386.c (m_TREMONT): Define.
2213 (processor_target_table): Add "tremont".
2214 (PTA_TREMONT): Define.
2215 (ix86_lea_outperforms): Add TARGET_TREMONT.
2216 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2217 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2218 and M_INTEL_GOLDMONT_PLUS.
2219 (fold_builtin_cpu): Add "tremont".
2220 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2221 (ix86_option_override_internal): Add "tremont".
2222 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2223 (processor_type): Add PROCESSOR_TREMONT.
2224 * config/i386/x86-tune.def: Add m_TREMONT.
2225 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2227 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2229 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2230 symbol defined for msp430i* devices to be lower case.
2232 gcc/testsuite/gcc.target/msp430/
2233 * msp430i-device-symbol.c: New test.
2234 * msp430f-device-symbol.c: New test.
2235 * msp430.h: New test header file.
2237 2018-06-07 Richard Biener <rguenther@suse.de>
2239 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2240 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2241 Properly wrap signed arithmetic if overflow wraps.
2243 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2245 PR tree-optimization/69615
2246 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2247 of a cast from a same precision integral SSA_NAME in a bb dominated
2248 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2249 cast to utype if rhs2 has already a compatible type.
2251 2018-06-07 Richard Biener <rguenther@suse.de>
2253 PR tree-optimization/85935
2254 * graphite-scop-detection.c (find_params_in_bb): Analyze
2255 condition operands with respect to the correct loop. Assert
2256 the analysis doesn't fail.
2258 2018-06-04 Carl Love <cel@us.ibm.com>
2260 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2261 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2262 as it is slightly cheaper.
2263 (first_match_or_eos_index_<mode>):
2264 Calculate index using natural element order.
2265 (first_match_index_<mode>):
2266 Calculate index using natural element order.
2267 (first_match_or_eos_index_<mode>):
2268 Calculate index using natural order.
2269 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2270 for BE and LE modes.
2271 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2272 P9V_BUILTIN_VCLZLSBB_V16QI.
2273 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2276 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2278 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2279 indentation and line wrap for many prototypes. Add missing
2280 @smallexample directives around block of prototypes for vec_xl and
2283 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2285 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2286 track if we pass or return IEEE 128-bit floating point.
2287 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2288 C++ mangling that is compatible with GCC 8.1.
2289 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2290 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2291 floating point types.
2292 (rs6000_function_arg_advance_1): Likewise.
2293 (rs6000_mangle_type): Optionally generate mangled names that match
2294 what GCC 8.1 generated for IEEE 128-bit floating point types.
2295 (rs6000_globalize_decl_name): If we have an external function that
2296 passes or returns IEEE 128-bit types, generate a weak reference
2297 from the mangled name used in GCC 8.1 to the current mangled
2299 (rs6000_init_builtins): Make __ibm128 use the long double type if
2300 long double is IBM extended double. Make __float128 use the long
2301 double type if long double is IEEE 128-bit.
2304 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2305 macro for __ibm128 built-in functions.
2306 (PACK_IF): Add __ibm128 pack/unpack functions.
2307 (UNPACK_IF): Likewise.
2308 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2309 enable long double built-in functions if long double is IEEE
2310 128-bit floating point.
2311 (rs6000_invalid_builtin): Update long double built-in function
2313 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2314 functions, adjust the built-in function to use the long double
2315 built-in function if __ibm128 and long double are the same type.
2316 * doc/extend.texi (PowerPC builtins): Update documention for
2317 __builtin_{,un}pack_longdouble. Add documentation for
2318 __builtin_{,un}pack_ibm128.
2320 2018-06-06 Jim Wilson <jimw@sifive.com>
2322 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2323 (struct machine_function): New field interrupt_mode.
2324 (riscv_handle_type_attribute): New function. Add forward declaration.
2325 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2326 (riscv_expand_epilogue): Check interrupt_mode field.
2327 (riscv_set_current_function): Check interrupt attribute args and
2328 set interrupt_mode field.
2329 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2330 (riscv_sret, riscv_uret): New.
2331 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2332 new arguments to interrupt attribute.
2334 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2337 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2338 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2340 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2342 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2343 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2344 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2345 several redundant entries.
2347 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2349 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2350 type from "rtx" to "rtx_insn *".
2351 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2352 for local "call_insn", removing cast.
2353 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2355 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2357 PR tree-optimization/86066
2358 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2359 for BIT_INSERT_EXPR stores.
2361 2018-06-06 Richard Biener <rguenther@suse.de>
2363 PR tree-optimization/86062
2364 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2365 component refs ontop
2366 of to be offsetted base.
2368 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2370 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2371 to be static and remove check on interrupt attribute name.
2373 gcc/testsuite/gcc.target/msp430/
2374 * function-attributes-4.c: New test.
2375 * static-interrupts.c: New test.
2377 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2379 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2380 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2382 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2385 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2387 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2388 Remove second argument, change how error is called.
2389 (aarch64_layout_arg): Remove second argument from
2390 aarch64_err_no_fpadvsimd call.
2391 (aarch64_init_cumulative_args): Ditto.
2392 (aarch64_gimplify_va_arg_expr): Ditto.
2393 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2395 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2397 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2398 (*simple_return_indirect_internal<mode>): Rename from
2399 simple_return_indirect_internal. Use W mode iterator.
2400 (rstorssp): New expander.
2401 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2402 (clrssbsy): New expander.
2403 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2405 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2407 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2409 (cmse_check_pointed_object): Likewise.
2411 2018-06-05 Martin Liska <mliska@suse.cz>
2413 PR gcov-profile/47618
2414 * doc/invoke.texi: Document how -fprofile-dir format
2417 2018-06-05 Richard Biener <rguenther@suse.de>
2419 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2420 removal pretend DOM info isn't available so we do not update
2421 it and only remove edges, not dominated blocks. Actually free
2422 DOM info in case we removed something. Remove unreachable blocks.
2423 (mfb_keep_latches): Work with either DOM info or marked backedges.
2424 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2425 first. Mark backedges if DOM info isn't available.
2426 (Re-)compute DOM info after cleanup_control_flow_pre.
2428 2018-06-05 Richard Biener <rguenther@suse.de>
2430 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2431 (locus_discrim_hasher::hash): Adjust.
2432 (locus_discrim_hasher::equal): Likewise.
2433 (next_discriminator_for_locus): Work on line directly.
2434 (same_line_p): Pass in expanded locus1 as well.
2435 (assign_discriminators): Avoid redundant location expansions.
2437 2018-06-05 Richard Biener <rguenther@suse.de>
2439 PR tree-optimization/86046
2440 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2441 if required after clearing TREE_ADDRESSABLE.
2443 2018-06-05 Richard Biener <rguenther@suse.de>
2445 PR tree-optimization/86047
2446 * tree-ssa-loop.c (for_each_index): Glob handling of all
2447 decls and constants and really handle all of them.
2449 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2452 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2453 qualifier_void_pointer and qualifier_const_void_pointer.
2454 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2455 (arm_init_builtins): Handle the above.
2456 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2457 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2458 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2461 2018-06-05 Martin Liska <mliska@suse.cz>
2463 * auto-profile.c (read_autofdo_file): Do not use
2464 gcov_ctr_summary struct.
2465 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2466 * coverage.c (struct counts_entry): Likewise.
2467 (read_counts_file): Read just single summary entry.
2468 (get_coverage_counts): Use gcov_summary struct.
2469 * coverage.h (get_coverage_counts): Likewise.
2470 * gcov-dump.c (dump_working_sets): Likewise.
2471 (tag_summary): Dump just single summary.
2472 * gcov-io.c (gcov_write_summary): Write just histogram
2474 (gcov_read_summary): Read just single summary.
2475 (compute_working_sets): Use gcov_summary struct.
2476 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2477 of GCOV_COUNTERS_SUMMABLE.
2478 (GCOV_COUNTERS_SUMMABLE): Remove.
2479 (GCOV_FIRST_VALUE_COUNTER): Replace with
2480 GCOV_COUNTER_V_INTERVAL.
2481 (struct gcov_ctr_summary): Remove.
2482 (struct gcov_summary): Directly use fields of former
2484 (compute_working_sets): Use gcov_summary struct.
2485 * gcov.c (read_count_file): Do not use ctrs fields.
2486 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2488 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2490 * profile.c: Likewise.
2491 * profile.h: Likewise.
2493 2018-06-05 Martin Liska <mliska@suse.cz>
2495 PR gcov-profile/84846
2496 * gcov.c (output_lines): Print working directory only
2497 in intermediate format.
2499 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2501 * config/s390/s390-builtin-types.def: Add void function type.
2502 * config/s390/s390-builtins.def: Use the function type for the
2505 2018-06-04 Jim Wilson <jimw@sifive.com>
2507 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2509 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2510 and maybe_eh_return. Change regno to unsigned int. Use new args to
2511 handle EH_RETURN_DATA_REGNO registers properly.
2512 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2513 (riscv_expand_epilogue): Update comment. Change argument name and
2514 type. Update code to use new name and type. Pass new args to
2515 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2517 * config/riscv/riscv.md (NORMAL_RETURN): New.
2518 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2519 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2520 (eh_return): Call gen_eh_return_internal and emit barrier.
2521 (eh_return_internal): Call riscv_expand_epilogue.
2523 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2525 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2526 bit_insertion field and declare can_be_merged_into method.
2527 (merged_store_group::can_be_merged_into): New method.
2528 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2529 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2530 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2532 2018-06-04 Richard Biener <rguenther@suse.de>
2534 PR tree-optimization/85955
2535 * builtins.c (fold_builtin_sincos): Convert pointers to
2536 destination to appropriate type before dereferencing.
2538 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2540 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2542 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2544 * expr.c (expand_expr_real_1): Force the operand into memory if
2545 its TYPE_MODE is BLKmode and if there is no integer mode for
2546 the number of bits being extracted.
2548 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2552 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2553 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2555 2018-06-04 Richard Biener <rguenther@suse.de>
2557 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2558 (cleanup_tree_cfg_noloop): ... single caller. Do
2559 start_recording_case_labels later.
2561 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2563 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2564 to _IMMINTRIN_H_INCLUDED.
2565 * config/i386/pconfigintrin.h: Ditto.
2566 * config/i386/waitpkgintrin.h: Ditto.
2567 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2568 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2569 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2570 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2571 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2572 waitpkgintrin.h and cldemoteintrin.h.
2574 2018-06-04 Richard Biener <rguenther@suse.de>
2576 PR tree-optimization/86038
2577 * tracer.c (find_best_successor): Check probability for
2578 being initialized, bail out if not.
2580 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2583 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2584 of bits to ignore when comparing architectures.
2586 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2588 PR tree-optimization/69615
2589 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2590 maximum or minimum of the type, try to merge it also as if
2591 range1 is + [-, x - 1] or + [x + 1, -].
2594 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2596 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2598 PR tree-optimization/86034
2599 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2600 the unsigned bitfield type in a bit insertion sequence if it does not
2601 have a larger precision than the bitfield size.
2602 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2604 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2606 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2608 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2610 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2611 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2612 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2613 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2615 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2617 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2618 Disable -fdelete-null-pointer-checks for ELF toolchain.
2620 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2621 Kito Cheng <kito.cheng@gmail.com>
2623 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2624 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2625 (nds32*-*-*): Add glibc and uclibc conditions.
2626 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2627 (TARGET_EXCEPT_UNWIND_INFO): Define.
2628 * config/nds32/elf.h: New file.
2629 * config/nds32/linux.h: New file.
2630 * config/nds32/nds32-elf.opt: New file.
2631 * config/nds32/nds32-linux.opt: New file.
2632 * config/nds32/nds32-fp-as-gp.c
2633 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2634 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2636 (nds32_asm_file_end): Ditto.
2637 (nds32_print_operand): Ditto.
2638 (nds32_insert_attributes): Ditto.
2639 (nds32_init_libfuncs): New function.
2640 (TARGET_HAVE_TLS): Define.
2641 (TARGET_INIT_LIBFUNCS): Define.
2642 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2644 (TARGET_ELF): Apply different mcmodel setting.
2645 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2646 been migrated into elf.h and linux.h files.
2647 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2648 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2649 (mcmodel): The content has been migrated into nds32-elf.opt and
2650 nds32-linux.opt files.
2651 * config/nds32/t-elf: New file.
2652 * config/nds32/t-linux: New file.
2654 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2655 Shiva Chen <shiva0217@gmail.com>
2657 * config/nds32/constants.md (unspec_volatile_element): Add
2658 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2659 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2661 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2662 (make_pass_nds32_fp_as_gp): Declare.
2663 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2665 (nds32_asm_function_end_prologue): Remove unused asm output.
2666 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2667 (nds32_asm_file_start): Output necessary fp_as_gp information.
2668 (nds32_option_override): Adjust register usage.
2669 (nds32_expand_prologue): Consider fp_as_gp situation.
2670 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2671 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2674 (simple_return): Ditto.
2675 (omit_fp_begin): Output special directive for fp_as_gp.
2676 (omit_fp_end): Output special directive for fp_as_gp.
2677 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2678 mforbid-fp-as-gp): New options.
2680 2018-06-01 Mark Wielaard <mark@klomp.org>
2682 * dwarf2out.c (dwarf2out_finish): Remove generation of
2683 DW_AT_loclists_base.
2685 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2687 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2688 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2689 (struct merged_store_group): Add bit_insertion field.
2690 (dump_char_array): Use standard hexadecimal format.
2691 (merged_store_group::merged_store_group): Set bit_insertion to false.
2692 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2693 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2694 also print the mask in the dump file.
2695 (pass_store_merging::gate): Minor tweak.
2696 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2697 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2698 stores with INTEGER_CST stores.
2699 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2700 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2701 and use it throughout. Generate bit insertion sequences if need be.
2702 (pass_store_merging::process_store): Remove redundant condition.
2703 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2705 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2707 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2708 the 128-bit floating point types. Fix function comment.
2710 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2712 * config/aarch64/aarch64-simd.md
2713 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2715 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2718 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2720 PR tree-optimization/85989
2721 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2723 (backprop::intersect_uses): Check it when deciding whether this
2724 is a backedge reference.
2725 (backprop::process_block): Add each phi to m_visited_phis
2726 after visiting it, then clear it at the end.
2728 2018-06-01 Richard Biener <rguenther@suse.de>
2730 * tree-vectorizer.h (vect_dr_stmt): New function.
2731 (vect_get_load_cost): Adjust.
2732 (vect_get_store_cost): Likewise.
2733 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2734 Use vect_dr_stmt instead of DR_SMTT.
2735 (vect_record_base_alignments): Likewise.
2736 (vect_calculate_target_alignment): Likewise.
2737 (vect_compute_data_ref_alignment): Likewise and make static.
2738 (vect_update_misalignment_for_peel): Likewise.
2739 (vect_verify_datarefs_alignment): Likewise.
2740 (vector_alignment_reachable_p): Likewise.
2741 (vect_get_data_access_cost): Likewise. Pass down
2742 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2743 (vect_get_peeling_costs_all_drs): Likewise.
2744 (vect_peeling_hash_get_lowest_cost): Likewise.
2745 (vect_enhance_data_refs_alignment): Likewise.
2746 (vect_find_same_alignment_drs): Likewise.
2747 (vect_analyze_data_refs_alignment): Likewise.
2748 (vect_analyze_group_access_1): Likewise.
2749 (vect_analyze_group_access): Likewise.
2750 (vect_analyze_data_ref_access): Likewise.
2751 (vect_analyze_data_ref_accesses): Likewise.
2752 (vect_vfa_segment_size): Likewise.
2753 (vect_small_gap_p): Likewise.
2754 (vectorizable_with_step_bound_p): Likewise.
2755 (vect_prune_runtime_alias_test_list): Likewise.
2756 (vect_analyze_data_refs): Likewise.
2757 (vect_supportable_dr_alignment): Likewise.
2758 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2759 (vect_gen_prolog_loop_niters): Likewise.
2760 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2761 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2763 (vect_recog_mask_conversion_pattern): Likewise.
2764 (vect_try_gather_scatter_pattern): Likewise.
2765 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2766 to vect_get_store_cost.
2767 (vect_get_store_cost): Get stmt_info instead of DR.
2768 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2769 (vect_get_load_cost): Get stmt_info instead of DR.
2771 2018-06-01 Richard Biener <rguenther@suse.de>
2774 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2775 vars wrapped in MEM_REFs.
2777 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2779 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2780 Fix subreg tests so that we only return a choice between
2781 GENERAL_REGS and FP_REGS if the original classes included both.
2783 2018-06-01 Richard Biener <rguenther@suse.de>
2786 * tree-ssa-structalias.c (get_function_part_constraint):
2787 Handle NULL fi->decl.
2788 (find_func_aliases_for_call): Properly handle indirect
2789 fi from direct call.
2790 (find_func_clobbers): Likewise.
2791 (ipa_pta_execute): Likewise.
2792 (create_variable_info_for): For functions that are ifunc_resolver
2793 resolve to a varinfo that contains the result of the resolver call.
2794 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2797 2018-05-31 Michael Collison <michael.collison@arm.com>
2799 * config/aarch64/aarch64.md:
2800 (*fix_to_zero_extenddfdi2): New pattern.
2801 * gcc.target/aarch64/fix_extend1.c: New testcase.
2803 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2807 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2808 and BUILT_IN_STRNCMP_EQ.
2809 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2810 BUILT_IN_STRNCMP_EQ.
2811 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2812 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2813 (gimple_fold_builtin): Likewise.
2814 * tree-ssa-strlen.c (compute_string_length): New function.
2815 (determine_min_obsize): New function.
2816 (handle_builtin_string_cmp): New function to handle calls to
2817 string compare functions.
2818 (strlen_optimize_stmt): Add handling to builtin string compare
2820 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2821 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2822 * tree.c (build_common_builtin_nodes): Add new defines of
2823 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2825 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2828 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2829 functions with naked attribute.
2831 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2833 * config/i386/sse.md (avx_vec_concat<mode>):
2834 Substitute concat_tg_mode mode attribute with xtg_mode.
2835 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2836 (concat_tg_mode): Remove mode attribute.
2838 2018-05-31 Martin Sebor <msebor@redhat.com>
2841 * calls.c (alloc_max_size): Correct a logic error/typo.
2842 Treat excessive arguments as infinite. Warn for invalid arguments.
2843 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2845 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2848 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2849 and movx for Haswell.
2851 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2852 Cesar Philippidis <cesar@codesourcery.com>
2855 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2856 when emitting error on private/firstprivate reductions.
2857 * omp-low.c (lower_omp_target): Avoid reference-type processing
2858 on pointers for firstprivate clause.
2860 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2862 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2865 * config/aarch64/aarch64-simd.md
2866 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2867 (aarch64_ld1_x3_<mode>): Likewise
2868 (aarch64_st1x2<VALLDIF:mode>): Likewise
2869 (aarch64_st1_x2_<mode>): Likewise
2870 (aarch64_st1x3<VALLDIF:mode>): Likewise
2871 (aarch64_st1_x3_<mode>): Likewise
2872 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2873 (vld1_s8_x3): Likewise.
2874 (vld1_u16_x3): Likewise.
2875 (vld1_s16_x3): Likewise.
2876 (vld1_u32_x3): Likewise.
2877 (vld1_s32_x3): Likewise.
2878 (vld1_u64_x3): Likewise.
2879 (vld1_s64_x3): Likewise.
2880 (vld1_f16_x3): Likewise.
2881 (vld1_f32_x3): Likewise.
2882 (vld1_f64_x3): Likewise.
2883 (vld1_p8_x3): Likewise.
2884 (vld1_p16_x3): Likewise.
2885 (vld1_p64_x3): Likewise.
2886 (vld1q_u8_x3): Likewise.
2887 (vld1q_s8_x3): Likewise.
2888 (vld1q_u16_x3): Likewise.
2889 (vld1q_s16_x3): Likewise.
2890 (vld1q_u32_x3): Likewise.
2891 (vld1q_s32_x3): Likewise.
2892 (vld1q_u64_x3): Likewise.
2893 (vld1q_s64_x3): Likewise.
2894 (vld1q_f16_x3): Likewise.
2895 (vld1q_f32_x3): Likewise.
2896 (vld1q_f64_x3): Likewise.
2897 (vld1q_p8_x3): Likewise.
2898 (vld1q_p16_x3): Likewise.
2899 (vld1q_p64_x3): Likewise.
2900 (vst1_s64_x2): Likewise.
2901 (vst1_u64_x2): Likewise.
2902 (vst1_f64_x2): Likewise.
2903 (vst1_s8_x2): Likewise.
2904 (vst1_p8_x2): Likewise.
2905 (vst1_s16_x2): Likewise.
2906 (vst1_p16_x2): Likewise.
2907 (vst1_s32_x2): Likewise.
2908 (vst1_u8_x2): Likewise.
2909 (vst1_u16_x2): Likewise.
2910 (vst1_u32_x2): Likewise.
2911 (vst1_f16_x2): Likewise.
2912 (vst1_f32_x2): Likewise.
2913 (vst1_p64_x2): Likewise.
2914 (vst1q_s8_x2): Likewise.
2915 (vst1q_p8_x2): Likewise.
2916 (vst1q_s16_x2): Likewise.
2917 (vst1q_p16_x2): Likewise.
2918 (vst1q_s32_x2): Likewise.
2919 (vst1q_s64_x2): Likewise.
2920 (vst1q_u8_x2): Likewise.
2921 (vst1q_u16_x2): Likewise.
2922 (vst1q_u32_x2): Likewise.
2923 (vst1q_u64_x2): Likewise.
2924 (vst1q_f16_x2): Likewise.
2925 (vst1q_f32_x2): Likewise.
2926 (vst1q_f64_x2): Likewise.
2927 (vst1q_p64_x2): Likewise.
2928 (vst1_s64_x3): Likewise.
2929 (vst1_u64_x3): Likewise.
2930 (vst1_f64_x3): Likewise.
2931 (vst1_s8_x3): Likewise.
2932 (vst1_p8_x3): Likewise.
2933 (vst1_s16_x3): Likewise.
2934 (vst1_p16_x3): Likewise.
2935 (vst1_s32_x3): Likewise.
2936 (vst1_u8_x3): Likewise.
2937 (vst1_u16_x3): Likewise.
2938 (vst1_u32_x3): Likewise.
2939 (vst1_f16_x3): Likewise.
2940 (vst1_f32_x3): Likewise.
2941 (vst1_p64_x3): Likewise.
2942 (vst1q_s8_x3): Likewise.
2943 (vst1q_p8_x3): Likewise.
2944 (vst1q_s16_x3): Likewise.
2945 (vst1q_p16_x3): Likewise.
2946 (vst1q_s32_x3): Likewise.
2947 (vst1q_s64_x3): Likewise.
2948 (vst1q_u8_x3): Likewise.
2949 (vst1q_u16_x3): Likewise.
2950 (vst1q_u32_x3): Likewise.
2951 (vst1q_u64_x3): Likewise.
2952 (vst1q_f16_x3): Likewise.
2953 (vst1q_f32_x3): Likewise.
2954 (vst1q_f64_x3): Likewise.
2955 (vst1q_p64_x3): Likewise.
2957 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2959 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2960 user_label_prefix to name.
2962 * tree-core.h: Update comment about the format of NAME string
2963 passed to handler in attribute_spec.
2965 * config/msp430/msp430.md: Remove erroneous subreg expression from
2966 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2967 zero_extend{q,h}isi2.
2969 2018-05-30 Borislav Petkov <bp@suse.de>
2971 * doc/extend.texi: Document some architecture specific
2972 constraints and sort entries.
2974 2018-05-30 Martin Sebor <msebor@redhat.com>
2977 * builtins.c (expand_builtin_stpcpy_1): New function.
2978 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2979 only if the former succeeds.
2981 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2983 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2986 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2988 * doc/invoke.texi (-flinker-output): Document
2990 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2992 * passes.c (ipa_write_summaries): Only modify statements if body
2994 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2995 incrementally linking.
2996 (ipa_passes): Likewise.
2997 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2998 pass down resolution info.
2999 * common.opt (flag_incremental_link): Update info.
3000 * gcc.c (plugin specs): Turn flinker-output=* to
3001 -plugin-opt=-linker-output-known
3002 * toplev.c (compile_file): Also cut compilation when doing incremental
3004 * flag-types. (enum lto_partition_model): Add
3005 LTO_LINKER_OUTPUT_NOLTOREL.
3006 (invoke.texi): Add -flinker-output docs.
3007 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
3008 link same way as WPA; do not stream in dead initializers.
3010 * dwarf2out.c (dwarf2out_die_ref_for_decl,
3011 darf2out_register_external_decl): Support incremental link.
3013 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3015 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
3017 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3019 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
3020 it down to simple_object_copy_lto_debug_sections.
3021 (run_gcc): Determine incremental LTO link time and configure
3022 lto1 into non-wpa mode, disable renaming of debug sections.
3024 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
3026 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
3027 descriptions of various incorrectly documented functions.
3029 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
3032 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3033 address check not strict.
3035 2018-05-30 Richard Biener <rguenther@suse.de>
3037 PR tree-optimization/85964
3038 * tracer.c (better_p): Drop initialized count check, we only
3039 call the function with initialized counts now.
3040 (find_best_successor): Do find a best edge if one
3041 has uninitialized count.
3042 (find_best_predecessor): Likewise. Do BB frequency check only
3043 if count is initialized.
3045 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3047 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
3048 (aarch64_ldrstr_offset_compare): New.
3049 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
3050 load/store orderings.
3051 (aarch64_gen_adjusted_ldpstp): Likewise.
3053 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3055 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
3056 Check for subset of GENERAL_REGS and FP_REGS.
3057 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
3060 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
3062 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
3063 and wi::to_poly_offset. Add the current offset and then check
3064 whether the sum fits, rather than using an unchecked addition of
3065 a checked term. Check for a shwi rather than a uhwi.
3066 * expr.c (get_bit_range): Use tree_to_poly_uint64.
3067 (store_constructor): Use poly_int_tree_p.
3068 (expand_expr_real_1): Likewise.
3069 * function.c (assign_temp): Likewise.
3070 * fold-const.c (const_binop): Use poly_int_tree_p and
3072 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
3074 * ipa-icf-gimple.c (func_checker::compare_operand): Use
3075 to_poly_offset for MEM offsets.
3076 * ipa-icf.c (sem_variable::equals): Likewise.
3077 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
3078 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
3079 wi::to_poly_offset for BIT_FIELD_REF offsets.
3080 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
3082 * var-tracking.c (emit_note_insn_var_location): Use
3083 tree_to_poly_uint64.
3085 2018-05-29 Jim Wilson <jimw@sifive.com>
3087 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
3089 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
3092 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
3093 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
3095 (sse4_1_round<mode>2): Use nonimmediate_operand
3096 for operand 1 predicate.
3098 2018-05-29 Martin Sebor <msebor@redhat.com>
3099 Richard Biener <rguenther@suse.de>
3102 * calls.c (get_size_range): Call determine_value_range instead
3103 of get_value_range..
3104 * tree-vrp.h (determine_value_range): Declared new function.
3105 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3107 2018-05-29 Richard Biener <rguenther@suse.de>
3109 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3110 sure to use non-pattern stmts for get_earlier_stmt arguments.
3111 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3112 called on pattern stmts.
3113 (get_later_stmt): Likewise.
3115 2018-05-29 Martin Liska <mliska@suse.cz>
3117 PR gcov-profile/85759
3118 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3121 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3123 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3125 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3129 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3130 VEC_PACK_FLOAT_EXPR): New tree codes.
3131 * tree-pretty-print.c (op_code_prio): Handle
3132 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3133 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3134 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3135 * tree-inline.c (estimate_operator_cost): Likewise.
3136 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3137 * fold-const.c (const_binop): Likewise.
3138 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3139 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3140 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3141 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3142 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3143 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3144 * expr.c (expand_expr_real_2): Likewise.
3145 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3146 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3147 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3149 * optabs.c (expand_widen_pattern_expr): For
3150 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3151 sign from result type rather than operand's type.
3152 (expand_binop_directly): For vec_packu_float_optab and
3153 vec_packs_float_optab allow result type to be different from operand's
3155 * optabs-tree.c (optab_for_tree_code): Handle
3156 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3157 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3158 * tree-vect-generic.c (expand_vector_operations_1): Handle
3159 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3160 VEC_PACK_FLOAT_EXPR.
3161 * tree-vect-stmts.c (supportable_widening_operation): Handle
3163 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3164 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3165 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3166 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3167 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3169 (vec_pack<floatprefix>_float_<mode>): New expander.
3170 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3172 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3173 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3174 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3175 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3176 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3178 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3179 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3180 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3181 VEC_PACK_FLOAT_EXPR): Document.
3183 2018-05-29 Richard Biener <rguenther@suse.de>
3185 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3187 (stmt_vec_info_vec): Make pointer.
3188 (init_stmt_vec_info_vec): Remove.
3189 (free_stmt_vec_info_vec): Likewise.
3190 (set_stmt_vec_info_vec): New function.
3191 (free_stmt_vec_infos): Likewise.
3192 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3193 (set_vinfo_for_stmt): Likewise.
3194 (get_earlier_stmt): Likewise.
3195 (get_later_stmt): Likewise.
3196 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3197 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3198 (vec_info::~vec_info): Free stmt_vec_infos.
3199 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3200 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3201 (pass_slp_vectorize::execute): Likewise.
3202 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3203 (free_stmt_vec_info_vec): Likewise.
3204 (set_stmt_vec_info_vec): New function.
3205 (free_stmt_vec_infos): Likewise.
3206 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3207 the global stmt_vec_info_vec.
3208 * tree-parloops.c (gather_scalar_reductions): Use
3209 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3212 2018-05-29 Richard Biener <rguenther@suse.de>
3214 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3216 2018-05-29 Martin Liska <mliska@suse.cz>
3217 David Malcolm <dmalcolm@redhat.com>
3219 * vec.c (test_reverse): New.
3220 (vec_c_tests): Add new test.
3221 * vec.h (vl_ptr>::reverse): New function.
3223 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3225 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3227 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3230 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3232 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3234 2018-05-28 Richard Biener <rguenther@suse.de>
3236 PR tree-optimization/85933
3237 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3238 look at stmts marked as vectorizable.
3240 2018-05-28 Richard Biener <rguenther@suse.de>
3242 PR tree-optimization/85934
3243 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3244 vector boolean check before scalar optimization.
3246 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3248 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3251 2018-05-28 Mark Wielaard <mark@klomp.org>
3253 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3254 if it is an expression containing a minus sign.
3256 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3258 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3260 2018-05-27 Paul Koning <ni1d@arrl.net>
3262 * config/pdp11/pdp11.md (truncsihi2): Remove.
3264 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3265 Chung-Ju Wu <jasonwucj@gmail.com>
3267 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3269 (unaligned_store_dw): Ditto.
3270 * config/nds32/nds32-memory-manipulation.c
3271 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3272 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3273 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3274 (emit_setmem_word_loop): Rename to ...
3275 (emit_setmem_doubleword_loop): ... this.
3276 (nds32_gen_dup_4_byte_to_word_value): New function.
3277 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3278 (nds32_expand_setmem_loop): Refine implementation.
3279 (nds32_expand_setmem_loop_v3m): Ditto.
3280 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3283 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3285 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3287 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3289 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3290 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3291 machine->attr_no_prologue_p.
3292 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3293 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3294 (nds32_expand_epilogue): Consider attr_naked_p.
3295 (nds32_expand_epilogue_v3pop): Likewise.
3296 (nds32_can_use_return_insn): Likewise.
3297 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3298 attr_no_prologue_p fields.
3299 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3301 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3304 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3306 * config/i386/sse.md
3307 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3309 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3311 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3313 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3315 (*<floatsuffix>floatv2div2sf2): Rename to ...
3316 (*float<floatunssuffix>v2div2sf2): ... this.
3317 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3318 (float<floatunssuffix>v2div2sf2_mask): ... this.
3319 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3320 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3321 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3323 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3325 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3327 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3329 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3331 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3333 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3334 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3335 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3336 gen_ufix_truncv8dfv8si2.
3337 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3338 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3339 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3340 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3341 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3342 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3343 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3344 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3346 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3350 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3352 2018-05-25 Jim Wilson <jimw@sifive.com>
3354 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3355 * config/riscv/riscv.c (struct machine_function): Add
3356 interrupt_handler_p and attribute_checked_p fields.
3357 (riscv_attribute_table): Add interrupt.
3358 (riscv_interrupt_type_p): New.
3359 (riscv_save_reg_p): Save extra regs for interrupt handler.
3360 (riscv_use_save_libcall): Return false for interrupt handler.
3361 (riscv_first_stack_step): Add forward declaration.
3362 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3363 for interrupt handler with large frame. Use it for saved reg list.
3364 (riscv_expand_prologue): Move flag_stack_usage_info support to
3365 eliminate duplication.
3366 (riscv_expand_epilogue): Generate mret for interrupt handler.
3367 (riscv_epilogue_uses): New.
3368 (riscv_can_use_return_insn): Return false for interrupt handler.
3369 (riscv_function_ok_for_sibcall): Likewise.
3370 (riscv_set_current_function): Add interrupt handler support.
3371 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3372 * config/riscv/riscv.md (UNSPECV_MRET): New.
3374 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3376 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3378 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3380 PR tree-optimization/85712
3381 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3382 this candidate has already been replaced in-situ by a copy.
3384 2018-05-25 Jason Merrill <jason@redhat.com>
3386 PR c++/80485 - inline function non-zero address.
3387 * symtab.c (nonzero_address): Check DECL_COMDAT.
3389 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3392 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3393 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3394 extension of SImode operation. Use const123_operand predicate.
3395 (*saddsi_1): Remove.
3396 (*saddl_se_1): Ditto.
3398 (*ssubl_se_1): Ditto.
3399 * config/alpha/predicates.md (const123_operand): New predicate.
3400 * config/alpha/constraints.md (P): Use IN_RANGE.
3402 2018-05-25 Richard Biener <rguenther@suse.de>
3404 * tree-ssa-alias.h (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-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3411 (ref_maybe_used_by_stmt_p): Likewise.
3412 (stmt_may_clobber_ref_p): Likewise.
3413 (stmt_may_clobber_ref_p_1): Likewise.
3414 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3415 the alias oracle to disambiguate DRs with stmts DR analysis
3417 (vect_analyze_data_refs): Do not give up on not analyzable
3418 DRs for BB vectorization. Remove code truncating the dataref
3421 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3424 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3425 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3426 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3428 2018-05-25 Richard Biener <rguenther@suse.de>
3430 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3431 function, combining stmt data ref gathering and fatal analysis
3433 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3434 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3435 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3436 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3437 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3439 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3441 PR tree-optimization/85720
3442 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3443 SCC if all partitions are builtins.
3444 (version_loop_by_alias_check): New parameter. Generate cancelable
3445 runtime alias check if all partitions are builtins.
3446 (distribute_loop): Update call to above function.
3448 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3450 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3451 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3452 (parm_default_def_partition_arg): Ditto.
3453 (set_parm_default_def_partition): Ditto.
3454 (get_parm_default_def_partitions): Ditto and make it static.
3455 (get_undefined_value_partitions): Ditto and make it static.
3456 (remove_ssa_form): Refactor call to init_var_map here.
3457 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3458 computation for loop region.
3459 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3460 (register_default_def): Delete.
3461 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3462 (parm_default_def_partition_arg): Ditto.
3463 (set_parm_default_def_partition): Ditto.
3464 (get_parm_default_def_partitions): Ditto and make it static.
3465 (get_undefined_value_partitions): Ditto and make it static.
3466 (coalesce_with_default, coalesce_with_default): Update comment.
3467 (create_coalesce_list_for_region): New func factored out from
3468 create_outofssa_var_map.
3469 (populate_coalesce_list_for_outofssa): New func factored out from
3470 create_outofssa_var_map and coalesce_ssa_name.
3471 (create_outofssa_var_map): Delete.
3472 (coalesce_ssa_name): Refactor to support live range computation.
3473 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3474 (get_parm_default_def_partitions): Delete.
3475 (get_undefined_value_partitions): Ditto.
3476 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3477 computation for loop region.
3478 (new_tree_live_info, loe_visit_block): Ditto.
3479 (live_worklist, set_var_live_on_entry): Ditto.
3480 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3481 * tree-ssa-live.h (struct _var_map): New fields.
3482 (init_var_map): Change decl.
3483 (region_contains_p): New.
3485 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3487 * tree-ssa-live.h (live_merge_and_clear): Delete.
3489 2018-05-25 Richard Biener <rguenther@suse.de>
3492 * tree-dump.c (dequeue_and_dump): Remove access to removed
3493 operand 2 of a SWITCH_EXPR.
3495 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3497 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3498 multiplication and division.
3499 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3500 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3501 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3502 (cond_udiv_optab, cond_umod_optab): New optabs.
3503 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3504 (IFN_COND_RDIV): New internal functions.
3505 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3506 TRUNC_MOD_EXPR and RDIV_EXPR.
3507 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3508 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3510 (SVE_INT_BINARY): Include mult.
3511 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3512 (optab, sve_int_op): Handle mult.
3513 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3515 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3516 for SVE_INT_BINARY_SD.
3518 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3520 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3521 (optab, sve_int_op): Handle div and udiv.
3522 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3523 for SVE_INT_BINARY_SD.
3524 (*<optab><mode>3): New insn for the same.
3526 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3528 * tree-vect-patterns.c: Include predict.h.
3529 (vect_recog_divmod_pattern): Restrict check for division support
3530 to when optimizing for size.
3532 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3534 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3535 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3536 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3537 (gimple_match_op::set_op): Likewise.
3538 (gimple_resimplify4): Declare.
3539 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3540 (expr::gen_transform): Likewise.
3541 (decision_tree::gen): Generate a simplification routine for 4 operands.
3542 * gimple-match-head.c (gimple_simplify): Add an overload for
3543 4 operands. In the top-level function, handle up to 4 call
3544 arguments and call gimple_resimplify4.
3545 (gimple_resimplify4): New function.
3546 (build_call_internal): Pass a fourth operand.
3547 (maybe_push_to_seq): Likewise.
3548 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3549 Fold VEC_COND_EXPRs of an operation and a default value into
3550 an IFN_COND_* function if possible.
3551 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3553 (SVE_COND_FP_BINARY): Include them.
3554 (optab, sve_fp_op): Handle them.
3555 (SVE_INT_BINARY_REV): New code iterator.
3556 (SVE_COND_FP_BINARY_REV): New int iterator.
3557 (commutative): New int attribute.
3558 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3560 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3562 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3563 (*cond_<optab><mode>): New patterns for reversed operands.
3565 2018-05-25 Richard Biener <rguenther@suse.de>
3567 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3568 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3569 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3570 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3571 * tree-vect-data-refs.c (everywhere): Adjust users.
3572 * tree-vect-loop.c (everywhere): Likewise.
3573 * tree-vect-slp.c (everywhere): Likewise.
3574 * tree-vect-stmts.c (everywhere): Likewise.
3575 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3577 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3579 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3581 (gcc_cv_as_section_exclude): ... this.
3582 Try Solaris as #exclude syntax.
3583 * configure: Regenerate.
3584 * config.in: Regenerate.
3585 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3587 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3588 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3590 * varasm.c (default_elf_asm_named_section): Don't check if
3591 HAVE_GAS_SECTION_EXCLUDE is defined.
3593 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3595 * doc/md.texi: Update the documentation of the cond_* optabs
3596 to mention the new final operand. Fix GET_MODE_NUNITS call.
3597 Describe the scalar case too.
3598 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3599 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3601 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3602 (get_conditional_internal_fn): Update comment.
3603 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3604 accumulator value as a final argument to conditional functions.
3605 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3606 a define_expand and add an "else" operand. Assert for now that
3607 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3608 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3609 (*cond_<optab><mode>): New patterns.
3610 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3611 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3612 (UNSPEC_COND_EOR): Delete.
3613 (optab): Remove associated mappings.
3614 (SVE_INT_BINARY): New code iterator.
3615 (sve_int_op): Remove int attribute and add "minus" to the code
3617 (SVE_COND_INT_OP): Delete.
3618 (SVE_COND_FP_OP): Rename to...
3619 (SVE_COND_FP_BINARY): ...this.
3621 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3623 * optabs.c (can_reuse_operands_p): New function.
3624 (maybe_legitimize_operands): Try to reuse the results for
3627 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3629 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3630 Add {q} suffix to insn mnemonic.
3632 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3634 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3635 (msp430_warn_func_return): New.
3637 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3639 * fold-const.c (tree_nonzero_bits): New function.
3640 * fold-const.h (tree_nonzero_bits): Likewise.
3641 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3642 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3644 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3648 * varasm.c (assemble_alias): Check ifunc_resolver only on
3651 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3654 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3655 when memory input operand is handled.
3657 2018-05-24 Luis Machado <luis.machado@linaro.org>
3659 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3661 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3663 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3665 * match.pd: Delay FMA folds until after vectorization.
3667 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3670 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3671 address check not strict.
3673 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3675 * gimple-match.h (gimple_match_op): New class.
3676 (mprts_hook): Replace parameters with a gimple_match_op *.
3677 (maybe_build_generic_op): Likewise.
3678 (gimple_simplified_result_is_gimple_val): Replace parameters with
3679 a const gimple_match_op *.
3680 (gimple_simplify): Replace code_helper * and tree * parameters with
3681 a gimple_match_op * parameter.
3682 (gimple_resimplify1): Replace code_helper *, tree and tree *
3683 parameters with a gimple_match_op * parameter.
3684 (gimple_resimplify2): Likewise.
3685 (gimple_resimplify3): Likewise.
3686 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3687 parameters with a gimple_match_op * parameter.
3688 * gimple-match-head.c (gimple_simplify): Change prototypes of
3689 auto-generated functions to take a gimple_match_op * instead of
3690 separate code_helper * and tree * parameters. Make the same
3691 change in the top-level overload and update calls to the
3692 gimple_resimplify routines. Update calls to the auto-generated
3693 functions and to maybe_push_res_to_seq in the publicly-facing
3694 operation-specific gimple_simplify overloads.
3695 (gimple_match_op::MAX_NUM_OPS): Define.
3696 (gimple_resimplify1): Replace rcode and ops with a single res_op
3697 parameter. Update call to gimple_simplify.
3698 (gimple_resimplify2): Likewise.
3699 (gimple_resimplify3): Likewise.
3700 (mprts_hook): Replace parameters with a gimple_match_op *.
3701 (maybe_build_generic_op): Likewise.
3702 (build_call_internal): Replace type, nargs and ops with
3703 a gimple_match_op *.
3704 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3705 with a single gimple_match_op *. Update calls to mprts_hook,
3706 build_call_internal and gimple_simplified_result_is_gimple_val.
3707 Factor out code that is common to the tree_code and combined_fn cases.
3708 * genmatch.c (expr::gen_transform): Replace tem_code and
3709 tem_ops with a gimple_match_op called tem_op. Update calls
3710 to the gimple_resimplify functions and maybe_push_res_to_seq.
3711 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3712 res_ops. Update call to the gimple_resimplify functions.
3713 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3714 (decision_tree::gen): Make the functions take a gimple_match_op *
3715 called res_op instead of separate res_code and res_ops parameters.
3716 Update call accordingly.
3717 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3718 and ops with a single res_op parameter. Update calls to
3719 maybe_build_generic_op and maybe_push_res_to_seq.
3720 (fold_stmt_1): Update calls to gimple_simplify and
3721 replace_stmt_with_simplification.
3722 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3723 and gimple_simplified_result_is_gimple_val.
3724 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3726 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3727 with a gimple_match_op *.
3728 (vn_nary_build_or_lookup): Likewise. Update call to
3729 vn_nary_build_or_lookup_1.
3730 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3731 gimple_match_op *. Update calls to the gimple_resimplify routines
3732 and to gimple_simplified_result_is_gimple_val.
3733 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3734 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3735 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3736 (visit_nary_op): Likewise.
3737 (visit_reference_op_load): Likewise.
3739 2018-05-23 Luis Machado <luis.machado@linaro.org>
3741 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3742 modifier for printing the step amount.
3744 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3747 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3750 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3752 * doc/sourcebuild.texi (Endianness): New subsubsection.
3754 2018-05-23 Luis Machado <luis.machado@linaro.org>
3756 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3757 <prefetch_dynamic_strides>: New const bool field.
3758 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3759 prefetch_dynamic_strides.
3760 (exynosm1_prefetch_tune): Likewise.
3761 (thunderxt88_prefetch_tune): Likewise.
3762 (thunderx_prefetch_tune): Likewise.
3763 (thunderx2t99_prefetch_tune): Likewise.
3764 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3766 (aarch64_override_options_internal): Update to set
3767 PARAM_PREFETCH_DYNAMIC_STRIDES.
3768 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3769 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3770 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3771 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3772 prefetch-dynamic-strides setting.
3774 2018-05-23 Luis Machado <luis.machado@linaro.org>
3776 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3777 <minimum_stride>: New const int field.
3778 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3779 minimum_stride field defaulting to -1.
3780 (exynosm1_prefetch_tune): Likewise.
3781 (thunderxt88_prefetch_tune): Likewise.
3782 (thunderx_prefetch_tune): Likewise.
3783 (thunderx2t99_prefetch_tune): Likewise.
3784 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3785 <default_opt_level>: Set to 3.
3786 (aarch64_override_options_internal): Update to set
3787 PARAM_PREFETCH_MINIMUM_STRIDE.
3788 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3789 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3790 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3791 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3792 stride is constant and is below the minimum stride threshold.
3794 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3796 * config/arm/arm-cpus.in (mode26): Delete.
3797 (armv4): Delete mode26 reference.
3798 * config/arm/arm.c (arm_configure_build_target): Delete use of
3801 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3803 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3805 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3806 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3808 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3809 (floatunsdidf2): Ditto.
3811 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3813 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3814 (fixuns_trunc<mode>si2_avx512f): Ditto.
3815 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3816 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3817 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3819 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3821 PR rtl-optimization/79985
3822 * df-scan.c (df_insn_refs_collect): Remove special case for
3823 global registers and asm statements.
3825 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3827 * extend.texi (Global Register Variables): Rewrite the bullet list.
3828 Note that the register is available for allocation. Note that access
3829 via inline asm must use constraints. Add note about async-signal
3830 handlers. Remove paragraph about automagic register selection.
3832 2018-05-23 Richard Biener <rguenther@suse.de>
3834 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3835 of fixed offset from memset VN.
3837 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3839 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3841 (alloc_cand_and_find_basis): Initialize first_interp field.
3842 (slsr_process_mul): Modify first_interp field.
3843 (slsr_process_add): Likewise.
3844 (slsr_process_cast): Modify first_interp field for each new
3846 (slsr_process_copy): Likewise.
3847 (dump_candidate): Dump first_interp field.
3848 (replace_mult_candidate): Process all interpretations, not just
3850 (replace_rhs_if_not_dup): Likewise.
3851 (replace_one_candidate): Likewise.
3853 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3855 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3857 (aarch64_needs_frame_chain): New function.
3858 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3860 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3863 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3864 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3865 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3866 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3867 as true for strict-align.
3868 (aarch64_can_inline_p): Perform checks even when callee has no
3869 attributes to check for strict alignment.
3870 * doc/extend.texi (AArch64 Function Attributes): Document
3872 * doc/invoke.texi: (AArch64 Options): Likewise.
3874 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3876 PR tree-optimization/85853
3877 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3878 the handling of the root of the node to...
3879 (vect_slp_analyze_node_operations_1): ...this new function,
3880 and run the whole thing with the child nodes' def types
3881 set according to their SLP node's def type.
3883 2018-05-23 Richard Biener <rguenther@suse.de>
3886 * tree-data-ref.c (create_runtime_alias_checks): Defer
3887 and ignore overflow warnings.
3889 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3891 PR tree-optimization/85822
3892 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3895 2018-05-23 Richard Biener <rguenther@suse.de>
3897 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3898 memset constants via native_interpret_expr.
3900 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3903 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3905 (cgraph_node::create_alias): Likewise.
3906 (cgraph_node::get_availability): Check ifunc_resolver instead
3907 of looking up ifunc attribute.
3908 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3909 * varasm.c (do_assemble_alias): Likewise.
3910 (assemble_alias): Likewise.
3911 (default_binds_local_p_3): Likewise.
3912 * cgraph.h (cgraph_node): Add ifunc_resolver.
3913 (cgraph_node::only_called_directly_or_aliased_p): Return false
3915 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3917 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3918 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3919 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3920 instead of looking up ifunc attribute.
3922 2018-05-22 Luis Machado <luis.machado@linaro.org>
3924 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3926 2018-05-22 Martin Sebor <msebor@redhat.com>
3929 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3930 only when expasion succeeds.
3931 (expand_builtin_strcmp): Same.
3932 (expand_builtin_strncmp): Same.
3934 2018-05-22 Martin Sebor <msebor@redhat.com>
3936 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3938 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3939 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3941 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3942 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3944 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3945 Add check for legitimate_address.
3946 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3947 (aarch64_swap_ldrstr_operands): New.
3948 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3951 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3952 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3954 * config/aarch64/aarch64.md: New patterns to generate stp
3956 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3957 single words and double words.
3958 (load_pair_sw, load_pair_dw): Likewise.
3959 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3961 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3963 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3964 for different mode ldpstp and add peephole for merged zero stores.
3966 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3968 (aarch64_gen_store_pair): Rename calls to match new patterns.
3969 (aarch64_gen_load_pair): Rename calls to match new patterns.
3970 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3971 (load_pair<DREG:mode><DREG2:mode>): ... This.
3972 (store_pair<mode>): Rename to...
3973 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3974 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3976 (V_INT_EQUIV): Handle SImode.
3977 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3980 2018-05-22 Martin Sebor <msebor@redhat.com>
3983 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3984 or ajust the presumed bound on an operation to avoid unnecessary
3987 2018-05-22 Martin Sebor <msebor@redhat.com>
3989 PR tree-optimization/85826
3990 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3991 assuming that a DECL necesarily has a constant size.
3993 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3996 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3998 2018-05-22 Richard Biener <rguenther@suse.de>
4000 PR tree-optimization/85834
4001 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
4002 non-constant and non-zero memset arguments.
4004 2018-05-22 Martin Liska <mliska@suse.cz>
4007 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
4009 2018-05-22 Richard Biener <rguenther@suse.de>
4011 PR tree-optimization/85863
4012 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
4013 comparisons when vectype is specified.
4014 (vectorizable_condition): Do not specify vectype for
4015 vect_is_simple_cond when SLP vectorizing.
4017 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
4020 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
4021 define __ibm128 as long double.
4022 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
4023 as a distinct type when IEEE 128-bit support is enabled.
4024 (init_float128_ieee): Fix up conversions between IFmode and IEEE
4025 128-bit types to use the correct functions.
4026 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
4027 convert between 128-bit floating point types that have different
4028 modes but the same representation, instead of using gen_lowpart to
4030 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
4032 (IFKF_reg): New attributes to give the register constraints for
4034 (extend<mode>tf2_internal): New insns to mark an explicit
4035 conversion between 128-bit floating point types that have a
4036 different mode but share the same representation.
4038 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
4040 PR tree-optimization/85814
4041 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
4042 a null return from get_strinfo when unsharing the next
4043 strinfo in the chain.
4045 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4048 * varasm.c (weak_finish): Clean up weak_decls.
4050 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4052 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
4053 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
4054 UNSPEC_UADALP values.
4055 * config/aarch64/iterators.md (ABAL): New int iterator.
4058 (sur): Add mappings for the above.
4059 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
4061 (aarch64_<sur>abal<mode>_4): Likewise.
4062 (aarch64_<sur>adalp<mode>_3): Likewise.
4063 (<sur>sadv16qi): New define_expand.
4065 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
4067 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
4068 (*movdf_internal): Ditto.
4069 (*rcpsf2_sse): Ditto.
4070 (*rsqrtsf2_sse): Ditto.
4071 (*sqrt<mode>2_sse): Ditto.
4073 2018-05-21 Tamar Christina <tamar.christina@arm.com>
4075 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
4077 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
4078 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
4079 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
4080 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
4082 * config/aarch64/arm_neon.h: Likewise.
4083 * config/aarch64/iterators.md (VQ_I): New.
4085 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
4087 * config.gcc: Add arc/t-multilib-linux to tmake_file for
4089 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
4092 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4094 * config/nds32/constraints.md (S): New constraint.
4095 * config/nds32/nds32.md (call_internal): Use constraint S.
4096 (call_value_internal): Likewise.
4097 (sibcall_internal): Likewise.
4098 (sibcall_value_internal): Likewise.
4100 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4101 Chung-Ju Wu <jasonwucj@gmail.com>
4103 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4106 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4107 Chung-Ju Wu <jasonwucj@gmail.com>
4109 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4110 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4111 (nds32_rtx_costs_impl): Simplify.
4112 (nds32_address_cost_impl): Simplify.
4113 (nds32_init_rtx_costs): New function.
4114 (nds32_rtx_costs_speed_prefer): Likewise.
4115 (nds32_rtx_costs_size_prefer): Likewise.
4116 (nds32_address_cost_speed_prefer): Likewise.
4117 (nds32_address_cost_speed_fwprop): Likewise.
4118 (nds32_address_cost_size_prefer): Likewise.
4119 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4120 * config/nds32/nds32.c (nds32_option_override): Use
4121 nds32_init_rtx_costs function.
4123 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4125 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4126 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4127 (TARGET_PIPELINE_N8): Likewise.
4128 (TARGET_PIPELINE_N10): Likewise.
4129 (TARGET_PIPELINE_N13): Likewise.
4130 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4132 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4134 * config/nds32/nds32-fpu.md: Update copyright year.
4136 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4138 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4140 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4142 * config/nds32/nds32.c
4143 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4144 * config/nds32/nds32.opt (minline-asm-r15): New option.
4146 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4148 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4150 * config/nds32/nds32.md (abssi2): New pattern.
4152 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4154 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4155 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4156 Merge insn pattern from sse_cvtsi2ss<round_name> and
4157 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4158 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4159 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4160 using SWI48 mode iterator.
4161 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4162 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4163 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4164 pattern from sse_cvttss2si<round_saeonly_name>
4165 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4166 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4167 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4168 using SWI48 mode iterator.
4169 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4170 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4171 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4172 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4173 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4174 using SWI48 mode iterator.
4175 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4176 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4177 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4178 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4179 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4180 SWI48 mode iterator.
4181 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4182 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4183 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4184 pattern from sse_cvttsd2si<round_saeonly_name>
4185 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4187 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4189 * config/nds32/nds32-md-auxiliary.c
4190 (nds32_valid_smw_lwm_base_p): Refine.
4191 (nds32_output_smw_single_word): Refine.
4192 (nds32_output_smw_double_word): New.
4193 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4195 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4197 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4198 (nds32_output_stack_pop): Refine.
4199 (nds32_expand_unaligned_load): Refine.
4200 (nds32_expand_unaligned_store): Refine.
4202 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4203 Chung-Ju Wu <jasonwucj@gmail.com>
4205 * config/nds32/constants.md: Add TP_REGNUM constant.
4206 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4207 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4209 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4210 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4211 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4212 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4213 and PIC code generation.
4214 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4216 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4218 * config/nds32/nds32.md: Support TLS and PIC.
4219 * config/nds32/nds32.c: Support TLS and PIC.
4220 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4221 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4224 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4226 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4227 mode with E_ prefix.
4229 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4230 Chung-Ju Wu <jasonwucj@gmail.com>
4232 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4233 * config/nds32/nds32-md-auxiliary.c
4234 (symbolic_reference_mentioned_p): New.
4235 (nds32_legitimize_ict_address): New.
4236 (nds32_expand_ict_move): New.
4237 (nds32_indirect_call_referenced_p): New.
4238 (nds32_symbol_binds_local_p): Delete.
4239 (nds32_long_call_p): Modify.
4240 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4241 * config/nds32/nds32-protos.h
4242 (symbolic_reference_mentioned_p): Declare.
4243 (nds32_legitimize_ict_address): Declare.
4244 (nds32_expand_ict_move): Declare.
4245 (nds32_indirect_call_referenced_p): Declare.
4246 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4247 (nds32_relax_group): Use nds32_ict_const_p as condition.
4248 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4249 (nds32_asm_file_start): Output ict_model directive in asm code.
4250 (nds32_legitimate_address_p): Consider indirect call.
4251 (nds32_print_operand): Consider indirect call.
4252 (nds32_print_operand_address): Consider indirect call.
4253 (nds32_insert_attributes): Handle "indirect_call" attribute.
4254 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4255 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4256 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4257 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4258 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4259 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4260 (TARGET_ICT_MODEL_SMALL): Define.
4261 (TARGET_ICT_MODEL_LARGE): Define.
4262 * config/nds32/nds32.md (movsi): Consider ict model.
4263 (call, call_value): Consider ict model.
4264 (sibcall, sibcall_value): Consider ict model.
4265 * config/nds32/nds32.opt (mict-model): New option.
4266 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4269 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4270 Monk Chiang <sh.chiang04@gmail.com>
4271 Jim Wilson <jimw@sifive.com>
4273 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4274 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4275 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4276 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4277 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4278 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4279 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4280 compute save_libcall_adjustment properly.
4281 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4282 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4283 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4284 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4285 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4286 (ABI_SPEC): Handle mabi=ilp32e.
4287 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4288 (RVE): Add RVE mask.
4289 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4290 <-march>: Add rv32e as an example.
4292 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4295 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4296 (intra_create_variable_infos): Handle C++ constructors.
4298 2018-05-18 Martin Liska <mliska@suse.cz>
4300 * passes.def: Remove a redundant pass.
4302 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4305 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4307 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4309 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4311 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4313 (armv2, armv2a, armv3, armv3m): Delete architectures.
4314 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4315 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4316 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4318 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4319 (*mulsidi3adddi): Likewise.
4320 (mulsidi3): Likewise.
4321 (*mulsidi3_nov6): Likewise.
4322 (umulsidi3): Likewise.
4323 (umulsidi3_nov6): Likewise.
4324 (umaddsidi4): Likewise.
4325 (*umulsidi3adddi): Likewise.
4326 (smulsi3_highpart): Likewise.
4327 (*smulsi3_highpart_nov6): Likewise.
4328 (umulsi3_highpart): Likewise.
4329 (*umulsi3_highpart_nov6): Likewise.
4330 * config/arm/arm.h (arm_arch3m): Delete.
4331 * config/arm/arm.c (arm_arch3m): Delete.
4332 (arm_option_override_internal): Update armv3-related comment.
4333 (arm_configure_build_target): Delete use of isa_bit_mode32.
4334 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4335 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4336 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4338 (mulusa3): Likewise.
4339 * config/arm/arm-protos.h (arm_arch3m): Delete.
4340 * config/arm/arm-tables.opt: Regenerate.
4341 * config/arm/arm-tune.md: Likewise.
4342 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4343 deleted architectures.
4345 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4347 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4348 (armv5t, armv5te): New features.
4349 (ARMv5, ARMv5e): Delete fgroups.
4350 (ARMv5t, ARMv5te): Adjust for above changes.
4352 (armv5, armv5e): Delete arches.
4353 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4355 (*call_reg_arm): Likewise.
4356 (*call_value_reg_armv5): Likewise.
4357 (*call_value_reg_arm): Likewise.
4358 (*call_symbol): Likewise.
4359 (*call_value_symbol): Likewise.
4360 (*sibcall_insn): Likewise.
4361 (*sibcall_value_insn): Likewise.
4363 (prefetch): Likewise.
4364 (define_split and define_peephole2 dependent on arm_arch5):
4366 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4368 (TARGET_ARM_QBIT): Likewise.
4369 (TARGET_DSP_MULTIPLY): Likewise.
4370 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4371 (arm_arch5, arm_arch5e): Delete.
4372 (arm_arch5t, arm_arch5te): Declare.
4373 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4374 (arm_arch5t): Declare.
4375 (arm_option_reconfigure_globals): Update for the above.
4376 (arm_options_perform_arch_sanity_checks): Update comment, replace
4377 use of arm_arch5 with arm_arch5t.
4378 (use_return_insn): Likewise.
4379 (arm_emit_call_insn): Likewise.
4380 (output_return_instruction): Likewise.
4381 (arm_final_prescan_insn): Likewise.
4382 (arm_coproc_builtin_available): Likewise.
4383 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4384 arm_arch5e with arm_arch5t and arm_arch5te.
4385 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4386 (arm_arch5t, arm_arch5te): Declare.
4387 * config/arm/arm-tables.opt: Regenerate.
4388 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4389 * config/arm/t-multilib: Likewise.
4390 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4391 instead of arm_arch5.
4392 (*call_reg_thumb1): Likewise.
4393 (*call_value_reg_thumb1_v5): Likewise.
4394 (*call_value_reg_thumb1): Likewise.
4395 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4397 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4399 2018-05-18 Martin Liska <mliska@suse.cz>
4401 PR gcov-profile/84846
4402 * doc/gcov.texi: Document -t option of gcov tool.
4404 2018-05-18 Martin Liska <mliska@suse.cz>
4406 PR gcov-profile/84846
4407 * gcov.c (print_usage): Add new -t option.
4408 (process_args): Handle the option.
4409 (generate_results): Use stdout as output when requested by
4412 2018-05-18 Martin Liska <mliska@suse.cz>
4414 PR gcov-profile/84846
4415 * coverage.c (coverage_init): Write PWD to .gcno file.
4416 * doc/gcov.texi: Document how working directory is printed.
4417 * gcov-dump.c (dump_gcov_file): Print PWD.
4418 * gcov.c (output_intermediate_file): Likewise.
4419 (read_graph_file): Read PWD string.
4420 (output_lines): Print PWD.
4422 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4425 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4426 for retval and return false if all args to phi are zero.
4428 2018-05-18 Richard Biener <rguenther@suse.de>
4430 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4432 (evrp_dom_walker::before_dom_children): Call it.
4434 2018-05-18 Richard Biener <rguenther@suse.de>
4436 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4437 results when processing array refs with variable index.
4439 2018-05-18 Toon Moene <toon@moene.org>
4441 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4442 directly after that of -floop-interchange. Indicate that both
4443 options are enabled by default when specifying -O3.
4445 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4447 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4448 iterator. Delete separate integer-mode vec_set<mode> expander.
4449 (aarch64_simd_vec_setv2di): Delete.
4450 (vec_setv2di): Delete.
4451 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4452 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4453 the "w, r" alternative.
4455 2018-05-18 Martin Liska <mliska@suse.cz>
4457 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4458 * tree-pass.h (make_pass_lower_switch_O0): New function.
4459 * tree-switch-conversion.c (node_has_low_bound): Remove.
4460 (node_has_high_bound): Likewise.
4461 (node_is_bounded): Likewise.
4462 (class pass_lower_switch): Make it a template type and create
4464 (pass_lower_switch::execute): Add template argument.
4465 (make_pass_lower_switch): New function.
4466 (make_pass_lower_switch_O0): New function.
4467 (do_jump_if_equal): Remove.
4468 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4469 all the hard work to tree optimization passes.
4471 2018-05-18 Martin Liska <mliska@suse.cz>
4473 * dbgcnt.c (limit_low): Renamed from limit.
4474 (limit_high): New variable.
4475 (dbg_cnt_is_enabled): Check for upper limit.
4476 (dbg_cnt): Adjust dumping.
4477 (dbg_cnt_set_limit_by_index): Add new argument for high
4479 (dbg_cnt_set_limit_by_name): Likewise.
4480 (dbg_cnt_process_single_pair): Parse new format.
4481 (dbg_cnt_process_opt): Use strtok.
4482 (dbg_cnt_list_all_counters): Remove 'value' and add
4484 * doc/invoke.texi: Document changes.
4486 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4488 * doc/sourcebuild.texi (scalar_all_fma): Document.
4489 * tree.def (FMA_EXPR): Delete.
4490 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4491 * internal-fn.c (ternary_direct): New macro.
4492 (expand_ternary_optab_fn): Likewise.
4493 (direct_ternary_optab_supported_p): Likewise.
4494 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4495 * builtins.c (fold_builtin_fma): Delete.
4496 (fold_builtin_3): Don't call it.
4497 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4498 * expr.c (expand_expr_real_2): Likewise.
4499 * fold-const.c (operand_equal_p): Likewise.
4500 (fold_ternary_loc): Likewise.
4501 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4502 * gimple.c (DEFTREECODE): Likewise.
4503 * gimplify.c (gimplify_expr): Likewise.
4504 * optabs-tree.c (optab_for_tree_code): Likewise.
4505 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4506 * tree-eh.c (operation_could_trap_p): Likewise.
4507 (stmt_could_throw_1_p): Likewise.
4508 * tree-inline.c (estimate_operator_cost): Likewise.
4509 * tree-pretty-print.c (dump_generic_node): Likewise.
4510 (op_code_prio): Likewise.
4511 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4512 * tree-ssa-operands.c (get_expr_operands): Likewise.
4513 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4514 * fold-const-call.h (fold_fma): Delete.
4515 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4516 CFN_FNMA and CFN_FNMS.
4518 * genmatch.c (combined_fn): New enum.
4519 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4520 (commutative_op): New function.
4521 (commutate): Use it. Handle more than 2 operands.
4522 (dt_operand::gen_gimple_expr): Use commutative_op.
4523 (parser::parse_expr): Allow :c to be used with non-binary
4524 operators if the commutative operand is known.
4525 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4526 CFN_FMS, CFN_FNMA and CFN_FNMS.
4527 (backprop::process_assign_use): Remove FMA_EXPR handling.
4528 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4529 (gen_hsa_fma): New function.
4530 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4531 IFN_FNMA and IFN_FNMS.
4532 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4533 * gimple-fold.h (follow_all_ssa_edges): Declare.
4534 * gimple-fold.c (follow_all_ssa_edges): New function.
4535 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4536 gimple_build interface and use follow_all_ssa_edges to fold the result.
4537 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4538 instead of checking for optabs directly.
4539 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4540 rather than FMA_EXPRs.
4541 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4542 call to IFN_FMA instead of an FMA_EXPR.
4544 2018-05-17 Jim Wilson <jimw@sifive.com>
4546 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4547 sign extended promoted subreg, and the range does not have the sign bit
4548 set, then do a sign extend.
4550 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4551 test, check for sign extended subreg and/or constant operands, and
4552 do a sign extend in that case.
4554 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4556 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4557 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4559 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4560 Change logics_shift_reg to logics_shift_imm.
4561 (thunderx2t99_fp_loadpair_basic): Delete.
4562 (thunderx2t99_fp_storepair_basic): Delete.
4563 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4564 (thunderx2t99_asimd_polynomial): Delete.
4565 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4566 and neon_fp_mul_d_scalar_q.
4567 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4568 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4569 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4570 (thunderx2t99_asimd_lut): Add missing tbl types.
4571 (thunderx2t99_asimd_ext): Delete.
4572 (thunderx2t99_asimd_load1_1_mult): Delete.
4573 (thunderx2t99_asimd_load1_2_mult): Delete.
4574 (thunderx2t99_asimd_load1_ldp): New.
4575 (thunderx2t99_asimd_load1): New.
4576 (thunderx2t99_asimd_load2): Add missing *load2* types.
4577 (thunderx2t99_asimd_load3): New.
4578 (thunderx2t99_asimd_load4): New.
4579 (thunderx2t99_asimd_store1_1_mult): Delete.
4580 (thunderx2t99_asimd_store1_2_mult): Delete.
4581 (thunderx2t99_asimd_store2_mult): Delete.
4582 (thunderx2t99_asimd_store2_onelane): Delete.
4583 (thunderx2t99_asimd_store_stp): New.
4584 (thunderx2t99_asimd_store1): New.
4585 (thunderx2t99_asimd_store2): New.
4586 (thunderx2t99_asimd_store3): New.
4587 (thunderx2t99_asimd_store4): New.
4589 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4591 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4592 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4594 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4595 Segher Boessenkool <segher@kernel.crashing.org>
4598 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4601 2018-05-17 Richard Biener <rguenther@suse.de>
4603 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4604 for pruning loop and prune defs feeding only already visited PHIs.
4606 2018-05-17 Richard Biener <rguenther@suse.de>
4608 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4610 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4611 Richard Biener <rguenther@suse.de>
4613 PR tree-optimization/85793
4614 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4615 for VMAT_ELEMENTWISE.
4617 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4619 * internal-fn.h (lookup_internal_fn): Declare
4620 * internal-fn.c (lookup_internal_fn): New function.
4621 * gimple.c (gimple_build_call_from_tree): Handle calls to
4623 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4624 internal function names.
4625 * tree-pretty-print.c (dump_generic_node): Likewise.
4626 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4628 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4630 * gimple-fold.h (gimple_build): Make the function forms take
4631 combined_fn rather than built_in_function.
4632 (gimple_simplify): Likewise.
4633 * gimple-match-head.c (gimple_simplify): Likewise.
4634 * gimple-fold.c (gimple_build): Likewise.
4635 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4636 rather than gimple_build_call_internal.
4637 (get_initial_defs_for_reduction): Likewise.
4638 (vect_create_epilog_for_reduction): Likewise.
4639 (vectorizable_live_operation): Likewise.
4641 2018-05-17 Martin Liska <mliska@suse.cz>
4643 * gimple-ssa-sprintf.c (format_directive): Do not use
4644 space in between 'G_' and '('.
4646 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4649 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4650 even if the mask is not all ones.
4653 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4655 (ix86_gimple_fold_builtin): Likewise.
4658 * config/i386/i386.c: Include tree-vector-builder.h.
4659 (ix86_vector_shift_count): New function.
4660 (ix86_fold_builtin): Fold shift builtins by scalar count.
4661 (ix86_gimple_fold_builtin): Likewise.
4663 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4664 _mm512_setzero): New intrinsics.
4666 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4667 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4669 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4670 code generation for cases where splatting a value is not useful.
4671 * simplify-rtx.c (simplify_ternary_operation): Simplify
4672 vec_merge across a vec_duplicate and a paradoxical subreg forming
4673 a vector mode to a vec_concat.
4675 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4677 * config.gcc: Support "goldmont-plus".
4678 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4680 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4681 PROCESSOR_GOLDMONT_PLUS.
4682 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4683 (processor_target_table): Add "goldmont-plus".
4684 (PTA_GOLDMONT_PLUS): Define.
4685 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4686 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4687 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4688 (fold_builtin_cpu): Add "goldmont-plus".
4689 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4690 (ix86_option_override_internal): Add "goldmont-plus".
4691 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4692 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4693 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4694 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4696 2018-05-17 Richard Biener <rguenther@suse.de>
4698 PR tree-optimization/85757
4699 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4700 remove defs that only feed that PHI from further processing.
4702 2018-05-16 Jim Wilson <jimw@sifive.com>
4704 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4706 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4708 2018-05-16 Mark Wielaard <mark@klomp.org>
4710 * dwarf2out.c (count_index_strings): New function.
4711 (output_indirect_strings): Call count_index_strings and generate
4712 header for dwarf_version >= 5.
4714 2018-05-16 Mark Wielaard <mark@klomp.org>
4716 * dwarf2out.c (dwarf_FORM): New function.
4717 (set_indirect_string): Use dwarf_FORM.
4718 (reset_indirect_string): Likewise.
4719 (size_of_die): Likewise.
4720 (value_format): Likewise.
4721 (output_die): Likewise.
4722 (add_skeleton_AT_string): Likewise.
4723 (output_macinfo_op): Likewise.
4724 (index_string): Likewise.
4725 (output_index_string_offset): Likewise.
4726 (output_index_string): Likewise.
4727 (count_index_strings): Likewise.
4729 2018-05-16 Carl Love <cel@us.ibm.com>
4731 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4732 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4734 2018-05-16 Martin Jambor <mjambor@suse.cz>
4736 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4737 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4739 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4741 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4742 (fnma<mode>4): Likewise.
4743 (fms<mode>4): Likewise.
4744 (fnms<mode>4): Likewise.
4745 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4746 (aarch64_fnma<mode>4): Likewise.
4747 (aarch64_fms<mode>4): Likewise.
4748 (aarch64_fnms<mode>4): Likewise.
4749 (aarch64_fnmadd<mode>4): Likewise.
4751 2018-05-16 Jason Merrill <jason@redhat.com>
4753 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4755 2018-05-16 Richard Biener <rguenther@suse.de>
4757 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4758 (dump_stmt_cost): Declare.
4759 (add_stmt_cost): Dump cost we add.
4760 (add_stmt_costs): New function.
4761 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4763 (vect_analyze_stmt): Adjust prototype.
4764 (vectorizable_condition): Likewise.
4765 (vectorizable_live_operation): Likewise.
4766 (vectorizable_reduction): Likewise.
4767 (vectorizable_induction): Likewise.
4768 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4769 cost vector to pass to vectorizable_ and record afterwards.
4770 (vect_model_reduction_cost): Take cost vector argument and adjust.
4771 (vect_model_induction_cost): Likewise.
4772 (vectorizable_reduction): Likewise.
4773 (vectorizable_induction): Likewise.
4774 (vectorizable_live_operation): Likewise.
4775 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4776 SLP_TREE_NUMBER_OF_VEC_STMTS.
4777 (vect_analyze_slp_cost_1): Remove.
4778 (vect_analyze_slp_cost): Likewise.
4779 (vect_slp_analyze_node_operations): Take visited args and
4780 a target cost vector. Avoid processing already visited stmt sets.
4781 (vect_slp_analyze_operations): Use a local cost vector to gather
4782 costs and register those of non-discarded instances.
4783 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4784 (vect_schedule_slp_instance): Remove copying of
4785 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4787 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4788 adding cost. Record cost entry location.
4789 (vect_prologue_cost_for_slp_op): Function to compute cost of
4790 a constant or invariant generated for SLP vect in the prologue,
4791 split out from vect_analyze_slp_cost_1.
4792 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4793 (vect_model_promotion_demotion_cost): Likewise.
4794 (vect_model_store_cost): Likewise, make static.
4795 (vect_model_load_cost): Likewise.
4796 (vectorizable_bswap): Add cost vector arg and adjust.
4797 (vectorizable_call): Likewise.
4798 (vectorizable_simd_clone_call): Likewise.
4799 (vectorizable_conversion): Likewise.
4800 (vectorizable_assignment): Likewise.
4801 (vectorizable_shift): Likewise.
4802 (vectorizable_operation): Likewise.
4803 (vectorizable_store): Likewise.
4804 (vectorizable_load): Likewise.
4805 (vectorizable_condition): Likewise.
4806 (vectorizable_comparison): Likewise.
4807 (can_vectorize_live_stmts): Likewise.
4808 (vect_analyze_stmt): Likewise.
4809 (vect_transform_stmt): Adjust calls to vectorizable_*.
4810 * tree-vectorizer.c: Include gimple-pretty-print.h.
4811 (dump_stmt_cost): New function.
4813 2018-05-16 Richard Biener <rguenther@suse.de>
4815 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4816 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4817 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4818 (check_name): New callback.
4819 (dse_classify_store): Track cycles via a visited bitmap of PHI
4820 defs and simplify handling of in-loop and across loop dead stores
4821 and properly fail for loop-variant refs. Handle byte-tracking with
4822 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4825 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4827 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4828 (vect_get_mask_type_for_stmt): Likewise.
4829 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4831 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4832 to determine the statement's vector type and the vector type that
4833 should be used for calculating nunits. Deal with cases in which
4834 the type has to be deferred.
4835 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4836 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4837 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4838 (vect_determine_vf_for_stmt): New functions, split out from...
4839 (vect_determine_vectorization_factor): ...here.
4840 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4841 (vect_get_mask_type_for_stmt): New functions, split out from
4842 vect_determine_vectorization_factor.
4844 2018-05-16 Richard Biener <rguenther@suse.de>
4846 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4847 verify the [VEC_]COND_EXPR embedded comparison.
4849 2018-05-15 Martin Sebor <msebor@redhat.com>
4851 PR tree-optimization/85753
4852 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4853 RECORD_TYPE in addition to ARRAY_TYPE.
4855 2018-05-15 Martin Sebor <msebor@redhat.com>
4858 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4860 2018-05-15 Richard Biener <rguenther@suse.de>
4862 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4863 add by_clobber_p one. Change algorithm to collect all defs
4864 representing uses we need to walk and try reducing them to
4865 a single one before failing.
4866 (dse_dom_walker::dse_optimize_stmt): Adjust.
4868 2018-05-13 Mark Wielaard <mark@klomp.org>
4870 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4871 (size_of_loc_descr): Likewise.
4872 (output_loc_operands): Likewise.
4873 (output_loc_operands_raw): Likewise.
4874 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4875 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4876 (hash_loc_operands): Likewise.
4877 (compare_loc_operands): Likewise.
4879 2018-05-14 Mark Wielaard <mark@klomp.org>
4881 * dwarf2out.c (count_index_addrs): New function.
4882 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4884 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4886 PR tree-optimization/83648
4887 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4888 return value as malloc candidate.
4890 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4893 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4894 param as true in call to suggest_attribute.
4896 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4898 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4899 -mreadonly-in-sdata.
4901 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4903 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4905 (aarch64_crypto_aesd_fused): Likewise.
4907 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4909 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4910 (movsi_aarch64): Likewise.
4911 (load_pairsi): Likewise.
4912 (load_pairdi): Likewise.
4913 (store_pairsi): Likewise.
4914 (store_pairdi): Likewise.
4915 (load_pairsf): Likewise.
4916 (load_pairdf): Likewise.
4917 (store_pairsf): Likewise.
4918 (store_pairdf): Likewise.
4919 (zero_extend): Likewise.
4920 (trunc): Swap alternatives.
4921 (fcvt_target): Add '?' to prefer w over r.
4923 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4926 * config/i386/i386.md: Disallow non-commutative arithmetics in
4927 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4928 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4929 in the peephole2 before it.
4931 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4933 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4934 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4935 (ix86_handle_option): Handle -mcldemote.
4936 * config.gcc: New header.
4937 * config/i386/cldemoteintrin.h: New file.
4938 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4939 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4941 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4942 OPTION_MASK_ISA_CLDEMOTE.
4943 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4944 (ix86_valid_target_attribute_inner_p): Ditto.
4945 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4946 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4947 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4948 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4949 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4951 * config/i386/i386.opt: Add -mcldemote.
4952 * config/i386/x86intrin.h: New header.
4953 * doc/invoke.texi: Add -mcldemote.
4955 2018-05-14 Richard Biener <rguenther@suse.de>
4957 * doc/match-and-simplify.texi: Adjust :s documentation.
4959 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4961 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4962 intended memcpy size.
4963 (REORDER_45): Likewise.
4965 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4967 * sort.cc: New file.
4968 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4969 * vec.c (qsort_chk): Use gcc_qsort.
4970 * Makefile.in (OBJS-libcommon): Add sort.o.
4971 (build/sort.o): New target. Use it...
4972 (BUILD_RTL): ... here, and...
4973 (build/gencfn-macros): ... here, and...
4974 (build/genmatch): ... here.
4976 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4977 Chung-Ju Wu <jasonwucj@gmail.com>
4979 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4980 * config/nds32/nds32-graywolf.md: New file.
4981 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4982 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4984 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4985 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4986 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4987 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4988 * config/nds32/pipelines.md: Include n15 settings.
4990 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4991 Chung-Ju Wu <jasonwucj@gmail.com>
4993 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4994 * config/nds32/nds32-n13.md: New file.
4995 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4996 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4998 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4999 * config/nds32/nds32.md (pipeline_model): Add n13.
5000 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
5001 * config/nds32/pipelines.md: Include n13 settings.
5003 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
5004 Chung-Ju Wu <jasonwucj@gmail.com>
5006 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
5007 * config/nds32/nds32-n10.md: New file.
5008 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
5009 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
5011 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
5012 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
5013 * config/nds32/nds32.md (pipeline_model): Add n10.
5014 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
5015 * config/nds32/pipelines.md: Include n10 settings.
5017 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
5018 Kito Cheng <kito.cheng@gmail.com>
5019 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
5022 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5023 Add enum values for DSP extension instructions.
5024 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
5026 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
5027 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
5029 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
5030 * config/nds32/nds32-dspext.md: New file for DSP implementation.
5031 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
5032 * config/nds32/nds32-intrinsic.md: Likewise.
5033 * config/nds32/nds32_intrinsic.h: Likewise.
5034 * config/nds32/nds32-md-auxiliary.c: Likewise.
5035 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
5036 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
5037 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
5038 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
5039 * config/nds32/nds32-protos.h: New declarations for DSP extension.
5040 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
5041 TYPE_DMAC in switch statement.
5042 * config/nds32/nds32.c: New checking and implementation for DSP
5043 extension instructions.
5044 * config/nds32/nds32.h: Likewise.
5045 * config/nds32/nds32.md: Likewise.
5046 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
5047 * config/nds32/predicates.md: Implement new predicates for DSP
5050 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5052 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
5053 Reformat alternatives and attributes so it is easier to identify
5054 which constraints/attributes go with which instruction.
5055 (mov<mode>_hardfloat32, FMOVE64): Likewise.
5056 (mov<mode>_softfloat32, FMOVE64): Likewise.
5057 (mov<mode>_hardfloat64, FMOVE64): Likewise.
5058 (mov<mode>_softfloat64, FMOVE64): Likewise.
5060 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
5062 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5064 (Basic PowerPC Built-in Functions): The new name of the
5065 subsection previously known as "PowerPC Built-in Functions".
5066 (Basic PowerPC Built-in Functions Available on all Configurations):
5068 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
5069 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
5070 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
5071 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
5073 2018-05-11 Martin Jambor <mjambor@suse.cz>
5076 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
5079 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5082 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
5084 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
5086 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
5087 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
5088 (ix86_handle_option): Handle -mwaitpkg.
5089 * config.gcc: New header.
5090 * config/i386/cpuid.h (bit_WAITPKG): New bit.
5091 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
5092 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
5094 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5095 OPTION_MASK_ISA_WAITPKG.
5096 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
5097 (ix86_option_override_internal): Add PTA_WAITPKG.
5098 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
5099 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
5100 IX86_BUILTIN_TPAUSE.
5101 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5102 __builtin_ia32_umwait and __builtin_ia32_tpause.
5103 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5104 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5105 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5106 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5107 UNSPECV_TPAUSE): New.
5108 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5109 * config/i386/i386.opt: Add -mwaitpkg.
5110 * config/i386/waitpkgintrin.h: New file.
5111 * config/i386/x86intrin.h: New header.
5112 * doc/invoke.texi: Add -mwaitpkg.
5114 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5117 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5119 (cortex-m0): Use armv6s-m isa.
5120 (cortex-m0plus): Likewise.
5121 (cortex-m1): Likewise.
5122 (cortex-m0.small-multiply): Likewise.
5123 (cortex-m0plus.small-multiply): Likewise.
5124 (cortex-m1.small-multiply): Likewise.
5126 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5127 Jakub Jelinek <jakub@redhat.com>
5129 PR tree-optimization/85692
5130 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5131 source permute as well.
5133 2018-05-11 Martin Liska <mliska@suse.cz>
5136 * doc/extend.texi: Document LLVM style format for no_sanitize
5139 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5141 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5142 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5143 (mode_supports_vsx_dform_quad): Likewise.
5144 (mode_supports_vmx_dform): Move these functions to be next to the
5145 other mode_supports functions.
5146 (mode_supports_dq_form): Likewise.
5147 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5148 mode_supports_dq_form.
5149 (reg_offset_addressing_ok_p): Likewise.
5150 (offsettable_ok_by_alignment): Likewise.
5151 (rs6000_legitimate_offset_address_p): Likewise.
5152 (legitimate_lo_sum_address_p): Likewise.
5153 (rs6000_legitimize_address): Likewise.
5154 (rs6000_legitimize_reload_address): Likewise.
5155 (rs6000_secondary_reload_inner): Likewise.
5156 (rs6000_preferred_reload_class): Likewise.
5157 (rs6000_output_move_128bit): Likewise.
5159 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5161 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5162 Generate SImode target register for null target.
5163 <case IX86_BUILTIN_XGETBV>: Ditto.
5164 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5165 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5167 2018-05-10 Carl Love <cel@us.ibm.com>
5169 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5170 dcbtt and dcbtstt if operands[2] is 0.
5172 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5175 * config/i386/sse.md (usadv64qi): New expander.
5177 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5179 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5180 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5181 -maltivec=be support.
5182 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5183 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5184 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5185 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5186 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5187 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5188 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5189 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5190 altivec_vsumsws): Adjust.
5191 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5192 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5193 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5195 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5196 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5197 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5198 (altivec_lve<VI_char>x): Delete expand.
5199 (*altivec_lve<VI_char>x_internal): Rename to...
5200 (altivec_lve<VI_char>x): ... this.
5201 (altivec_lvxl_<mode>): Delete expand.
5202 (*altivec_lvxl_<mode>_internal): Rename to ...
5203 (altivec_lvxl_<mode>): ... this.
5204 (altivec_stvxl_<mode>): Delete expand.
5205 (*altivec_stvxl_<mode>_internal): Rename to ...
5206 (altivec_stvxl_<mode>): ... this.
5207 (altivec_stve<VI_char>x): Delete expand.
5208 (*altivec_stve<VI_char>x_internal): Rename to ...
5209 (altivec_stve<VI_char>x): ... this.
5210 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5211 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5212 reduc_plus_scal_<mode>): Adjust.
5213 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5215 (rs6000_cpu_cpp_builtins): Adjust.
5216 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5217 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5218 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5219 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5220 -maltivec=be support.
5221 (rs6000_split_vec_extract_var): Adjust.
5222 (rs6000_split_v4si_init): Adjust.
5223 (swap_selector_for_mode): Delete.
5224 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5225 altivec_expand_stvex_be): Delete.
5226 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5227 -maltivec=be support.
5228 (rs6000_gimple_fold_builtin): Ditto.
5229 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5231 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5232 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5233 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5234 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5235 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5236 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5237 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5238 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5239 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5240 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5241 anonymous split): Adjust.
5242 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5243 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5245 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5247 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5248 when --with-gxx-include-dir is also specified.
5249 * configure: Regenerate.
5251 2018-05-09 Jim Wilson <jimw@sifive.com>
5254 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5255 * config/riscv/t-withmultilib: New.
5256 * config/riscv/withmultilib.h: New.
5257 * doc/install.texi: Document RISC-V --with-multilib-list support.
5259 2018-05-09 Richard Biener <rguenther@suse.de>
5261 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5263 (vect_bb_vectorization_profitable_p): Adjust. Compute
5264 actual scalar cost using the cost vector and the add_stmt_cost
5267 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5269 PR rtl-optimization/85645
5270 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5271 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5273 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5275 PR rtl-optimization/85645
5276 * shrink-wrap.c (spread_components): Return a boolean saying if
5277 anything was changed.
5278 (try_shrink_wrapping_separate): Iterate spread_components until
5279 nothing changes anymore.
5281 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5283 PR rtl-optimization/85645
5284 * regrename.c (build_def_use): Also kill the chains that include the
5285 destination of a REG_CFA_REGISTER note.
5287 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5289 PR rtl-optimization/85645
5290 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5291 insn that has a REG_CFA_REGISTER note.
5293 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5295 * cfgexpand.c (expand_clobber): New function.
5296 (expand_gimple_stmt_1): Use it.
5297 * tree-vect-stmts.c (vect_clobber_variable): New function,
5299 (vectorizable_simd_clone_call): ...here.
5300 (vectorizable_store): Emit a clobber either side of an
5301 IFN_STORE_LANES sequence.
5302 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5304 2018-05-09 Tom de Vries <tom@codesourcery.com>
5307 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5308 (define_insn "trap_if_false"): Add exit after trap.
5310 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5312 PR rtl-optimization/85638
5313 * bb-reorder.c: Include common/common-target.h.
5314 (create_forwarder_block): New function extracted from...
5315 (fix_up_crossing_landing_pad): ...here. Rename into...
5316 (dw2_fix_up_crossing_landing_pad): ...this.
5317 (sjlj_fix_up_crossing_landing_pad): New function.
5318 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5319 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5320 from both partitions and exit the loop after one iteration.
5322 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5325 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5327 (Basic PowerPC Built-in Functions): The new name of the
5328 subsection previously known as "PowerPC Built-in Functions".
5329 (Basic PowerPC Built-in Functions Available on all Configurations):
5331 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5333 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5335 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5337 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5340 2018-05-08 Jim Wilson <jimw@sifive.com>
5342 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5343 (LD_EMUL_SUFFIX): New.
5344 (LINK_SPEC): Use it.
5346 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5348 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5350 (Basic PowerPC Built-in Functions): The new name of the
5351 subsection previously known as "PowerPC Built-in Functions".
5352 (Basic PowerPC Built-in Functions Available on all Configurations):
5354 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5356 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5358 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5360 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5363 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5366 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5367 after cmpelim optimization.
5369 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5371 * config.gcc: Support "goldmont".
5372 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5373 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5375 * config/i386/i386.c (m_GOLDMONT): Define.
5376 (processor_target_table): Add "goldmont".
5377 (PTA_GOLDMONT): Define.
5378 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5379 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5380 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5381 (fold_builtin_cpu): Add "goldmont".
5382 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5383 (ix86_option_override_internal): Add "goldmont".
5384 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5385 (processor_type): Add PROCESSOR_GOLDMONT.
5386 * config/i386/i386.md: Add CPU "glm".
5387 * config/i386/glm.md: New file.
5388 * config/i386/x86-tune.def: Add m_GOLDMONT.
5389 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5391 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5394 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5396 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5397 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5398 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5401 * config/i386/i386.c (ix86_fold_builtin): Handle
5402 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5405 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5406 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5408 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5411 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5412 (check_arch): Likewise.
5413 (check_fpu): Return the result rather than printing it.
5414 (end arch): Fix operator precedence.
5415 (end cpu): Likewise.
5416 (END): Print the result from check_fpu.
5418 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5419 Alan Hayward <alan.hayward@arm.com>
5420 David Sherwood <david.sherwood@arm.com>
5422 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5423 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5424 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5425 (*fcmuo<mode>_and): New patterns.
5427 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5429 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5430 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5431 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5432 (cmp_op, sve_imm_con): New code attributes.
5433 (SVE_COND_INT_CMP, imm_con): Delete.
5434 (cmp_op): Remove above unspecs from int attribute.
5435 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5437 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5438 comparison-specific unspecs.
5439 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5440 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5441 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5442 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5443 (*vec_fcm<cmp_op><mode>): Rename to...
5444 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5445 (*vec_fcmuo<mode>): Rename to...
5446 (*fcmuo<mode>): ...this and adjust likewise.
5447 (*pred_fcm<cmp_op><mode>): New pattern.
5448 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5449 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5451 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5453 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5454 (aarch64_emit_sve_predicated_cond): New function.
5455 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5456 (aarch64_emit_unspec_cond_or): Replace with...
5457 (aarch64_emit_sve_or_conds): ...this new function. Use
5458 aarch64_emit_sve_ptrue_op for the individual comparisons and
5459 aarch64_emit_binop to OR them together.
5460 (aarch64_emit_inverted_unspec_cond): Replace with...
5461 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5462 aarch64_emit_sve_ptrue_op for the comparison and
5463 aarch64_emit_unop to invert the result.
5464 (aarch64_expand_sve_vec_cmp_float): Update after the above
5465 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5467 2018-05-07 Nathan Sidwell <nathan@acm.org>
5469 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5470 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5471 (Backwards Compatibility): Likewise.
5473 2018-05-07 Luis Machado <luis.machado@linaro.org>
5477 2018-05-07 Luis Machado <luis.machado@linaro.org>
5479 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5480 <prefetch_dynamic_strides>: New const bool field.
5481 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5482 prefetch_dynamic_strides.
5483 (exynosm1_prefetch_tune): Likewise.
5484 (thunderxt88_prefetch_tune): Likewise.
5485 (thunderx_prefetch_tune): Likewise.
5486 (thunderx2t99_prefetch_tune): Likewise.
5487 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5489 (aarch64_override_options_internal): Update to set
5490 PARAM_PREFETCH_DYNAMIC_STRIDES.
5491 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5492 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5493 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5494 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5495 prefetch-dynamic-strides setting.
5497 2018-05-07 Luis Machado <luis.machado@linaro.org>
5499 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5500 <minimum_stride>: New const int field.
5501 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5502 minimum_stride field.
5503 (exynosm1_prefetch_tune): Likewise.
5504 (thunderxt88_prefetch_tune): Likewise.
5505 (thunderx_prefetch_tune): Likewise.
5506 (thunderx2t99_prefetch_tune): Likewise.
5507 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5508 (aarch64_override_options_internal): Update to set
5509 PARAM_PREFETCH_MINIMUM_STRIDE.
5510 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5511 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5512 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5513 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5514 stride is constant and is below the minimum stride threshold.
5516 2018-05-07 Luis Machado <luis.machado@linaro.org>
5518 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5521 2018-05-07 Luis Machado <luis.machado@linaro.org>
5523 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5524 <prefetch_dynamic_strides>: New const bool field.
5525 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5526 prefetch_dynamic_strides.
5527 (exynosm1_prefetch_tune): Likewise.
5528 (thunderxt88_prefetch_tune): Likewise.
5529 (thunderx_prefetch_tune): Likewise.
5530 (thunderx2t99_prefetch_tune): Likewise.
5531 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5533 (aarch64_override_options_internal): Update to set
5534 PARAM_PREFETCH_DYNAMIC_STRIDES.
5535 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5536 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5537 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5538 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5539 prefetch-dynamic-strides setting.
5541 2018-05-07 Luis Machado <luis.machado@linaro.org>
5543 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5544 <minimum_stride>: New const int field.
5545 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5546 minimum_stride field.
5547 (exynosm1_prefetch_tune): Likewise.
5548 (thunderxt88_prefetch_tune): Likewise.
5549 (thunderx_prefetch_tune): Likewise.
5550 (thunderx2t99_prefetch_tune): Likewise.
5551 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5552 (aarch64_override_options_internal): Update to set
5553 PARAM_PREFETCH_MINIMUM_STRIDE.
5554 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5555 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5556 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5557 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5558 stride is constant and is below the minimum stride threshold.
5560 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5563 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5564 the type is addressable. Don't force op into register if it has
5567 2018-05-05 Roland McGrath <mcgrathr@google.com>
5570 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5571 any section for which we don't know a specific type it should have,
5572 regardless of name. Previously this was done only for the exact
5573 names ".init_array", ".fini_array", and ".preinit_array".
5574 (default_elf_asm_named_section): Add comment about
5575 relationship with default_section_type_flags and SECTION_NOTYPE.
5576 (get_section): Don't consider it a type conflict if one side has
5577 SECTION_NOTYPE and the other doesn't, as long as neither has the
5578 SECTION_BSS et al used in the default_section_type_flags logic.
5580 2018-05-05 Tom de Vries <tom@codesourcery.com>
5583 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5584 (workaround_barsyncs): New function.
5585 (nvptx_reorg): Use workaround_barsyncs.
5586 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5587 (define_expand "nvptx_membar_cta"): New define_expand.
5588 (define_insn "*nvptx_membar_cta"): New insn.
5590 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5592 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5593 To improve optimization opportunities.
5594 * builtin-types.def: The new needed builtin types for the above.
5596 2018-05-04 Richard Biener <rguenther@suse.de>
5598 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5599 * gimple-ssa-store-merging.c
5600 (imm_store_chain_info::output_merged_store): Remove redundant create,
5601 release split_store vector contents on failure.
5602 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5603 scalar stmt vector on cache hit.
5605 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5607 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5609 * config.gcc (powerpc-xilinx-eabi*): Remove.
5610 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5612 (fusion_addis_mem_combo_load): Ditto.
5613 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5615 (rs6000_cpu_cpp_builtins): Ditto.
5616 * config/rs6000/rs6000-linux.c
5617 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5618 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5619 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5621 (rs6000_setup_reg_addr_masks): Ditto.
5622 (rs6000_init_hard_regno_mode_ok): Ditto.
5623 (rs6000_option_override_internal): Ditto.
5624 (legitimate_lo_sum_address_p): Ditto.
5625 (rs6000_legitimize_address): Ditto.
5626 (rs6000_legitimize_reload_address): Ditto.
5627 (rs6000_legitimate_address_p): Ditto.
5628 (abi_v4_pass_in_fpr): Ditto.
5629 (setup_incoming_varargs): Ditto.
5630 (rs6000_gimplify_va_arg): Ditto.
5631 (rs6000_split_multireg_move): Ditto.
5632 (rs6000_savres_strategy): Ditto.
5633 (rs6000_emit_prologue_components): Ditto.
5634 (rs6000_emit_epilogue_components): Ditto.
5635 (rs6000_emit_prologue): Ditto.
5636 (rs6000_emit_epilogue): Ditto.
5637 (rs6000_elf_file_end): Ditto.
5638 (rs6000_function_value): Ditto.
5639 (rs6000_libcall_value): Ditto.
5640 * config/rs6000/rs6000.h: Ditto.
5641 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5642 (TARGET_MINMAX): ... this. New.
5643 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5644 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5645 (*movsi_internal1_single): Delete.
5646 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5647 mfpu=, mxilinx-fpu): Delete.
5648 * config/rs6000/singlefp.h: Delete.
5649 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5650 * config/rs6000/t-rs6000: Ditto.
5651 * config/rs6000/t-xilinx: Delete.
5652 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5653 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5654 (VStype_simple): Delete.
5655 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5656 * config/rs6000/xfpu.h: Delete.
5657 * config/rs6000/xfpu.md: Delete.
5658 * config/rs6000/xilinx.h: Delete.
5659 * config/rs6000/xilinx.opt: Delete.
5660 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5661 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5663 2018-05-04 Tom de Vries <tom@codesourcery.com>
5666 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5669 2018-05-04 Richard Biener <rguenther@suse.de>
5672 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5673 (expand_complex_div_wide): Likewise.
5674 (expand_complex_operations_1): Likewise.
5675 (expand_complex_libcall): Preserve EH info of the original stmt.
5676 (tree_lower_complex): Handle removed blocks.
5677 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5678 on complex multiplication and division libcall builtins.
5680 2018-05-04 Richard Biener <rguenther@suse.de>
5683 * fold-const.c (negate_expr_p): Restrict negation of operand
5684 zero of a division to when we know that can happen without
5686 (fold_negate_expr_1): Likewise.
5688 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5691 * real.h (real_nextafter): Declare.
5692 * real.c (real_nextafter): New function.
5693 * fold-const-call.c (fold_const_nextafter): New function.
5694 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5695 CASE_CFN_NEXTTOWARD.
5696 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5697 even when arg1_mode is different from arg0_mode.
5699 2018-05-03 Nathan Sidwell <nathan@acm.org>
5701 * doc/extend.texi (Deprecated Features): Remove
5703 (Backwards Compatibility): Likewise.
5704 * doc/invoke.texi (C++ Language Options): Likewise.
5705 (C++ Dialect Options): Likewise.
5707 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5710 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5711 _mm512_mask_mullox_epi64): New intrinsics.
5713 2018-05-03 Tom de Vries <tom@codesourcery.com>
5716 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5717 dump files): Add offload-tree.
5719 2018-05-03 Richard Biener <rguenther@suse.de>
5721 PR tree-optimization/85615
5722 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5723 to loops not nested in BBs loop father to avoid creating multi-entry
5726 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5728 PR tree-optimization/70291
5729 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5730 arguments. Change return type to tree. Emit libcall as a new
5731 statement rather than replacing existing one when inplace_p is true.
5732 (expand_complex_multiplication_components): New function.
5733 (expand_complex_multiplication): Expand floating-point complex
5734 multiplication using the above.
5735 (expand_complex_division): Rename inner_type parameter to type.
5736 Update expand_complex_libcall call-site.
5737 (expand_complex_operations_1): Update expand_complex_multiplication
5738 and expand_complex_division call-sites.
5740 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5743 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5744 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5745 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5746 the highest significant bit of the shift count mask is clear. In
5747 check whether and[sq]i3 is needed verify that all significant bits
5748 of the shift count other than the highest are set.
5750 2018-05-02 Tom de Vries <tom@codesourcery.com>
5753 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5754 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5755 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5756 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5757 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5758 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5759 __builtin_goacc_parlevel_size.
5761 2018-05-02 Richard Biener <rguenther@suse.de>
5763 PR tree-optimization/85597
5764 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5765 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5768 2018-05-02 Tom de Vries <tom@codesourcery.com>
5771 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5772 dump files): Add ltrans-tree.
5774 2018-05-02 Tom de Vries <tom@codesourcery.com>
5777 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5778 dump files): Add wpa-ipa.
5780 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5782 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5783 powerpc*-*-linux*paired* target.
5784 * config/rs6000/750cl.h: Delete.
5785 * config/rs6000/paired.h: Delete.
5786 * config/rs6000/paired.md: Delete.
5787 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5789 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5790 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5791 comment. Remove paired float support.
5792 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5793 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5795 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5796 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5798 * config/rs6000/rs6000.c: Remove paired float support.
5799 (paired_expand_vector_init, paired_expand_vector_move,
5800 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5801 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5802 paired_expand_builtin, paired_expand_predicate_builtin,
5803 paired_init_builtins): Delete.
5804 * config/rs6000/rs6000.h: Remove paired float support.
5805 * config/rs6000/rs6000.md: Remove paired float support.
5806 (move_from_CR_ov_bit): Delete.
5807 * config/rs6000/rs6000.opt (mpaired): Delete.
5808 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5809 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5811 2018-05-02 Richard Biener <rguenther@suse.de>
5814 * gimplify.c (gimplify_save_expr): When in SSA form allow
5815 SAVE_EXPRs to compute to SSA vars.
5817 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5820 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5821 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5822 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5823 clobber operands[2], instead use a new pseudo. Formatting fixes.
5825 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5827 PR tree-optimization/85586
5828 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5829 exit early for statements in the same group if the accesses are
5832 2018-05-02 Tom de Vries <tom@codesourcery.com>
5835 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5838 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5840 PR tree-optimization/85143
5841 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5843 2018-05-01 Tom de Vries <tom@codesourcery.com>
5846 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5847 not found" error message.
5849 2018-05-01 Tom de Vries <tom@codesourcery.com>
5852 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5853 * vec.c (test_ordered_remove_if): New function.
5854 (vec_c_tests): Call test_ordered_remove_if.
5855 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5856 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5857 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5858 VEC_ORDERED_REMOVE_IF.
5860 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5862 PR tree-optimization/82665
5863 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5864 pointer subtraction where arguments come from a memchr call.
5866 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5868 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5869 --push-state --as-needed and --pop-state instead of --as-needed and
5870 --no-as-needed if ld supports it.
5871 * configure: Regenerated.
5874 * doc/install.texi2html: Replace _002d with - and _002a with * in
5875 generated html files using sed.
5877 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5880 * gcc-rich-location.c (blank_line_before_p): New function.
5881 (use_new_line): New function.
5882 (gcc_rich_location::add_fixit_insert_formatted): New function.
5883 * gcc-rich-location.h
5884 (gcc_rich_location::add_fixit_insert_formatted): New function.
5886 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5888 * selftest.c (assert_streq): Rename "expected" and "actual" to
5889 "val1" and "val2". Extend NULL-handling to cover both inputs
5890 symmetrically, while still requiring both to be non-NULL for a pass.
5891 * selftest.h (assert_streq): Rename "expected" and "actual" to
5893 (ASSERT_EQ): Likewise.
5894 (ASSERT_EQ_AT): Likewise.
5895 (ASSERT_KNOWN_EQ): Likewise.
5896 (ASSERT_KNOWN_EQ_AT): Likewise.
5897 (ASSERT_NE): Likewise.
5898 (ASSERT_MAYBE_NE): Likewise.
5899 (ASSERT_MAYBE_NE_AT): Likewise.
5900 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5901 the assertion to pass.
5902 (ASSERT_STREQ_AT): Likewise.
5904 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5906 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5907 interaction with -pie.
5909 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5911 * selftest.h: Fix alphabetization of per-source-file selftest
5914 2018-04-30 Jason Merrill <jason@redhat.com>
5916 PR c++/61982 - dead stores to destroyed objects.
5917 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5920 2018-04-30 Jason Merrill <jason@redhat.com>
5922 * tree.c (build_clobber): New.
5923 * tree.h: Declare it.
5924 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5926 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5928 * diagnostic-show-locus.c (layout::layout): Update for
5929 location_get_source_line returning a char_span.
5930 (struct char_span): Move to input.h.
5931 (struct correction): Update for fields in char_span becoming
5933 (struct source_line): Update for location_get_source_line
5934 returning a char_span.
5935 (layout::print_line): Likewise.
5936 * edit-context.c (edited_file::print_content): Likewise.
5937 (edited_file::print_diff_hunk): Likewise.
5938 (edited_file::print_run_of_changed_lines): Likewise.
5939 (edited_file::get_num_lines): Likewise.
5940 (edited_line::edited_line): Likewise.
5941 * final.c (asm_show_source): Likewise.
5942 * input.c (location_get_source_line): Convert return type
5943 from const char * to char_span, losing the final "line_len"
5945 (dump_location_info): Update for the above.
5946 (get_substring_ranges_for_loc): Likewise. Use a char_span
5947 when handling the literal within the line.
5948 (test_reading_source_line): Update for location_get_source_line
5949 returning a char_span.
5950 * input.h (class char_span): Move here from
5951 diagnostic-show-locus.c, converting from a struct to a class.
5952 Make data members private.
5953 (char_span::operator bool): New.
5954 (char_span::length): New.
5955 (char_span::get_buffer): New.
5956 (char_span::operator[]): New.
5957 (char_span::subspan): Make const.
5958 (char_span::xstrdup): New.
5959 (location_get_source_line): Convert return type from const char *
5960 to char_span, losing the final "line_size" param.
5962 2018-04-30 Jan Hubicka <jh@suse.cz>
5964 * lto-wrapper.c (ltrans_priorities): New static var.
5965 (cmp_priority): New.
5966 (run_gcc): Read priorities and if doing parallel build order
5967 the Makefile by them.
5969 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5971 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5973 2018-04-30 Richard Biener <rguenther@suse.de>
5975 * tree-cfg.c (verify_address): Remove base argument, add
5976 flag whether to check TREE_ADDRESSABLE and do that.
5977 (verify_expr): Remove.
5978 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5979 (verify_gimple_assign_single): Likewise.
5980 (verify_gimple_switch): Likewise.
5981 (verify_expr_location_1): Dereference tp once. Add (disabled)
5982 piece from verify_expr.
5983 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5985 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5987 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5989 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5991 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5992 (small_data_pattern): Likewise.
5993 (arc_rewrite_small_data): Likewise.
5994 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5995 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5996 (get_symbol_alignment): New function.
5997 (legitimate_small_data_address_p): Likewise.
5998 (legitimate_scaled_address): Update, call
5999 legitimate_small_data_address_p.
6000 (output_sdata): New static variable.
6001 (arc_print_operand): Update how we handle small data operands.
6002 (arc_print_operand_address): Likewise.
6003 (arc_legitimate_address_p): Update, use
6004 legitimate_small_data_address_p.
6005 (arc_rewrite_small_data_p): Remove.
6006 (arc_rewrite_small_data_1): Likewise.
6007 (arc_rewrite_small_data): Likewise.
6008 (small_data_pattern): Likewise.
6009 (compact_sda_memory_operand): Update to use
6010 legitimate_small_data_address_p and get_symbol_alignment.
6011 (prepare_move_operands): Don't rewite sdata pattern.
6012 (prepare_extend_operands): Remove.
6013 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
6015 (zero_extendqisi2): Likewise.
6016 (zero_extendhisi2): Likewise.
6017 (extendqihi2): Likewise.
6018 (extendqisi2): Likewise.
6019 (extendhisi2): Likewise.
6023 * config/arc/constraints.md (Usd): Change it to memory constraint.
6025 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
6027 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
6028 as source of std instructions.
6029 * config/arc/arc.md (movsi_insn): Update pattern predicate to
6030 allow 6-bit constants as source for store instructions.
6031 (movdi_insn): Update instruction pattern to allow 6-bit constants
6032 as source for store instructions.
6034 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
6036 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
6038 2018-04-30 Nathan Sidwell <nathan@acm.org>
6039 Sandra Loosemore <sandra@codesourcery.com>
6041 * dumpfile.c (dump_open): Allow '-' for stdout.
6042 * doc/invoke.texi (Developer Options): Document dump filename
6043 determination early. Document stdin/stdout selection.
6045 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
6047 Microblaze Target: PIC data text relative
6049 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
6050 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
6052 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
6053 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
6054 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
6055 New addressing mode for data-text relative position indepenedent code.
6056 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
6057 'ADDRESS_SYMBOLIC_TXT_REL'.
6058 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
6059 (microblaze_legitimate_pic_operand): Exclude function calls from
6060 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
6061 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
6063 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6064 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
6065 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
6066 for 'address + offset'.
6067 (microblaze_expand_prologue): Add new function prologue call for
6069 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
6070 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
6071 table in case of TARGET_PIC_DATA_TEXT_REL.
6072 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
6073 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
6074 Add new macros 'UNSPEC_TEXT',
6075 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
6076 + exclude function calls from 'UNSPEC_PLT' in case of data text
6078 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6079 new target hook for generating address diff vector tables in case of
6081 * doc/tm.texi : Regenerate.
6082 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
6083 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
6084 of addr diff vector generation.
6085 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
6086 target hook definition.
6087 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
6088 Add default function for generate_pic_addr_diff_vec -> flag_pic.
6089 * doc/invoke.texi (Add new pic option): Add new microblaze pic
6090 option for data text relative.
6092 2018-04-30 Richard Biener <rguenther@suse.de>
6094 * tree-chrec.h (evolution_function_is_constant_p): Remove
6096 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
6098 2018-04-30 Richard Biener <rguenther@suse.de>
6101 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6103 2018-04-30 Richard Biener <rguenther@suse.de>
6105 PR tree-optimization/28364
6106 PR tree-optimization/85275
6107 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6108 copying first exit test.
6110 2018-04-28 Mark Wielaard <mark@klomp.org>
6112 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6114 (dwarf_AT): Handle DW_AT_addr_base.
6115 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6117 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6120 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6121 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6122 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6123 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6125 2018-04-28 Richard Biener <rguenther@suse.de>
6127 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6128 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6129 to reflect use. Only add interesting stmts.
6131 2018-04-27 Martin Jambor <mjambor@suse.cz>
6134 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6135 the jump function allows for passing through aggregate values.
6137 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6139 * input.h (in_system_header_at): Convert from macro to inline
6141 (from_macro_expansion_at): Likewise.
6142 (from_macro_definition_at): Likewise.
6144 2018-04-27 Jeff Law <law@redhat.com>
6146 * config.gcc: Mark tile* targets as deprecated/obsolete.
6148 2018-04-27 Richard Biener <rguenther@suse.de>
6150 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6153 2018-04-27 Richard Biener <rguenther@suse.de>
6155 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6157 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6159 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6160 with Yd constraint. Set "preferred_for_speed" attribute from
6161 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6163 (*movdi_internal): Ditto.
6164 (movti_interunit splitters): Remove
6165 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6166 (movdi_interunit splitters): Ditto.
6167 * config/i386/constraints.md (Ye): Remove.
6168 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6170 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6173 * config/aarch64/constraints.md (Usg): Limit to 31.
6176 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6178 PR tree-optimization/85529
6179 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6180 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6181 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6182 zero extension or masking of the MSB bit.
6183 (optimize_range_tests): Add FIRST_BB argument, pass it through
6184 to optimize_range_tests_var_bound.
6185 (maybe_optimize_range_tests, reassociate_bb): Adjust
6186 optimize_range_tests callers.
6188 2018-04-26 Richard Biener <rguenther@suse.de>
6189 Jakub Jelinek <jakub@redhat.com>
6191 * cgraph.h (symbol_table): Just declare debug method here.
6192 * symtab.c (symbol_table::debug): Define.
6194 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6196 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6198 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6200 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6201 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6202 (*movdi_internal): Substitute Yi and Yj constraint with x
6203 and Ym and Yn constraint with y constraint. Update "isa"
6204 attribute and set "preferred_for_speed" attribute from
6205 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6206 (*movsi_internal): Ditto.
6207 (*movdf_internal): Ditto.
6208 (*movsf_internal): Ditto.
6209 (*zero_extendsidi2): Ditto.
6210 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6211 (sse2_loadld): Ditto.
6212 (*vec_extract<ssevecmodelower>_0): Ditto.
6213 (*vec_extractv4si_0_zext_sse4): Ditto.
6214 (vec_concatv2di): Ditto.
6215 (*vec_dup<mode>): Ditto.
6216 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6217 * config/i386/constraints.md (Yi): Remove.
6222 2018-04-26 Nathan Sidwell <nathan@acm.org>
6224 * dumpfile.c (dump_open): New.
6225 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6226 (dump_finish): Detect stdio/stderr by value not name.
6228 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6230 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6232 2018-04-26 Tom de Vries <tom@codesourcery.com>
6235 * config/nvptx/nvptx.c (verify_neutering_jumps)
6236 (verify_neutering_labels): New function
6237 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6239 2018-04-26 Tom de Vries <tom@codesourcery.com>
6242 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6243 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6246 2018-04-26 Richard Biener <rguenther@suse.de>
6247 Tom de Vries <tom@codesourcery.com>
6250 * lto-streamer-out.c (output_function): Fixup loops if required to match
6251 discovery done in the reader.
6253 2018-04-26 Richard Biener <rguenther@suse.de>
6255 PR tree-optimization/85116
6256 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6257 have a loop exit from the single latch predecessor. Remove
6258 case of header with just condition.
6259 (ch_base::copy_headers): Exclude infinite loops from any
6261 (pass_ch::execute): Record exits.
6263 2018-04-26 Richard Biener <rguenther@suse.de>
6265 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6266 prologue cost vector and pass it to vect_get_load_cost.
6267 (vect_get_peeling_costs_all_drs): Likewise.
6268 (vect_peeling_hash_get_lowest_cost): Likewise.
6269 (vect_enhance_data_refs_alignment): Likewise.
6271 2018-04-26 Richard Biener <rguenther@suse.de>
6274 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6275 checking of integer<->pointer conversions.
6276 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6277 sign-/zero-extending pointer types.
6278 (expand_omp_for_static_chunk): Likewise.
6280 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6281 Jean Lee <xiaoyur347@gmail.com>
6283 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6284 (TARGET_ASAN_SHADOW_OFFSET): Define.
6285 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6286 true for -fsanitize=address.
6288 2018-04-25 Mark Wielaard <mark@klomp.org>
6290 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6293 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6295 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6296 than "alu", remove explicit "memory" and "imm_disp" attributes.
6297 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6300 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6301 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6302 gen_lowpart_no_emit.
6304 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6307 * config/i386/i386.c (ix86_expand_builtin): Change memory
6308 operand to XI, extend p0 to Pmode.
6309 * config/i386/i386.md: Change unspec volatile and operand
6310 1 mode to XI, change operand 0 mode to P.
6312 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6314 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6315 GET_MODE_MASK before any checking.
6316 (nds32_can_use_bset_p): Likewise.
6317 (nds32_can_use_btgl_p): Likewise.
6319 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6321 * config/nds32/nds32-doubleword.md: New define_split pattern for
6322 illegal register number.
6324 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6326 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6328 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6330 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6332 2018-04-25 Richard Biener <rguenther@suse.de>
6334 * lto-streamer.h (LTO_major_version): Bump to 8.
6336 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6338 * BASE-VER: Set to 9.0.0.
6340 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6342 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6343 in __abskf2 and __powikf2.
6345 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6348 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6349 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6350 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6351 Use the above on operand 2. Reindent.
6352 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6354 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6357 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6359 * config/i386/i386.opt (mcet): Removed.
6360 * doc/install.texi: Remove -mcet documentation.
6361 * doc/invoke.texi: Likewise.
6363 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6366 * doc/install.texi: Remove -mcet from bootstrap-cet.
6368 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6371 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6372 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6376 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6377 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6378 containing a CONST_VECTOR.
6380 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6382 * doc/install.texi: Update newlib dependency for nvptx.
6384 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6387 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6388 instead of INTVAL when shifting x left.
6390 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6392 PR tree-optimization/85478
6393 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6394 vect_grouped_store_supported for single element vectors.
6396 2018-04-24 Richard Biener <rguenther@suse.de>
6399 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6400 load cost increase to the case of non-constant step.
6402 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6405 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6406 destination if any_malformed_asm.
6408 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6411 * expr.c (store_field): In the bitfield case, if the value comes from
6412 a function call and is returned in registers by means of a PARALLEL,
6413 do not change the mode of the temporary unless BLKmode and VOIDmode.
6415 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6417 PR rtl-optimization/85423
6418 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6419 dependencies to debug insns when the previous insn is non-debug.
6421 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6423 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6424 enums into a single definition.
6425 (fls): Fix predicates and printing.
6428 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6430 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6431 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6432 and short u6 immediate.
6433 (check_if_valid_sleep_operand): Remove.
6434 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6436 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6438 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6439 flag_always_save_lp condition.
6440 * config/nds32/nds32.opt (malways-save-lp): New option.
6442 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6444 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6445 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6446 * config/nds32/nds32.h
6447 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6448 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6450 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6452 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6453 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6455 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6456 Chung-Ju Wu <jasonwucj@gmail.com>
6458 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6460 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6461 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6463 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6465 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6467 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6469 * config/nds32/nds32-protos.h (nds32_data_alignment,
6470 nds32_local_alignment): Declare.
6471 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6472 nds32_local_alignment): New functions.
6473 (TARGET_CONSTANT_ALIGNMENT): Define.
6474 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6476 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6478 * config/nds32/nds32.c
6479 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6480 (TARGET_MODES_TIEABLE_P): Likewise.
6482 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6484 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6487 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6488 Chung-Ju Wu <jasonwucj@gmail.com>
6490 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6491 for unaligned access.
6492 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6494 * config/nds32/nds32-intrinsic.md: Likewise.
6495 * config/nds32/nds32_intrinsic.h: Likewise.
6496 * config/nds32/nds32.h (nds32_builtins): Likewise.
6497 * config/nds32/nds32.opt (munaligned-access): New option.
6498 * config/nds32/nds32.c (nds32_asm_file_start): Display
6499 flag_unaligned_access status.
6501 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6503 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6504 -mno-relax is present.
6505 * config/riscv/linux.h (LINK_SPEC): Ditto.
6507 2018-04-20 Martin Sebor <msebor@redhat.com>
6510 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6512 (gimple_fold_builtin_stxcpy_chk): Same.
6513 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6515 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6518 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6519 __powikf2 when long double is IEEE 128-bit.
6521 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6523 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6524 step to make sure stack always aligned.
6526 2018-04-20 Carl Love <cel@us.ibm.com>
6529 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6530 size check for arg0.
6532 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6533 Tom de Vries <tom@codesourcery.com>
6536 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6537 Emit insns for calls too.
6538 (nvptx_find_par): Always look for worker-level predecessor insn.
6539 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6541 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6542 (nvptx_process_pars): Propagate frames for calls.
6544 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6547 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6549 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6550 (ix86_handle_option): Don't handle OPT_mibt.
6551 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6553 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6555 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6556 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6557 (ix86_target_macros): Define __CET__ with flag_cf_protection
6558 for -fcf-protection.
6559 * config/i386/i386.c (isa2_opts): Remove -mibt.
6560 * config/i386/i386.h (TARGET_IBT): Removed.
6561 (TARGET_IBT_P): Likewise.
6562 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6563 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6564 * config/i386/i386.opt (mcet): Update help message.
6567 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6568 -mcet as an alias for -mshstk.
6570 2018-04-20 Richard Biener <rguenther@suse.de>
6573 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6574 complexity by forcing a single use of the multiply operand.
6576 2018-04-20 Martin Jambor <mjambor@suse.cz>
6579 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6580 recursion dependency to only apply to non-clones.
6582 2018-04-20 Martin Jambor <mjambor@suse.cz>
6585 * ipa-cp.c (create_specialized_node): Check that clones of
6586 self-recursive edges exist during IPA-CP.
6588 2018-04-19 Toon Moene <toon@moene.org>
6590 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6593 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6595 PR tree-optimization/85467
6596 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6597 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6598 VECTOR_CST element to type.
6600 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6603 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6604 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6605 (builtin_longjmp): Likewise.
6606 (save_stack_nonlocal): New pattern.
6607 (restore_stack_nonlocal): Likewise.
6609 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6612 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6613 Replace ASM_OUTPUT_LABEL with fprintf.
6615 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6618 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6619 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6620 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6621 define __IBT__ and __SHSTK__ for -fcf-protection.
6622 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6624 (ix86_trampoline_init): Likewise.
6625 (x86_output_mi_thunk): Likewise.
6626 (ix86_notrack_prefixed_insn_p): Likewise.
6627 (ix86_option_override_internal): Don't disallow -fcf-protection.
6628 * config/i386/i386.md (rdssp<mode>): Also enable for
6630 (incssp<mode>): Likewise.
6631 (nop_endbr): Likewise.
6632 * config/i386/i386.opt (mcet): Change help message to built-in
6636 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6637 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6638 enable CET built-in functions.
6640 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6642 * common/config/i386/i386-common.c
6643 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6644 OPTION_MASK_ISA_MOVDIRI_UNSET,
6645 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6646 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6647 * config.gcc (movdirintrin.h): New header.
6648 * config/i386/cpuid.h (bit_MOVDIRI,
6649 bit_MOVDIR64B): New bits.
6650 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6652 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6653 (VOID, PVOID, PCVOID)): New function types.
6654 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6655 __builtin_ia32_directstoreu_u64,
6656 __builtin_ia32_movdir64b): New builtins.
6657 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6658 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6660 (ix86_valid_target_attribute_inner_p): Ditto.
6661 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6662 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6663 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6664 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6665 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6666 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6667 (movdiri<mode>, movdir64b_<mode>): New.
6668 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6669 * config/i386/immintrin.h: Include movdirintrin.h.
6670 * config/i386/movdirintrin.h: New file.
6671 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6673 2018-04-19 Richard Biener <rguenther@suse.de>
6676 * cfg.c (clear_bb_flags): When loop state says we have
6677 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6679 2018-04-19 Richard Biener <rguenther@suse.de>
6681 PR tree-optimization/84737
6682 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6683 copying restrict info.
6684 (vect_setup_realignment): Use it.
6685 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6686 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6687 the first DR to all generated stores.
6688 (vectorizable_load): Likewise for loads.
6690 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6692 PR tree-optimization/85446
6693 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6694 the integral and pointer types to have the same precision.
6696 * doc/install.texi: Document --disable-cet being the default and
6699 2018-04-18 Martin Liska <mliska@suse.cz>
6701 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6704 2018-04-18 Martin Liska <mliska@suse.cz>
6707 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6710 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6711 arguments if they are comparable.
6713 2018-04-18 Martin Liska <mliska@suse.cz>
6716 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6719 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6722 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6725 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6726 ENDBR after calling __morestack.
6728 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6731 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6732 by using gcc_base_ver to generate a gcc_driver_version, and use
6733 it when generating GCC_DRIVER_NAME.
6734 * configure: Regenerate.
6736 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6739 * config.gcc: Obsolete powerpc*-*-*spe*.
6741 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6744 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6745 (stabstr_D): Change type of unum from unsigned int to
6746 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6749 2018-04-17 Jim Wilson <jimw@sifive.com>
6752 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6753 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6754 Set arg_pointer_offset after using pretend_args_size.
6756 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6758 PR rtl-optimization/85431
6759 * dse.c (record_store): Ignore zero width stores.
6762 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6763 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6764 __builtin_stack_restore rather than after it.
6765 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6766 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6767 argument instead of virtual_dynamic_stack_rtx.
6769 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6771 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6773 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6774 Add note to error message to explain internal mapping of overloaded
6775 built-in function name to non-overloaded built-in function name.
6776 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6779 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6782 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6783 where the inputs overlap with the output.
6785 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6787 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6788 (=v, v) alternative and explicit "memory" attribute.
6789 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6790 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6792 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6793 "sselog1" type instead of "sselog".
6794 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6795 "sselog". Remove explicit "memory" attribute.
6796 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6797 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6799 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6800 "sselog1" type instead of "sselog". Remove explicit "memory"
6802 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6803 use "sselog1" type instead of "sselog". Remove explicit "memory"
6805 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6806 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6808 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6809 "sselog1" type instead of "sselog". Remove explicit "memory"
6811 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6812 use "sselog1" type instead of "sselog". Remove explicit "memory"
6816 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6819 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6822 2018-04-17 Martin Jambor <mjambor@suse.cz>
6825 * ipa-cp.c (create_specialized_node): Call
6826 expand_all_artificial_thunks if necessary.
6828 2018-04-17 Martin Liska <mliska@suse.cz>
6831 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6832 in message, remote space in between '_G' and '('.
6834 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6837 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6838 avx512f_vmcmp<mode>3<round_saeonly_name>,
6839 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6840 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6841 avx512f_rndscale<mode><round_saeonly_name>,
6842 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6843 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6844 Use %<iptr>2 instead of %2 for -masm=intel.
6845 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6846 avx512f_vcvttss2usi<round_saeonly_name>,
6847 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6849 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6850 avx512f_vcvttsd2usi<round_saeonly_name>,
6851 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6852 Use %q1 instead of %1 for -masm=intel.
6853 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6854 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6855 of %3 for -masm=intel.
6856 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6858 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6860 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6862 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6863 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6865 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6867 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6868 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6869 %g1 and one with %0 and %1.
6870 (avx512er_vmrcp28<mode><round_saeonly_name>,
6871 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6873 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6874 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6875 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6876 of %0 and %{%4%} for -masm=intel.
6877 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6878 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6879 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6880 order of %0 and %{%5%}%{z%} for -masm=intel.
6882 2018-04-17 Jan Hubicka <jh@suse.cz>
6885 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6887 2018-04-17 Martin Liska <mliska@suse.cz>
6890 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6891 for target_clone error message. Make default implementation
6892 clone to be a local declaration.
6893 (separate_attrs): Add new argument and check for an empty
6895 (expand_target_clones): Handle it.
6896 (ipa_target_clone): Make redirection just for target_clones
6899 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6900 Tom de Vries <tom@codesourcery.com>
6903 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6904 tiled basic blocks without omp continue statements.
6906 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6909 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6910 vec_extract expression as having side effects to make sure it gets
6913 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6916 * config/i386/i386.c (get_builtin_code_for_version): Check
6919 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6922 * gcc/config.gcc: Support "skylake".
6923 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6925 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6926 (processor_target_table): Add "skylake".
6927 (ix86_option_override_internal): Add "skylake".
6928 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6929 PROCESSOR_CANNONLAKE.
6930 (get_builtin_code_for_version): Fix priority for
6931 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6932 PROCESSOR_SKYLAKE-AVX512.
6933 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6934 (processor_type): Add PROCESSOR_SKYLAKE.
6936 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6937 Jason Merrill <jason@redhat.com>
6940 * convert.c (convert_to_integer_1): Use direct recursion for
6941 enumeral types and types with a precision less than the number
6942 of bits in their mode.
6944 2018-04-16 Julia Koval <julia.koval@intel.com>
6947 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6948 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6950 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6953 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6954 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6955 and -mno-direct-move.
6957 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6960 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6961 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6962 Ensure negative shifts result in {0}.
6964 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6966 PR rtl-optimization/79916
6967 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6968 regs (if any) to define how to gnerate SD moves when LRA is in
6971 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6973 PR rtl-optimization/85393
6974 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6975 * except.c (expand_dw2_landing_pad_for_region): Make static.
6976 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6977 a label and unconditional jump to old_bb, rather than
6978 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6981 PR rtl-optimization/85376
6982 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6983 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6984 instead of a specific value.
6986 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6987 Bin Cheng <bin.cheng@arm.com>
6989 PR tree-optimization/82965
6990 PR tree-optimization/83991
6991 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6992 by_profile_only parameter.
6993 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6994 information if the loop was predicted to iterate too many times.
6995 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6997 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
7000 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
7003 2018-04-13 Martin Liska <mliska@suse.cz>
7004 Jakub Jelinek <jakub@redhat.com>
7007 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
7008 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
7009 * builtins.c (expand_builtin_memory_copy_args): Use
7010 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
7011 handle dest_addr == pc_rtx.
7013 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
7016 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
7017 asked to not generate direct moves.
7018 (fix_trunc<mode>si2_stfiwx): Similar.
7019 (fix_trunc<mode>si2_internal): Similar.
7021 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7024 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
7025 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
7026 lookup if dest in some wider mode is known to be const0_rtx and
7027 if so, record permanent equivalence for it to be ZERO_EXTEND of
7028 the narrower mode destination.
7030 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7032 * lto-streamer-out.c (output_function): Revert 259346.
7033 * omp-expand.c (expand_oacc_for): Likewise.
7035 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
7037 PR rtl-optimization/85354
7038 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
7039 * sel-sched.c (sel_global_init): ... here.
7041 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
7044 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
7045 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
7046 mode for PE-COFF targets.
7047 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
7048 (i386_pe_asm_lto_end): Likewise.
7049 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
7050 (TARGET_ASM_LTO_END): Likewise.
7051 * config/i386/winnt.c (saved_debug_info_level): New static variable.
7052 (i386_pe_asm_lto_start): New function.
7053 (i386_pe_asm_lto_end): Likewise.
7055 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
7056 Richard Biener <rguenther@suse.de>
7059 * lto-streamer-out.c (output_function): Fix CFG loop state before
7061 * omp-expand.c (expand_oacc_for): Handle calls to internal
7062 functions like regular functions.
7064 2018-04-12 Richard Biener <rguenther@suse.de>
7067 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
7068 for the early LTO debug to properly generate references to it
7069 during DIE emission. Do not re-use that for the skeleton for
7071 (dwarf2out_early_finish): Likewise.
7073 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7076 * config/i386/sse.md
7077 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
7078 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7079 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
7080 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
7081 and output is a reg, avoid creating invalid lowpart subreg, but
7082 instead split into a 512-bit move. Don't split if not AVX512VL,
7083 input is xmm16+ reg and output is a mem.
7084 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
7085 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
7086 xmm16+ reg and output is a mem.
7088 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7090 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
7091 also for flag_dwarf2_cfi_asm.
7093 2018-04-12 Jakub Jelinek <jakub@redhat.com>
7095 PR rtl-optimization/85342
7096 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
7097 a bool scalar var inside of the loop instead. Don't try to update
7098 recog_data.operand after failed apply_change_group.
7100 2018-04-12 Tom de Vries <tom@codesourcery.com>
7103 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7104 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7105 array with flexible array member as array without given dimension.
7106 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7107 argument for undefined param to true.
7109 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7112 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7113 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7114 from PowerPC section.
7115 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7116 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7117 help text that is too long.
7118 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7119 help text that is too long.
7120 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7121 help text that is too long.
7123 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7125 * config/alpha/alpha.md (stack_probe_internal): Rename
7126 from "probe_stack". Update all callers.
7128 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7130 PR rtl-optimization/84566
7131 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7132 sched_macro_fuse_insns.
7134 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7137 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7138 (compute_block_dependences): ... from here.
7140 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7142 PR tree-optimization/85331
7143 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7144 from int to HOST_WIDE_INT.
7146 2018-04-11 Martin Jambor <mjambor@suse.cz>
7149 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7150 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7151 not the same as the source val.
7152 (cgraph_edge_brings_value_p): New parameter.
7153 (gather_edges_for_value): Pass destination value to
7154 cgraph_edge_brings_value_p.
7155 (perhaps_add_new_callers): Likewise.
7156 (get_info_about_necessary_edges): Likewise and exclude values brought
7157 only by self-recursive edges.
7158 (create_specialized_node): Redirect only clones of self-calling edges.
7159 (+self_recursive_pass_through_p): New function.
7160 (find_more_scalar_values_for_callers_subset): Use it.
7161 (find_aggregate_values_for_callers_subset): Likewise.
7162 (known_aggs_to_agg_replacement_list): Removed.
7163 (decide_whether_version_node): Re-calculate known constants for all
7164 remaining context clones.
7166 2018-04-11 Richard Biener <rguenther@suse.de>
7169 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7170 from early DWARF output.
7171 (dwarf2out_early_finish): Output line info unconditionally into
7172 early DWARF and add reference to it.
7174 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7177 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7178 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7179 other than V2DFmode using iptr mode attribute.
7180 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7182 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7184 PR rtl-optimization/84659
7185 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7187 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7190 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7192 (output_loc_list): Pass address of a dummy size variable even in the
7193 locview handling loop.
7194 (index_location_lists): Add comment on why skip_loc_list_entry can't
7197 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7200 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7203 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7206 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7207 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7208 and -mstring-compare-inline-limit.
7210 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7213 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7214 for stack clash protection in a register whenever we need it to be in
7217 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7219 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7220 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7222 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7225 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7227 (mlong-double-): Ditto.
7228 * config/rs6000/sysv4.opt (msdata=): Ditto.
7229 (mtls-size=): Ditto.
7231 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7233 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7234 erroneous entries for
7235 "vector int vec_ldl (int, long int *)", and
7236 "vector unsigned int vec_ldl (int, unsigned long int *)".
7237 Add comments and entries for
7238 "vector bool char vec_ldl (int, bool char *)",
7239 "vector bool short vec_ldl (int, bool short *)",
7240 "vector bool int vec_ldl (int, bool int *)",
7241 "vector bool long long vec_ldl (int, bool long long *)",
7242 "vector pixel vec_ldl (int, pixel *)",
7243 "vector long long vec_ldl (int, long long *)",
7244 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7245 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7246 type tree bool_long_long_type_node and correct definition of
7247 bool_V2DI_type_node to make reference to this new type tree.
7248 (rs6000_mangle_type): Replace erroneous reference to
7249 bool_long_type_node with bool_long_long_type_node.
7250 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7251 comments to emphasize sign distinctions for char and int types and
7252 replace RS6000_BTI_bool_long constant with
7253 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7254 use of RS6000_BTI_pixel.
7255 (bool_long_type_node): Remove this macro definition.
7256 (bool_long_long_type_node): New macro definition
7258 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7260 PR rtl-optimization/85300
7261 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7262 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7263 simplify_unary_operation fails.
7265 2018-04-10 Martin Liska <mliska@suse.cz>
7267 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7268 cgraph_edge and ipa_ref.
7270 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7274 * config/i386/sse.md
7275 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7276 computation of the VEC_MERGE selector from mask.
7277 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7278 Fix decoding of the VEC_MERGE selector into mask.
7280 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7282 PR tree-optimization/85286
7283 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7285 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7287 * final.c (final_1): Set insn_last_address as well as
7288 insn_current_address.
7290 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7293 * explow.c (emit_stack_probe): Call validize_mem on memory location
7294 before passing it to gen_probe_stack. Create address operand and
7295 legitimize it for the probe_stack_address case.
7297 2018-04-09 Jan Hubicka <jh@suse.cz>
7300 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7301 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7302 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7304 (free_lang_data): Rebuild type inheritance graph.
7306 2018-04-09 Martin Sebor <msebor@redhat.com>
7308 * invoke.texi (-finline-small-functions): Mention other optimization
7310 (-findirect-inlining, -fpartial-inlining): Same.
7311 (-finline-functions-called-once): Same.
7312 (-freorder-blocks-and-partition): Same.
7314 2018-04-09 Jan Hubicka <jh@suse.cz>
7317 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7318 jumps; choose last target that matches the criteria (i.e.
7319 no partition changes for non-crossing jumps).
7320 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7321 support for redirecting crossing jumps to non-crossing.
7323 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7325 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7326 also for naked functions.
7328 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7330 * config/arc/arc.md (add_shift): New pattern.
7331 (add_shift2): Likewise.
7332 (sub_shift): Likewise.
7333 (sub_shift_cmp0_noout): Likewise.
7334 (compare_si_ashiftsi): Likewise.
7335 (xbfu_cmp0_noout): New combine pattern.
7336 (xbfu_cmp0"): Likewise.
7337 (movsi_set_cc_insn): Place the predicable variant first.
7338 (commutative_binary_cmp0_noout): Remove clobber.
7339 (commutative_binary_cmp0): New pattern.
7340 (noncommutative_binary_cmp0): Likewise.
7341 (noncommutative_binary_cmp0_noout): Likewise.
7342 (noncommutative_binary_comparison_result_used): Removed.
7343 (rsub_cmp0): New pattern.
7344 (rsub_cmp0_noout): Likewise.
7345 (extzvsi): Changed, keep only meaningful variants.
7346 (SQH, SEZ): New iterators.
7347 (SQH_postfix): New mode attribute.
7348 (SEZ_prefix): New code attribute.
7349 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7350 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7351 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7353 (noncommutative_operator): Check the availability of barrel
7356 2018-04-09 Richard Biener <rguenther@suse.de>
7358 PR tree-optimization/85284
7359 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7360 Only use the niter constraining form of simple_iv when the exit
7363 2018-04-09 Tom de Vries <tom@codesourcery.com>
7366 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7367 (define_expand "*memory_barrier"): New define_expand.
7368 (define_insn "memory_barrier"): New insn.
7370 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7372 PR rtl-optimization/80463
7373 PR rtl-optimization/83972
7374 PR rtl-optimization/83480
7376 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7377 correct producer for the insn.
7378 (tidy_control_flow): Fixup seqnos in case of debug insns.
7380 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7382 PR rtl-optimization/83913
7384 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7385 different sched-times when merging exprs.
7387 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7389 PR rtl-optimization/83962
7391 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7392 tidy_fallthru_edge and tidy_control_flow.
7394 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7396 PR rtl-optimization/83530
7398 * sel-sched.c (force_next_insn): New global variable.
7399 (remove_insn_for_debug): When force_next_insn is true, also leave only
7400 next insn in the ready list.
7401 (sel_sched_region): When the region wasn't scheduled, make another pass
7402 over it with force_next_insn set to 1.
7404 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7406 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7408 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7409 for interrupt control.
7410 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7411 functions for interrupt control.
7412 * config/nds32/nds32-intrinsic.md: Likewise.
7413 * config/nds32/nds32_intrinsic.h: Likewise.
7414 * config/nds32/nds32.h (nds32_builtins): Likewise.
7416 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7418 * config/nds32/nds32.c (nds32_init_machine_status,
7419 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7420 strict_aligned_p field.
7421 (nds32_expand_to_rtl_hook): New function.
7422 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7423 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7425 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7426 Chung-Ju Wu <jasonwucj@gmail.com>
7428 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7429 * config/nds32/nds32-n7.md: New file.
7430 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7431 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7433 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7434 * config/nds32/nds32.md (pipeline_model): Add n7.
7435 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7436 * config/nds32/pipelines.md: Include n7 settings.
7438 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7439 Chung-Ju Wu <jasonwucj@gmail.com>
7441 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7442 * config/nds32/nds32-e8.md: New file.
7443 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7444 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7446 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7447 * config/nds32/nds32.md (pipeline_model): Add e8.
7448 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7449 * config/nds32/pipelines.md: Include e8 settings.
7451 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7452 Chung-Ju Wu <jasonwucj@gmail.com>
7454 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7455 * config/nds32/nds32-n8.md: New file.
7456 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7457 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7459 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7460 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7461 * config/nds32/nds32.md (pipeline_model): Add n8.
7462 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7463 * config/nds32/pipelines.md: Include n8 settings.
7465 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7466 Chung-Ju Wu <jasonwucj@gmail.com>
7468 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7469 * config/nds32/nds32-n9-2r1w.md: New file.
7470 * config/nds32/nds32-n9-3r2w.md: New file.
7471 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7472 nds32_register_ports): New or modify for cpu n9.
7473 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7475 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7476 * config/nds32/nds32-utils.c: New file.
7477 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7478 TARGET_MUL_SLOW): Define.
7479 * config/nds32/nds32.md (pipeline_model): New attribute.
7480 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7481 New options that support cpu n9.
7482 * config/nds32/pipelines.md: Include n9 settings.
7483 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7485 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7487 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7488 information if necessary.
7489 (output_cond_branch_compare_zero): Likewise.
7490 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7491 (nds32_target_alignment): Refine for alignment.
7492 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7493 (FUNCTION_BOUNDARY): Modify.
7494 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7496 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7498 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7500 * config/nds32/constants.md (unspec_volatile_element): Add values for
7501 TLB operation and data prefetch.
7502 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7503 functions for TLB operation and data prefetch.
7504 * config/nds32/nds32-intrinsic.md: Likewise.
7505 * config/nds32/nds32_intrinsic.h: Likewise.
7506 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7507 (nds32_print_operand): Likewise.
7508 * config/nds32/nds32.h (nds32_builtins): Likewise.
7510 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7511 Andrew Pinski <pinsika@gcc.gnu.org>
7514 * match.pd: Use constant_boolean_node of correct type instead of
7515 boolean_true_node or boolean_false_node for simplifying
7516 pointer comparisons to zero.
7518 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7520 PR tree-optimization/80021
7521 * tree.c (verify_type_variant): Make error call in verify_variant_match
7522 translatable and remove final full stop.
7524 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7526 * config/nds32/constants.md (unspec_volatile_element): Add
7527 UNSPEC_VOLATILE_EH_RETURN.
7528 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7529 nds32_output_stack_pop): Support dwarf exception handling process.
7530 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7531 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7532 exception handling process.
7533 (nds32_compute_stack_frame): Likewise.
7534 (nds32_return_addr_rtx): Likewise.
7535 (nds32_initial_elimination_offset): Likewise.
7536 (nds32_expand_prologue): Likewise.
7537 (nds32_expand_epilogue): Likewise.
7538 (nds32_dynamic_chain_address): New function.
7539 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7541 (DYNAMIC_CHAIN_ADDRESS): Define.
7542 (EH_RETURN_DATA_REGNO): Define.
7543 (EH_RETURN_STACKADJ_RTX): Define.
7544 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7545 patterns for dwarf exception handling.
7547 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7549 * config/nds32/nds32.h: Clean up obsolete macros.
7551 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7553 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7554 Add enum values for particular instructions.
7555 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7556 particular intrinsic functions.
7557 * config/nds32/nds32-intrinsic.md: Likewise.
7558 * config/nds32/nds32_intrinsic.h: Likewise.
7559 * config/nds32/nds32.h (nds32_builtins): Likewise.
7560 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7561 (btst, ave): New patterns for particular instructions.
7563 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7565 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7566 Add enum values for atomic load/store and memory sync.
7567 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7569 * config/nds32/nds32-intrinsic.md: Likewise.
7570 * config/nds32/nds32_intrinsic.h: Likewise.
7571 * config/nds32/nds32.h (nds32_builtins): Likewise.
7573 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7575 PR tree-optimization/85257
7576 * fold-const.c (native_encode_vector): If not all elts could fit
7577 and off is -1, return 0 rather than offset.
7578 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7579 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7580 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7581 adjust buffer in native_interpret_expr call.
7583 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7585 * config/nds32/constants.md (unspec_volatile_element): Add cache
7586 control enum values.
7587 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7588 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7589 * config/nds32/nds32.c (nds32_cctl_names): New.
7590 (nds32_print_operand): Handle cache control register names.
7591 * config/nds32/nds32.h (nds32_builtins): New enum values.
7592 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7594 * config/nds32/nds32.md (type): Add mmu.
7595 * config/nds32/pipelines.md (simple_insn): Add mmu.
7597 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7599 * config/nds32/nds32.md (type): Remove call.
7600 * config/nds32/pipelines.md (simple_insn): Likewise.
7602 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7604 * config/nds32/constants.md (unspec_volatile_element): Add
7605 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7606 UNSPEC_VOLATILE_FMFCFG.
7607 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7608 description for fmfcfg and fmfcsr.
7609 (bdesc_1arg): Add fmtcsr.
7610 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7611 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7612 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7613 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7614 unspec_fmfcfg): New patterns.
7615 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7616 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7617 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7618 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7619 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7620 __nds32__fmfcfg): Define.
7622 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7624 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7625 intrinsic register names.
7626 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7627 intrinsic register enum values and macros.
7629 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7631 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7632 for load/store addressing form.
7633 (nds32_print_operand_address): Likewise.
7635 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7638 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7639 based on LABEL_REF. Remove useless assertion.
7640 (pic_address_needs_scratch): Fix formatting.
7641 (sparc_legitimize_pic_address): Minor tweaks.
7642 (sparc_delegitimize_address): Adjust assertion accordingly.
7643 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7644 into symbolic_operand.
7645 (movsi_high_pic_label_ref): Likewise.
7646 (movsi_lo_sum_pic_label_ref): Likewise.
7647 (movdi_pic_label_ref): Likewise.
7648 (movdi_high_pic_label_ref): Likewise.
7649 (movdi_lo_sum_pic_label_ref): Likewise.
7651 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7653 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7654 custom LIB_SPEC setup.
7656 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7657 Kito Cheng <kito.cheng@gmail.com>
7659 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7660 * config/riscv/freebsd.h: New.
7662 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7664 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7665 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7668 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7669 Kito Cheng <kito.cheng@gmail.com>
7671 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7672 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7673 * config/nds32/nds32-protos.h (nds32_output_call,
7674 nds32_output_return): Declare.
7675 * config/nds32/nds32.md: Refine all the call and return patterns.
7677 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7680 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7681 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7683 PR rtl-optimization/84872
7684 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7685 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7688 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7690 * expr.c (copy_blkmode_to_reg): Revert 254862.
7691 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7693 2018-04-06 Richard Biener <rguenther@suse.de>
7696 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7697 after seeing a component reference with an adjacent field. Treat
7698 refs to arrays at struct end of external decls similar to
7699 refs to unconstrained commons.
7701 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7704 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7705 look through SAVE_EXPRs with non-side-effects argument. Adjust
7707 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7710 2018-04-06 Richard Biener <rguenther@suse.de>
7713 * alias.c (find_base_term): New wrapper around find_base_term
7714 unwinding CSELIB_VAL_PTR changes.
7715 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7718 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7720 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7722 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7723 constant definitions.
7724 ("nop"): lr 0,0 -> nopr r0
7725 ("nop_lr0", "nop_lr1"): New insn definitions.
7727 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7729 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7730 NDS32_V3PUSH_AVAILABLE_P macro.
7732 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7733 Chung-Ju Wu <jasonwucj@gmail.com>
7735 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7736 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7737 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7738 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7739 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7740 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7741 * config/nds32/constraints.md: New constraints and checking for hard
7742 float configuration.
7743 * config/nds32/iterators.md: New mode iterator and attribute for hard
7744 float configuration.
7745 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7747 * config/nds32/nds32-fpu.md: New file.
7748 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7749 deal with hard float code generation.
7750 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7752 (abi_type, float_reg_number): New enum type.
7753 * config/nds32/nds32-predicates.c: New predicates for hard float.
7754 * config/nds32/nds32-protos.h: Declare functions for hard float.
7755 * config/nds32/nds32.c: Implementation for hard float configuration.
7756 * config/nds32/nds32.h: Definitions for hard float configuration.
7757 * config/nds32/nds32.md: Include hard float machine description and
7758 modify patterns for hard float configuration.
7759 * config/nds32/nds32.opt: New options for hard float configuration.
7760 * config/nds32/predicates.md: New predicates for hard float
7763 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7765 * common/config/nds32/nds32-common.c
7766 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7768 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7771 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7772 CONSTRUCTOR_ELT (ctor, ...)->value.
7774 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7777 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7779 2018-04-05 Tom de Vries <tom@codesourcery.com>
7782 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7785 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7786 Kito Cheng <kito.cheng@gmail.com>
7788 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7789 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7790 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7791 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7793 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7794 Kito Cheng <kito.cheng@gmail.com>
7796 * config/nds32/constraints.md (Ufe): New memory constraint.
7797 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7798 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7799 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7801 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7802 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7804 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7806 * config/nds32/nds32.md: Use optimize_size in the condition for
7807 alu-shift instructions.
7809 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7811 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7813 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7815 * config/nds32/nds32.md (negsi2): Refine pattern.
7817 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7818 Chung-Ju Wu <jasonwucj@gmail.com>
7820 * config/nds32/iterators.md (shift_rotate): New code iterator.
7821 (shift): New code attribute.
7822 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7823 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7824 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7825 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7826 bit-wise operations.
7827 (andsi3, *andsi3): Ditto.
7828 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7829 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7830 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7831 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7832 nds32_ior_operand, nds32_xor_operand): New predicates.
7834 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7836 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7837 (addsi3, subsi3): ... this.
7839 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7841 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7843 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7845 * config/nds32/nds32.md: Adjust indention.
7847 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7849 * config/nds32/nds32.md (feature): New attribute.
7851 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7853 * config/nds32/nds32.md (subtype): New attribute.
7855 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7858 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7859 expansion to perform a bitwise AND of the argument followed by a
7860 boolean negation of the result.
7862 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7864 PR rtl-optimization/84878
7865 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7866 the basic block. Assert the use reference is not artificial and that
7867 it has an associated insn.
7869 2018-04-04 Michael Matz <matz@suse.de>
7871 * builtins.c (compute_objsize): Pass correct operand
7872 to array_at_struct_end_p.
7874 2018-04-04 Richard Biener <rguenther@suse.de>
7877 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7878 from contexts for DINFO_LEVEL_TERSE and below.
7880 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7882 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7883 resiter_operand condition.
7884 * config/nds32/nds32.md (*move<mode>): Ditto.
7886 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7887 Monk Chiang <sh.chiang04@gmail.com>
7889 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7891 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7893 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7895 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7896 Kito Cheng <kito.cheng@gmail.com>
7898 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7899 nds32_cond_code_str, output_cond_branch,
7900 output_cond_branch_compare_zero, nds32_expand_cbranch,
7901 nds32_expand_cstore, nds32_expand_movcc,
7902 nds32_output_cbranchsi4_equality_zero,
7903 nds32_output_cbranchsi4_equality_reg,
7904 nds32_output_cbranchsi4_equality_reg_or_const_int,
7905 nds32_output_cbranchsi4_greater_less_zero: New functions.
7906 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7907 nds32_expand_cstore, nds32_expand_movcc,
7908 nds32_output_cbranchsi4_equality_zero,
7909 nds32_output_cbranchsi4_equality_reg,
7910 nds32_output_cbranchsi4_equality_reg_or_const_int,
7911 nds32_output_cbranchsi4_greater_less_zero): Declare.
7912 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7913 nds32_rimm11s_operand): New predicates.
7914 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7915 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7918 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7920 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7921 * config/nds32/nds32.md: Ditto.
7922 * config/nds32/pipelines.md: Ditto.
7924 2018-04-04 Richard Biener <rguenther@suse.de>
7926 PR tree-optimization/85168
7927 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7928 propagating abnormals.
7930 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7932 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7934 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7935 Kito Cheng <kito.cheng@gmail.com>
7937 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7938 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7939 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7940 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7941 * config/nds32/nds32.md (sibcall_internal): New.
7942 (sibcall_register): Remove.
7943 (sibcall_immediate): Remove.
7944 (sibcall_value_internal): New.
7945 (sibcall_value_register): Remove.
7946 (sibcall_value_immediate): Remove.
7947 * config/nds32/predicates.md (nds32_general_register_operand): New.
7948 (nds32_call_address_operand): New.
7950 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7952 PR rtl-optimization/85167
7953 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7954 bb_defs if *split_p, instead preinitialize it to NULL.
7956 PR tree-optimization/85156
7957 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7958 evaluating the argument multiple times.
7960 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7962 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7964 (_mm_cvtpd_ps): Likewise.
7965 (_mm_cvttpd_epi32): Likewise.
7966 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7967 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7968 vector, pixel, and bool following altivec.h include.
7970 2018-04-03 Martin Sebor <msebor@redhat.com>
7972 * doc/extend.texi (Common Function Attributes): Clarify.
7973 (const attribute): Likewise.
7974 (pure attribute): Likewise.
7976 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7979 * config/i386/i386.c (ix86_expand_vector_set): Use
7980 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7982 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7984 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7985 instructions when changing rounding bits to preserve precision bits
7986 in the x87 control word.
7988 2018-04-03 Martin Liska <mliska@suse.cz>
7990 PR tree-optimization/82491
7991 * rtl.h (strip_offset_and_add): Replace += suboffset with
7992 poly_uint64 () + suboffset.
7994 2018-03-29 Martin Liska <mliska@suse.cz>
7995 Martin Jambor <mjambor@suse.cz>
7998 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7999 param_type is not an integral or pointer type.
8001 2018-04-03 Richard Biener <rguenther@suse.de>
8003 * sese.h (recompute_all_dominators): Remove.
8005 2018-04-02 Martin Sebor <msebor@redhat.com>
8007 * doc/invoke.texi (-Wrestrict): Fix typos.
8009 2018-04-02 Jim Wilson <jimw@sifive.com>
8011 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
8012 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
8013 (<optab>di3, <optab>si3_extend): Likewise.
8014 (<optab>si3_mask, <optab>si3_mask_1): New.
8015 (<optab>di3_mask, <optab>di3_mask_1): New.
8016 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
8017 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
8018 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
8020 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
8022 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
8025 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
8027 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
8028 (nds32_canonicalize_comparison): New function.
8030 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8031 Kito Cheng <kito.cheng@gmail.com>
8032 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8034 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
8035 * config/nds32/constants.md (unspec_volatile_element): Add
8036 UNSPEC_VOLATILE_RELAX_GROUP.
8037 * config/nds32/nds32-relax-opt.c: New file.
8038 * config/nds32/nds32-predicates.c
8039 (nds32_symbol_load_store_p): New function.
8040 * config/nds32/nds32-protos.h
8041 (nds32_symbol_load_store_p): Declare function.
8042 (make_pass_nds32_relax_opt): Declare new rtl pass function.
8043 * config/nds32/nds32.c
8044 (nds32_register_pass): New function to register pass.
8045 (nds32_register_passes): New function to register passes.
8046 * config/nds32/nds32.md (relax_group): New pattern.
8047 * config/nds32/nds32.opt (mrelax-hint): New option.
8048 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
8050 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
8052 * config/nds32/t-nds32: Modify files dependency.
8054 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8056 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
8057 (PROFILE_HOOK): Define its implementation.
8059 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
8061 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
8062 type and 32-bit size.
8064 2018-04-01 Jakub Jelinek <jakub@redhat.com>
8067 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
8068 (V_128_256): New mode iterator.
8069 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
8070 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
8071 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
8073 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
8074 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
8076 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
8079 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
8080 NaN inputs correctly.
8082 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
8085 * config/rs6000/vsx.md (??r): New mode attribute.
8086 (*vsx_mov<mode>_64bit): Use it.
8087 (*vsx_mov<mode>_32bit): Likewise.
8089 2018-03-30 Martin Sebor <msebor@redhat.com>
8091 PR tree-optimization/84818
8092 * builtins.c (check_access): Use warning_n.
8094 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8097 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
8099 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8102 2018-03-30 Julia Koval <julia.koval@intel.com>
8105 * x86-tune.def (movx, partial_reg_dependency): Enable for
8108 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8111 * lra-constraints.c (process_alt_operands): Move setting
8112 this_alternative_matches below.
8114 2018-03-29 Martin Liska <mliska@suse.cz>
8117 * doc/invoke.texi: Document how LTO works with debug info.
8118 Describe auto-load support of binutils. Mention 'x86-64'
8119 as valid option value of -march option.
8121 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8123 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8126 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8127 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8128 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8131 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8134 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8135 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8136 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8137 for RS6000_BTM_POWERPC64.
8138 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8139 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8140 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8145 2018-03-28 Carl Love <cel@us.ibm.com>
8148 2017-09-27 Carl Love <cel@us.ibm.com>
8150 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8151 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8152 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8155 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8157 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8158 instead of __vector bool.
8159 (_mm_max_pu8): Likewise.
8160 (_mm_min_pi16): Likewise.
8162 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8165 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8166 (DIVWEUO): Likewise.
8168 (DIVDEUO): Likewise.
8169 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8170 DIVWEUO and DIVDEUO.
8171 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8172 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8173 (div_extend): Likewise.
8174 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8176 (__builtin_divweuo): Likewise.
8177 (__builtin_divdeo): Likewise.
8178 (__builtin_divdeuo): Likewise.
8180 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8183 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8184 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8186 PR tree-optimization/82004
8187 * gimple-match-head.c (optimize_pow_to_exp): New function.
8188 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8189 Don't fold to exp if optimize_pow_to_exp is false.
8191 2018-03-28 Martin Liska <mliska@suse.cz>
8194 * calls.c (initialize_argument_information): Fix trailing space.
8195 * common.opt: Fix typo and provide better explanation for
8196 -fsanitize-coverage option.
8197 * config/i386/i386.opt: Fix typo.
8199 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8200 Martin Liska <mliska@suse.cz>
8203 * gimplify.c (asan_poison_variable): Don't do the check for
8204 gimplify_omp_ctxp here.
8205 (gimplify_decl_expr): Do it here.
8206 (gimplify_target_expr): Likewise.
8208 2018-03-28 Martin Liska <mliska@suse.cz>
8211 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8212 chkp_type_bounds_count if MPX is not enabled.
8214 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8216 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8218 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8221 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8222 function to create the function decl for complex long double
8223 multiply and divide for -mabi=ieeelongdouble.
8224 (init_float128_ieee): Call it.
8226 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8229 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8230 -fcf-protection=branch -mibt.
8231 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8233 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8236 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8237 arm_word_relocations.
8239 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8242 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8243 extern array declarations.
8245 2018-03-27 Richard Biener <rguenther@suse.de>
8248 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8249 explicit single_use checks.
8251 2018-03-27 Richard Biener <rguenther@suse.de>
8253 PR tree-optimization/85082
8254 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8257 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8259 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8260 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8261 Turn on fasynchronous-unwind-tables and funwind-tables.
8263 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8266 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8267 (*bmi_blsr_<mode>_ccz): Ditto.
8269 2018-03-26 Tom de Vries <tom@codesourcery.com>
8271 PR tree-optimization/85063
8272 * omp-general.c (offloading_function_p): New function. Factor out
8274 * omp-offload.c (pass_omp_target_link::gate): ... here.
8275 * omp-general.h (offloading_function_p): Declare.
8276 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8277 with attribute omp declare target for offloading functions.
8279 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8281 PR tree-optimization/84005
8282 * tree-data-ref.h (get_base_for_alignment): Declare.
8283 * tree-data-ref.c (get_base_for_alignment_1): New function.
8284 (get_base_for_alignment): Likewise.
8285 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8286 get_base_for_alignment to find a suitable base object, instead
8287 of always using drb->base_address.
8289 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8292 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8293 known size by default.
8295 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8298 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8299 and non BLKmode operands.
8301 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8304 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8305 Clean up attributes.
8307 2018-03-23 Richard Biener <rguenther@suse.de>
8310 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8311 we are going to emit early debug for LTO.
8313 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8316 * function.c (match_asm_constraints_1): Don't optimize if input
8317 doesn't satisfy general_operand predicate for output's mode.
8320 * alias.c (write_dependence_p): Don't require for x_canonicalized
8321 non-VOIDmode if x has VOIDmode.
8324 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8325 just don't try to optimize it rather than assert it never happens.
8327 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8329 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8330 macro expansions for definition of ST_INTERNAL_<mode> and
8331 LD_INTERNAL_<mode> builtins.
8332 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8334 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8336 (altivec_expand_st_builtin): Likewise.
8337 (altivec_expand_builtin): Remove calls to deleted functions.
8338 (rs6000_address_for_altivec): Delete this function.
8339 * config/rs6000/vector.md: Remove expands for
8340 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8342 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8345 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8346 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8347 re-computing once computed.
8348 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8349 (arm_init_machine_status): Initialize
8350 machine->static_chain_stack_bytes.
8352 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8355 * doc/extend.texi: Add four new prototypes for vec_ld.
8356 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8357 definitions for more logical presentation.
8358 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8359 entries for V1TI variants of __builtin_altivec_ld builtin.
8360 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8361 handling of V1TI variant of LVX icode pattern.
8362 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8363 (rs6000_gimple_fold_builtin): Likewise.
8364 (altivec_init_builtins): Add code to define
8365 __builtin_altivec_lvx_v1ti function.
8367 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8370 * function.c (match_asm_constraints_1): Don't do the optimization
8371 if input isn't a REG, SUBREG, MEM or constant.
8373 2018-03-22 Tom de Vries <tom@codesourcery.com>
8375 PR tree-optimization/84956
8376 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8377 bb_has_abnormal_pred.
8379 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8382 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8383 DECL_INITIAL (decl) to decl at the end.
8384 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8387 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8389 * doc/extend.texi (__builtin_tgmath): Document when complex
8390 integer types are treated as _Complex _Float64.
8392 2018-03-21 Tom de Vries <tom@codesourcery.com>
8394 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8396 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8398 PR tree-optimization/84960
8399 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8400 if it is ENTRY block, move them into single succ of ENTRY in that case.
8402 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8404 PR tree-optimization/84811
8405 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8406 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8407 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8408 (known_subrange_p): Update accordingly. Cast each value involved
8409 in the size comparison, rather than casting the result of the
8412 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8414 PR tree-optimization/84982
8415 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8416 by flipping the least significant bit rather than all bits from
8417 bitpos to bitpos + bitsize - 1.
8419 2018-03-21 Nathan Sidwell <nathan@acm.org>
8421 * doc/extend.texi (Deprecated Features): Remove mention of
8422 long-deleted deprecations.
8424 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8427 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8428 * configure: Regenerate.
8430 2018-03-21 Tom de Vries <tom@codesourcery.com>
8432 PR tree-optimization/83126
8433 * tree-parloops.c (num_phis): New function.
8434 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8436 2018-03-21 Nathan Sidwell <nathan@acm.org>
8438 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8439 mention anon-struct/union members and trailing attributes.
8441 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8443 PR tree-optimization/84969
8444 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8445 builtin memset partitions if they set different rhs values.
8447 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8449 PR rtl-optimization/84989
8450 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8451 VEC_DUPLICATE with scalar result mode.
8453 2018-03-21 Martin Liska <mliska@suse.cz>
8456 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8457 not intended return statement.
8459 2018-03-21 Martin Liska <mliska@suse.cz>
8462 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8463 (chkp_find_bound_slots_1): Limit number of iterations.
8465 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8468 * Minor grammar fixes for x86 options.
8470 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8473 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8474 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8476 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8479 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8480 (altivec_lvx_<mode>_1op): Likewise.
8481 (altivec_stvx_<mode>_2op): Likewise.
8482 (altivec_stvx_<mode>_1op): Likewise.
8483 (altivec_lvx_<VM2:mode>): New define_expand.
8484 (altivec_stvx_<VM2:mode>): Likewise.
8485 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8486 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8487 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8488 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8489 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8490 (rs6000_gen_lvx): Likewise.
8491 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8492 (altivec_expand_stv_builtin): Likewise.
8493 (altivec_expand_builtin): Likewise.
8494 * config/rs6000/vector.md: Likewise.
8496 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8499 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8502 2018-03-20 Richard Biener <rguenther@suse.de>
8505 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8506 sign-conversions as zero, fall back to standard scalar_stmt
8509 2018-03-20 Martin Liska <mliska@suse.cz>
8512 * predict.c (rebuild_frequencies): Handle case when we have
8513 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8515 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8518 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8519 flag_section_anchors.
8520 * varasm.c (use_blocks_for_decl_p): Remove hack for
8521 dw2_force_const_mem.
8524 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8526 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8527 be created, use lowpart_subreg of operands[0] rather than operands[0]
8529 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8530 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8531 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8532 and n constraint instead of aarch64_shift_imm_di and Usd.
8533 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8534 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8536 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8539 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8540 to favor GPR over NEON registers.
8541 (<shift>di3_neon): Likewise.
8543 2018-03-20 Tom de Vries <tom@codesourcery.com>
8546 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8547 (nvptx_process_pars): Emit bar.sync asap and alap.
8549 2018-03-20 Tom de Vries <tom@codesourcery.com>
8552 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8553 seen_label if seen_label is already set.
8555 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8558 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8559 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8560 Use 1U instead of 1. Formatting fixes.
8563 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8564 instead of TREE_TYPE (s1) for the return value.
8566 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8568 PR tree-optimization/84946
8569 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8570 bitsize + bitsize in poly_uint64 rather than poly_int64.
8573 * dwarf2asm.c: Include fold-const.c.
8574 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8575 of decl rather than decl itself.
8577 PR rtl-optimization/84643
8578 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8580 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8583 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8584 calling assemble_variable.
8586 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8589 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8590 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8592 2018-03-19 Jim Wilson <jimw@sifive.com>
8595 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8596 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8597 (riscv_first_stack_step): Likewise.
8598 (riscv_option_override): Use STACK_BOUNDARY instead of
8600 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8602 (BIGGEST_ALIGNMENT): Set to 128.
8603 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8604 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8607 2018-03-19 Richard Biener <rguenther@suse.de>
8609 PR tree-optimization/84933
8610 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8611 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8613 2018-03-19 Richard Biener <rguenther@suse.de>
8615 PR tree-optimization/84859
8616 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8617 (cond_if_else_store_replacement): Perform sinking operation on
8618 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8619 Generalize what a BB with a single eligible store is.
8621 2018-03-19 Richard Biener <rguenther@suse.de>
8623 PR tree-optimization/84929
8624 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8625 chrec_is_positive against non-chrec arg.
8627 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8630 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8632 2018-03-18 Martin Liska <mliska@suse.cz>
8634 PR rtl-optimization/84635
8635 * regrename.c (build_def_use): Use matches_mode only when
8638 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8640 PR tree-optimization/84913
8641 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8642 vectorize chains of COND_EXPRs.
8644 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8646 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8648 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8650 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8652 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8654 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8656 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8657 Kito Cheng <kito.cheng@gmail.com>
8659 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8660 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8661 (nds32_adjust_reg_alloc_order): New function.
8662 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8664 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8666 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8667 nds32_print_operand, nds32_print_operand_address): Use
8668 HOST_WIDE_INT_PRINT_DEC instead.
8670 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8672 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8674 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8677 * config/i386/i386.c (initial_ix86_tune_features,
8678 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8680 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8681 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8682 rather than 1u << ix86_tune. Formatting fix.
8683 (ix86_option_override_internal): Change ix86_arch_mask from
8684 unsigned int to unsigned HOST_WIDE_INT, initialize to
8685 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8686 (ix86_function_specific_restore): Likewise.
8688 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8691 * postreload.c (reload_combine_recognize_pattern): Perform
8692 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8693 truncate_int_for_mode the result for the destination's mode.
8696 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8697 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8699 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8702 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8705 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8706 regno_allocno_class_array and sorted_pseudos.
8707 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8708 insns where regno is used.
8710 2018-03-16 Martin Liska <mliska@suse.cz>
8713 * multiple_target.c (create_dispatcher_calls): Redirect
8714 reference in the symbol table.
8716 2018-03-16 Martin Liska <mliska@suse.cz>
8719 * multiple_target.c (create_dispatcher_calls): Redirect also
8722 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8726 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8727 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8728 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8730 2018-03-16 Julia Koval <julia.koval@intel.com>
8732 * doc/invoke.texi (Skylake Server): Add CLWB.
8733 Cannonlake): Remove CLWB.
8735 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8737 PR tree-optimization/84841
8738 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8740 (FLOAT_ONE_CONST_TYPE): Define.
8741 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8742 (sort_by_operand_rank): Put entries with higher constant_type last
8743 rather than first to match comments.
8745 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8747 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8750 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8753 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8754 check and use address of target always.
8756 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8759 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8760 (indirect_thunk_bnd_needed): Likewise.
8761 (indirect_thunks_used): Likewise.
8762 (indirect_thunks_bnd_used): Likewise.
8763 (indirect_return_needed): New.
8764 (indirect_return_bnd_needed): Likewise.
8765 (output_indirect_thunk_function): Add a bool argument for
8767 (output_indirect_thunk_function): Don't generate alias for
8768 function return thunk.
8769 (ix86_code_end): Call output_indirect_thunk_function to generate
8770 function return thunks.
8771 (ix86_output_function_return): Set indirect_return_bnd_needed
8772 and indirect_return_needed instead of indirect_thunk_bnd_needed
8773 and indirect_thunk_needed.
8775 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8777 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8778 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8779 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8781 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8782 Paul Hua <paul.hua.gm@gmail.com>
8785 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8787 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8789 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8790 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8793 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8796 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8797 instead of GET_MODE_SIZE when comparing Units.
8799 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8802 * varasm.c (hash_section): Return an unchangeble hash value
8803 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8804 Return !aarch64_can_use_per_function_literal_pools_p ().
8806 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8809 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8810 rather than address of cmode as last argument to prepare_cmp_insn.
8812 2018-03-15 Julia Koval <julia.koval@intel.com>
8814 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8815 F_AVX512VNNI, F_AVX512BITALG): New.
8817 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8820 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8821 insn for floating-point loads and stores.
8823 2018-03-14 Carl Love <cel@us.ibm.com>
8825 * config/rs6000/rs6000-c.c: Add macro definitions for
8826 ALTIVEC_BUILTIN_VEC_PERMXOR.
8827 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8828 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8829 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8830 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8832 * config/doc/extend.texi: Add prototypes for vec_permxor.
8834 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8837 * diagnostic-show-locus.c (class layout_point): Convert m_line
8838 from int to linenum_type.
8839 (line_span::comparator): Use linenum "compare" function when
8840 comparing line numbers.
8841 (test_line_span): New function.
8842 (layout_range::contains_point): Convert param "row" from int to
8844 (layout_range::intersects_line_p): Likewise.
8845 (layout::will_show_line_p): Likewise.
8846 (layout::print_source_line): Likewise.
8847 (layout::should_print_annotation_line_p): Likewise.
8848 (layout::print_annotation_line): Likewise.
8849 (layout::print_leading_fixits): Likewise.
8850 (layout::annotation_line_showed_range_p): Likewise.
8851 (struct line_corrections): Likewise for field m_row.
8852 (line_corrections::line_corrections): Likewise for param "row".
8853 (layout::print_trailing_fixits): Likewise.
8854 (layout::get_state_at_point): Likewise.
8855 (layout::get_x_bound_for_row): Likewise.
8856 (layout::print_line): Likewise.
8857 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8858 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8859 * input.c (selftest::test_linenum_comparisons): New function.
8860 (selftest::input_c_tests): Call it.
8861 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8862 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8863 * selftest.h (ASSERT_GT): New macro.
8864 (ASSERT_GT_AT): New macro.
8865 (ASSERT_LT): New macro.
8866 (ASSERT_LT_AT): New macro.
8868 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8870 PR rtl-optimization/84780
8871 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8873 2018-03-14 Martin Liska <mliska@suse.cz>
8875 * tree.c (record_node_allocation_statistics): Use
8876 get_stats_node_kind.
8877 (get_stats_node_kind): New function extracted from
8878 record_node_allocation_statistics.
8879 (free_node): Use get_stats_node_kind.
8881 2018-03-14 Richard Biener <rguenther@suse.de>
8883 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8884 that the value-set of ANTIC_IN doesn't grow.
8887 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8889 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8890 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8892 2018-03-14 Julia Koval <julia.koval@intel.com>
8894 * config.gcc (icelake-client, icelake-server): New.
8896 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8897 (initial_ix86_arch_features): Ditto.
8898 (PTA_SKYLAKE): Add SGX.
8899 (PTA_ICELAKE): Remove.
8900 (PTA_ICELAKE_CLIENT): New.
8901 (PTA_ICELAKE_SERVER): New.
8902 (ix86_option_override_internal): Split up icelake on icelake client and
8904 (get_builtin_code_for_version): Ditto.
8905 (fold_builtin_cpu): Ditto.
8906 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8907 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8908 * config/i386/i386.h (processor_type): Ditto.
8909 * doc/invoke.texi: Ditto.
8911 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8914 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8915 INTEGER_CST offset, add it together with bitpos / 8 and
8916 sign extend based on POINTER_SIZE.
8920 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8923 * config/i386/constraints.md (Yc): New register constraint.
8924 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8925 Use Yc constraint for alternative 2 of operand 0. Remove
8926 preferred_for_speed attribute.
8928 2018-03-14 Richard Biener <rguenther@suse.de>
8930 PR tree-optimization/84830
8931 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8932 with the old one to avoid oscillations.
8934 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8937 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8939 (assign_by_spills): Return a flag of reload assignment failure.
8940 Do not process the reload assignment failures. Do not spill other
8941 reload pseudos if they has the same reg class. Update n if
8943 (lra_assign): Add a return arg. Set up from the result of
8944 assign_by_spills call.
8945 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8946 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8947 usage_insns if it is not NULL.
8948 (spill_hard_reg_in_range): New function.
8949 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8950 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8951 function prototypes.
8952 (lra_assign): Change prototype.
8953 * lra.c (lra): Add code to deal with fails by splitting hard reg
8956 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8958 * config/riscv/riscv.opt (mrelax): New option.
8959 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8960 "norelax" when riscv_mrelax is disabled.
8961 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8963 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8966 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8967 reassociation for int modes.
8969 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8971 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8972 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8974 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8975 * config/aarch64/aarch64-sve.md
8976 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8977 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8978 (*extend<mode><Vwide>2): Rename to...
8979 (aarch64_sve_extend<mode><Vwide>2): ...this.
8980 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8981 renaming the old pattern to...
8982 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8984 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8985 define_expand, renaming the old pattern to...
8986 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8987 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8988 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8989 account when deciding which SVE instruction the optab should use.
8990 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8992 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8994 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8995 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8996 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8997 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8998 (tlsdesc_small_<mode>): Turn a define_expand and use
8999 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
9000 (tlsdesc_small_advsimd_<mode>): ...this.
9001 (tlsdesc_small_sve_<mode>): New pattern.
9003 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
9005 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
9006 (UNSPEC_UMUL_HIGHPART): New constants.
9007 (MUL_HIGHPART): New int iteraor.
9008 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
9009 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
9011 (*<su>mul<mode>3_highpart): New define_insn.
9013 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9016 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9019 2018-03-13 Martin Liska <mliska@suse.cz>
9022 * (sem_item_optimizer::sem_item_optimizer): Initialize new
9024 (sem_item_optimizer::~sem_item_optimizer): Release it.
9025 (sem_item_optimizer::merge_classes): Register variable aliases.
9026 (sem_item_optimizer::fixup_pt_set): New function.
9027 (sem_item_optimizer::fixup_points_to_sets): Likewise.
9028 * ipa-icf.h: Declare new variables and functions.
9030 2018-03-13 Jakub Jelinek <jakub@redhat.com>
9033 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
9034 integer_pow2p@2 and test integer_pow2p in condition.
9035 (A < 0 ? C : 0): Similarly for @1.
9038 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
9039 characters starting at p contain '\0' character, don't look beyond
9043 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
9044 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
9047 * reg-stack.c (change_stack): Change update_end var from int to
9048 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
9049 also call set_block_for_insn on the newly added insns and rescan.
9052 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
9053 on the last operand.
9056 * tree.c (stabilize_reference_1): Return save_expr (e) for
9057 STATEMENT_LIST even if it doesn't have side-effects.
9059 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
9061 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
9063 2018-03-12 Renlin Li <renlin.li@arm.com>
9065 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
9066 aarch64_output_scalar_simd_mov_immediate.
9068 2018-03-12 Martin Sebor <msebor@redhat.com>
9070 PR tree-optimization/83456
9071 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
9072 for perfectly overlapping calls to memcpy.
9073 (gimple_fold_builtin_memory_chk): Same.
9074 (gimple_fold_builtin_strcpy): Handle no-warning.
9075 (gimple_fold_builtin_stxcpy_chk): Same.
9076 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
9078 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9080 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
9081 parameter. Use it for SFmode.
9082 (rs6000_function_arg_advance_1): Adjust.
9083 (rs6000_function_arg): Adjust.
9084 (rs6000_gimplify_va_arg): Pass false for that new parameter.
9086 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
9088 PR rtl-optimization/84169
9089 PR rtl-optimization/84780
9090 * combine.c (can_combine_p): Check for a 2-insn combination whether
9091 the destination register is used between the two insns, too.
9093 2018-03-12 Richard Biener <rguenther@suse.de>
9095 PR tree-optimization/84803
9096 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
9097 for refs DR analysis didn't process.
9099 2018-03-12 Richard Biener <rguenther@suse.de>
9101 PR tree-optimization/84777
9102 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9103 force-vectorize loops ignore whether we are optimizing for size.
9105 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9107 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9108 (TARGET_MD_ASM_ADJUST): Define.
9110 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9111 Kito Cheng <kito.cheng@gmail.com>
9112 Chung-Ju Wu <jasonwucj@gmail.com>
9114 * config/nds32/nds32.c (nds32_compute_stack_frame,
9115 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9116 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9117 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9118 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9119 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9120 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9121 * config/nds32/nds32.md (prologue, epilogue): Use macro
9122 NDS32_V3PUSH_AVAILABLE_P to do checking.
9124 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9127 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9128 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9129 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9130 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9131 addition of most attributes on !orig_type_die or the attribute not
9132 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9134 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9135 Chung-Ju Wu <jasonwucj@gmail.com>
9137 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9139 * config/nds32/nds32.opt (mvh): New option.
9141 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9142 Chung-Ju Wu <jasonwucj@gmail.com>
9144 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9146 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9147 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9150 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9151 Chung-Ju Wu <jasonwucj@gmail.com>
9153 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9155 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9156 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9158 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9159 Kito Cheng <kito.cheng@gmail.com>
9160 Chung-Ju Wu <jasonwucj@gmail.com>
9162 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9163 UNSPEC_FFMISM and UNSPEC_FLMISM.
9164 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9165 for ffb, ffmism and flmism.
9166 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9167 (unspec_ffmism): Ditto.
9168 (unspec_flmism): Ditto.
9169 (nds32_expand_builtin_impl): Check if string extension is available.
9170 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9171 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9173 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9176 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9179 * lra-assigns.c (assign_by_spills): Return a flag of reload
9180 assignment failure. Do not process the reload assignment
9181 failures. Do not spill other reload pseudos if they has the same
9183 (lra_assign): Add a return arg. Set up from the result of
9184 assign_by_spills call.
9185 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9186 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9187 usage_insns if it is not NULL.
9188 (spill_hard_reg_in_range): New function.
9189 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9190 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9191 function prototypes.
9192 (lra_assign): Change prototype.
9193 * lra.c (lra): Add code to deal with fails by splitting hard reg
9196 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9199 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9201 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9204 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9205 (dw_val_node): Add val_symbolic_view.
9206 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9207 (symview_upper_bound): New.
9208 (new_line_info_table): Initialize symviews_since_reset.
9209 (dwarf2out_source_line): Count symviews_since_reset and set
9210 symview_upper_bound.
9211 (dw_val_equal_p): Handle symview.
9212 (add_AT_symview): New.
9213 (print_dw_val): Handle symview.
9214 (attr_checksum, attr_checksum_ordered): Likewise.
9215 (same_dw_val_p, size_of_die): Likewise.
9216 (value_format, output_die): Likewise.
9217 (add_high_low_attributes): Use add_AT_symview for entry_view.
9218 (dwarf2out_finish): Reset symview_upper_bound, clear
9221 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9224 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9225 Add strict argument and use it.
9226 (rs6000_split_multireg_move): Update for new strict argument.
9227 (mem_operand_gpr): Disallow all non-offsettable addresses.
9228 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9230 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9233 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9234 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9235 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9238 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9239 decl, use remap_type if we want to use the type.
9241 2018-03-09 Martin Sebor <msebor@redhat.com>
9243 PR tree-optimization/84526
9244 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9246 (builtin_access::generic_overlap): Be prepared to handle non-array
9249 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9251 PR rtl-optimization/84682
9252 * lra-constraints.c (process_address_1): Check is_address flag
9253 for address constraints.
9254 (process_alt_operands): Likewise.
9255 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9256 preprocess_constraints.
9257 * recog.h (preprocess_constraints): Add oploc parameter.
9259 * recog.c (preprocess_constraints): Test address_operand for
9260 CT_ADDRESS constraints.
9262 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9265 * lra-assigns.c (assign_by_spills): Return a flag of reload
9266 assignment failure. Do not process the reload assignment
9267 failures. Do not spill other reload pseudos if they has the same
9269 (lra_assign): Add a return arg. Set up from the result of
9270 assign_by_spills call.
9271 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9272 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9273 usage_insns if it is not NULL.
9274 (spill_hard_reg_in_range): New function.
9275 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9276 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9277 function prototypes.
9278 (lra_assign): Change prototype.
9279 * lra.c (lra): Add code to deal with fails by splitting hard reg
9282 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9285 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9286 Accept complain bool parameter. Only emit errors if it is true.
9287 (arm_parse_cpu_option_name): Likewise.
9288 (arm_target_thumb_only): Adjust callers of the above.
9289 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9290 prototype to take a default true bool parameter.
9291 (arm_parse_arch_option_name): Likewise.
9293 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9294 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9298 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9299 * configure: Regenerate.
9300 * configure.ac ("linker --version-script option"): New.
9301 ("linker soname option"): New.
9303 2018-03-09 Richard Biener <rguenther@suse.de>
9305 PR tree-optimization/84775
9306 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9307 immediate uses of predicate stmts and mark them modified.
9310 PR tree-optimization/84178
9311 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9313 (version_loop_for_if_conversion): Delay update_ssa call.
9314 (tree_if_conversion): Delay update_ssa until after predicate
9317 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9320 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9321 when the function accesses prior frames.
9323 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9326 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9327 gen_llsym, otherwise call maybe_gen_llsym.
9330 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9331 has ',' character inside of it.
9333 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9336 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9337 as clobbering CC_REGNUM.
9339 2018-03-08 Richard Biener <rguenther@suse.de>
9342 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9343 (follow_copies_to_constant): Do not follow SSA names registered
9346 2018-03-08 Richard Biener <rguenther@suse.de>
9348 PR tree-optimization/84178
9349 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9351 (version_loop_for_if_conversion): Delay update_ssa call.
9352 (tree_if_conversion): Delay update_ssa until after predicate
9355 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9357 PR tree-optimization/84178
9358 * tree-if-conv.c (release_bb_predicate): Remove the
9359 the assertion that the stmts have NULL use_ops.
9360 Discard the statements, asserting that they haven't
9361 yet been added to a BB.
9363 2018-03-08 Richard Biener <rguenther@suse.de>
9365 PR tree-optimization/84746
9366 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9367 (phi_translate): Pass in destination ANTIC_OUT set.
9368 (phi_translate_1): Likewise. For a simplified result lookup
9369 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9370 (phi_translate_set): Adjust.
9371 (do_pre_regular_insertion): Likewise.
9372 (do_pre_partial_partial_insertion): Likewise.
9374 2018-03-08 Martin Liska <mliska@suse.cz>
9376 PR gcov-profile/84735
9377 * doc/gcov.texi: Document usage of profile files.
9378 * gcov-io.h: Document changes in the format.
9380 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9384 * dwarf2out.c (struct dw_line_info_table): Update comments for
9386 (FORCE_RESET_NEXT_VIEW): New.
9387 (FORCE_RESETTING_VIEW_P): New.
9388 (RESETTING_VIEW_P): Check for -1 too.
9389 (ZERO_VIEW_P): Likewise.
9390 (new_line_info_table): Force-reset next view.
9391 (dwarf2out_begin_function): Likewise.
9392 (dwarf2out_source_line): Simplify zero_view_p initialization.
9393 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9394 view directly. Omit view when omitting .loc at line 0.
9396 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9398 PR tree-optimization/84740
9399 * tree-switch-conversion.c (process_switch): Call build_constructors
9400 only if info.phi_count is non-zero.
9402 PR tree-optimization/84739
9403 * tree-tailcall.c (find_tail_calls): Check call arguments against
9404 DECL_ARGUMENTS (current_function_decl) rather than
9405 DECL_ARGUMENTS (func) when checking for tail recursion.
9407 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9409 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9410 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9411 Volker Reichelt's entry and add entries for people that perform
9412 GCC fuzzy testing and report numerous bugs.
9414 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9417 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9418 readonly data in sdata, if that is disabled.
9419 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9420 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9421 -mreadonly-in-sdata option.
9423 2018-03-07 Martin Sebor <msebor@redhat.com>
9425 PR tree-optimization/84468
9426 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9427 basic block when looking for nul assignment.
9429 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9432 * except.h (output_function_exception_table): Adjust prototype.
9433 * except.c (output_function_exception_table): Remove FNNAME parameter
9434 and add SECTION parameter. Ouput one part of the table at a time.
9435 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9436 the first part of the exception table and emit unwind directives.
9437 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9438 (i386_pe_seh_cold_init): Likewise.
9439 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9440 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9441 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9442 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9443 * config/i386/winnt.c: Include except.h.
9444 (struct seh_frame_state): Add reg_offset, after_prologue and
9445 in_cold_section fields.
9446 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9447 (i386_pe_seh_cold_init): New function.
9448 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9449 to seh->in_cold_section.
9450 (seh_emit_push): Record the offset of the push.
9451 (seh_emit_save): Record the offet of the save.
9452 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9453 Test seh->after_prologue to disregard the epilogue.
9454 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9455 (i386_pe_end_cold_function): New function.
9457 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9460 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9461 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9464 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9466 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9467 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9470 * multiple_target.c: Include tree-inline.h and intl.h.
9471 (expand_target_clones): Diagnose and fail if node->definition and
9472 !tree_versionable_function_p (node->decl).
9474 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9476 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9478 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9479 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9480 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9482 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9485 * combine.c (try_combine): Use reg_or_subregno instead of handling
9486 just paradoxical SUBREGs and REGs.
9488 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9490 * config/arc/arc.c (arc_finalize_pic): Remove function.
9491 (arc_must_save_register): We use single base PIC register, remove
9492 checks to save/restore the PIC register.
9493 (arc_expand_prologue): Likewise.
9494 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9496 (arc_verify_short): Likewise.
9497 (arc_attr_type): Likewise.
9498 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9499 (walk_stores): Likewise.
9500 (arc_address_cost): Make it static.
9501 (arc_verify_short): Likewise.
9502 (branch_dest): Likewise.
9503 (arc_attr_type): Likewise.
9504 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9505 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9506 (arc_final_prescan_insn): Remove inserting the nops due to
9507 hardware hazards. It is done in reorg step.
9508 (insn_length_variant_t): Remove.
9509 (insn_length_parameters_t): Likewise.
9510 (arc_insn_length_parameters): Likewise.
9511 (arc_get_insn_variants): Likewise.
9512 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9514 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9517 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9520 PR tree-optimization/84687
9521 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9523 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9525 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9527 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9528 Rename to ppc_speculation_barrier.
9529 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9530 __builtin_ppc_speculation_barrier.
9532 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9535 * combine.c (combine_simplify_rtx): Don't try to simplify if
9536 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9539 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9541 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9542 to 32 bytes when compiling for POWER9.
9544 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9547 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9548 regparm >= 3 with no arg reg available also for calls with
9549 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9552 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9554 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9556 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9559 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9561 2018-03-05 Richard Biener <rguenther@suse.de>
9563 PR tree-optimization/84486
9564 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9565 When inserting a __builtin_assume_aligned call set the LHS
9566 SSA name alignment info accordingly.
9568 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9570 PR tree-optimization/84114
9571 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9572 Avoid reassociation of FLOAT_MODE addition.
9574 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9576 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9577 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9578 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9579 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9580 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9581 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9582 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9584 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9585 __builtin_ia32_wbinvd): New builtins.
9586 (SPECIAL_ARGS2): New.
9587 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9588 (SPECIAL_ARGS2): New.
9589 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9590 (ix86_valid_target_attribute_inner_p): Ditto.
9591 (ix86_init_mmx_sse_builtins): Add special_args2.
9592 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9593 TARGET_WBNOINVD_P): New.
9594 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9595 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9596 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9597 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9598 * config/i386/pconfigintrin.h: New file.
9599 * config/i386/wbnoinvdintrin.h: Ditto.
9600 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9602 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9604 2018-03-05 Richard Biener <rguenther@suse.de>
9606 PR tree-optimization/84670
9607 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9609 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9610 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9611 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9612 doesn't grow if all successors (recursively) were visited at least
9615 2018-03-05 Richard Biener <rguenther@suse.de>
9617 PR tree-optimization/84650
9618 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9619 if executed in the loop pipeline.
9621 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9623 * doc/configfiles.texi (Configuration Files): Move info about
9624 conditionalizing $target-protos.h to...
9625 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9626 differs from $target-protos.h.
9628 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9629 Chung-Ju Wu <jasonwucj@gmail.com>
9631 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9632 * config/nds32/nds32-multiple.md (setmemsi): Define.
9633 * config/nds32/nds32-memory-manipulation.c
9634 (nds32_gen_dup_4_byte_to_word_value): New.
9635 (emit_setmem_word_loop): New.
9636 (emit_setmem_byte_loop): New.
9637 (nds32_expand_setmem_loop): New.
9638 (nds32_expand_setmem_loop_v3m): New.
9639 (nds32_expand_setmem_unroll): New.
9640 (nds32_expand_setmem): New.
9642 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9643 Chung-Ju Wu <jasonwucj@gmail.com>
9645 * config/nds32/nds32-memory-manipulation.c
9646 (nds32_emit_load_store): New.
9647 (nds32_emit_post_inc_load_store): New.
9648 (nds32_emit_mem_move): New.
9649 (nds32_emit_mem_move_block): New.
9650 (nds32_expand_movmemsi_loop_unknown_size): New.
9651 (nds32_expand_movmemsi_loop_known_size): New.
9652 (nds32_expand_movmemsi_loop): New.
9653 (nds32_expand_movmemsi_unroll): New.
9654 (nds32_expand_movmemqi): Rename ...
9655 (nds32_expand_movmemsi): ... to this.
9656 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9657 (movmemsi): ... to this.
9658 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9659 (nds32_expand_movmemsi): ... to this.
9661 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9662 Monk Chiang <sh.chiang04@gmail.com>
9663 Chung-Ju Wu <jasonwucj@gmail.com>
9665 * config/nds32/nds32-protos.h
9666 (nds32_expand_load_multiple): New arguments.
9667 (nds32_expand_store_multiple): Ditto.
9668 (nds32_valid_multiple_load_store): Rename ...
9669 (nds32_valid_multiple_load_store_p): ... to this.
9670 * config/nds32/nds32-memory-manipulation.c
9671 (nds32_expand_load_multiple): Refine implementation.
9672 (nds32_expand_store_multiple): Ditto.
9673 * config/nds32/nds32-multiple.md
9674 (load_multiple): Update nds32_expand_load_multiple interface.
9675 (store_multiple): Update nds32_expand_store_multiple interface.
9676 * config/nds32/nds32-predicates.c
9677 (nds32_valid_multiple_load_store): Rename ...
9678 (nds32_valid_multiple_load_store_p): ... to this and refine
9680 * config/nds32/predicates.md
9681 (nds32_load_multiple_and_update_address_operation): New predicate.
9682 (nds32_store_multiple_and_update_address_operation): New predicate.
9684 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9685 Chung-Ju Wu <jasonwucj@gmail.com>
9687 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9688 (combo): New attribute.
9689 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9691 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9693 * config/nds32/nds32.opt: Change -mcmodel= default value.
9695 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9696 Monk Chiang <sh.chiang04@gmail.com>
9697 Chung-Ju Wu <jasonwucj@gmail.com>
9699 * config/nds32/constants.md (unspec_element): New enum.
9700 * config/nds32/constraints.md (Umw): New constraint.
9701 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9702 * config/nds32/nds32-intrinsic.md: Likewise.
9703 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9704 (nds32_valid_smw_lwm_base_p): New.
9705 (nds32_output_smw_single_word): New.
9706 (nds32_output_lmw_single_word): New.
9707 (nds32_expand_unaligned_load): New.
9708 (nds32_expand_unaligned_store): New.
9709 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9710 (nds32_output_smw_single_word): Declare.
9711 (nds32_output_lmw_single_word): Declare.
9712 (nds32_expand_unaligned_load): Declare.
9713 (nds32_expand_unaligned_store): Declare.
9714 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9715 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9716 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9717 NDS32_BUILTIN_UASTORE_DW.
9718 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9721 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9722 Kito Cheng <kito.cheng@gmail.com>
9723 Chung-Ju Wu <jasonwucj@gmail.com>
9725 * config/nds32/nds32-intrinsic.c
9726 (nds32_expand_builtin_null_ftype_reg): Delete.
9727 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9728 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9729 (nds32_read_argument): New.
9730 (nds32_legitimize_target): Ditto.
9731 (nds32_legitimize_argument): Ditto.
9732 (nds32_check_constant_argument): Ditto.
9733 (nds32_expand_unop_builtin): Ditto.
9734 (nds32_expand_unopimm_builtin): Ditto.
9735 (nds32_expand_binop_builtin): Ditto.
9736 (nds32_builtin_decl_impl): Ditto.
9737 (builtin_description): Ditto.
9738 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9739 (nds32_init_builtins_impl): Ditto.
9740 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9741 (nds32_builtin_decl): New.
9742 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9743 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9745 2018-03-02 Jeff Law <law@redhat.com>
9747 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9748 (redundant_insn, fill_simple_delay_slots): Likewise.
9749 (fill_slots_from_thread): Likewise.
9750 * resource.c (mark_referenced_resources): Likewise.
9751 (mark_set_resources, find_dead_or_set_registers): Likewise.
9753 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9755 * substring-locations.h (format_warning_va): Formatting fix for
9757 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9759 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9760 * substring-locations.c: Include intl.h.
9761 (format_warning_va): Turned into small wrapper around
9762 format_warning_n_va, renamed to ...
9763 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9764 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9766 (format_warning_at_substring_n): New function.
9767 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9768 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9769 format_warning_at_substring with just a shorter name instead of
9770 const function pointer.
9771 (fmtwarn_n): New function.
9772 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9773 appropriate, get rid of all the fmtstr temporaries, move conditionals
9774 with G_() wrapped string literals directly into fmtwarn arguments,
9775 cast dir.len to (int), formatting fixes.
9777 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9779 * doc/invoke.texi: Remove "Cilk Plus" references.
9781 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9782 Richard Biener <rguenther@suse.de>
9785 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9786 for error or warning attributes if CALL_FROM_THUNK_P is set.
9789 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9792 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9793 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9796 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9797 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9799 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9800 instead of -1U in last predictors element's probability member.
9802 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9805 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9806 arguments if they are comparable.
9808 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9810 PR tree-optimization/84634
9811 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9812 masks and masked_loop_p with a single loop_masks, making sure it's
9813 null for bb vectorization.
9815 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9817 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9818 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9819 loop->force_vectorize to check whether there is no alias.
9821 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9824 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9826 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9828 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9829 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9830 instead of a loop around prev_real_insn.
9831 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9835 * config/i386/i386.c (ix86_print_operand): Use conditional
9836 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9839 2018-03-02 Richard Biener <rguenther@suse.de>
9841 PR tree-optimization/84427
9842 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9843 (bitmap_set_subtract_values): Rewrite to handle multiple
9846 (prune_clobbered_mems): Likewise.
9847 (phi_translate): Take edge instead of pred/phiblock.
9848 (phi_translate_1): Likewise.
9849 (phi_translate_set): Likewise. Insert all translated
9850 exprs for a value into the set, keeping possibly multiple
9851 expressions per value.
9852 (compute_antic_aux): Adjust for phi_translate changes.
9853 When intersecting union the expressions and prune those
9854 not in the final value set, keeping possibly multiple
9855 expressions per value. Do not use value-insertion
9856 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9857 all expressions. Add verification that the value-sets
9858 only shrink during iteration.
9859 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9860 (do_pre_regular_insertion): Likewise.
9861 (do_pre_partial_partial_insertion): Likewise.
9863 2018-03-02 Richard Biener <rguenther@suse.de>
9866 * config/darwin.c (saved_debug_info_level): New static global.
9867 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9868 (darwin_asm_lto_end): Restore debug info generation settings.
9870 2018-03-01 Martin Liska <mliska@suse.cz>
9873 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9876 2018-03-01 Richard Biener <rguenther@suse.de>
9879 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9880 type annotation with LTO when debug was disabled at compile-time.
9882 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9884 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9886 (mips_final_postscan_insn): Likewise.
9888 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9890 PR rtl-optimization/84528
9891 * alias.c (init_alias_target): Add commentary.
9892 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9893 a unique base value if the frame pointer is not eliminated
9894 to the stack pointer.
9896 2018-03-01 Tom de Vries <tom@codesourcery.com>
9898 PR rtl-optimization/83327
9899 * lra-int.h (hard_regs_spilled_into): Declare.
9900 * lra.c (hard_regs_spilled_into): Define.
9901 (init_reg_info): Init hard_regs_spilled_into.
9902 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9903 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9904 (process_bb_lives): Handle hard_regs_spilled_into.
9905 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9906 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9908 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9910 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9911 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9912 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9913 * config/rs6000/aix72.h: New file.
9915 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9917 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9918 instead of warning_at with conditional singular and plural messages
9922 * stor-layout.c (update_alignment_for_field): For
9923 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9924 && !DECL_PACKED (field), do the alignment update, just use
9925 only desired_align instead of MAX (type_align, desired_align)
9927 (place_field): Don't do known_align < desired_align handling
9928 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9929 is non-NULL, instead do it after rli->prev_field handling and
9930 only if not within a bitfield word. For DECL_PACKED (field)
9931 use type_align of BITS_PER_UNIT.
9933 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9935 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9936 superfluous parentheses and trailing spaces.
9938 2018-02-28 Richard Biener <rguenther@suse.de>
9940 PR tree-optimization/84584
9941 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9942 SCoPs with fake exit edge.
9944 2018-02-28 Martin Liska <mliska@suse.cz>
9947 * timevar.c (timer::print): Fix format to properly print 100%
9950 2018-02-28 Richard Biener <rguenther@suse.de>
9953 * genmatch.c (capture_info::walk_match): Do not mark
9954 captured expressions without operands as expr_p given
9955 they act more like predicates and should be subject to
9956 "lost tail" side-effect preserving.
9958 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9960 PR rtl-optimization/81611
9961 * auto-inc-dec.c (attempt_change): Move dead note from
9962 mem_insn if it's the next use of regno
9963 (find_address): Take address use of reg holding
9964 non-incremented value. Add parm to limit search to the named
9966 (merge_in_block): Attempt to use a mem insn that is the next
9967 use of the original regno.
9969 2018-02-27 Martin Sebor <msebor@redhat.com>
9972 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9973 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9975 2018-02-27 Martin Sebor <msebor@redhat.com>
9977 PR translation/84207
9978 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9979 n argument to unsigned HOST_WIDE_INT.
9980 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9981 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9982 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9985 2018-02-27 Richard Biener <rguenther@suse.de>
9987 PR tree-optimization/84512
9988 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9989 Do not use the estimate returned from record_stmt_cost for
9990 the scalar iteration cost but sum properly using add_stmt_cost.
9992 2018-02-27 Richard Biener <rguenther@suse.de>
9994 PR tree-optimization/84466
9995 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9996 Adjust last change to less strictly validate use operands.
9998 2018-02-27 Martin Liska <mliska@suse.cz>
10000 PR gcov-profile/84548
10001 * gcov.c (process_file): Allow partial overlap and consider it
10002 also as group functions.
10003 (output_lines): Properly calculate range of lines for a group.
10005 2018-02-27 Martin Liska <mliska@suse.cz>
10007 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
10008 'ggc' suffixes. Change first column width.
10009 (timer::print): Fix formatting of the column.
10011 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
10013 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
10014 preserve inline entry blocks for the sake of debug inline
10015 entry point markers alone.
10016 (remove_unused_locals): Suggest in comments a better place to
10017 force the preservation of inline entry blocks that are
10018 otherwise unused, but do not preserve them.
10020 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10022 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
10024 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10027 * config/i386/constraints.md (Bs): Replace
10028 ix86_indirect_branch_register with
10029 TARGET_INDIRECT_BRANCH_REGISTER.
10031 * config/i386/i386.md (indirect_jump): Likewise.
10032 (tablejump): Likewise.
10033 (*sibcall_memory): Likewise.
10034 (*sibcall_value_memory): Likewise.
10035 Peepholes of indirect call and jump via memory: Likewise.
10036 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
10037 (*sibcall_value_GOT_32): Likewise.
10038 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10039 (GOT_memory_operand): Likewise.
10040 (call_insn_operand): Likewise.
10041 (sibcall_insn_operand): Likewise.
10042 (GOT32_symbol_operand): Likewise.
10043 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
10045 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10047 PR rtl-optimization/83496
10048 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
10049 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
10050 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
10051 redundant insn, if any.
10052 (relax_delay_slots): Likewise.
10053 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
10055 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
10057 PR tree-optimization/83965
10058 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
10059 that grouped statements are part of a reduction chain. Return
10060 true if the statement is not marked as a reduction itself but
10061 is part of a group.
10062 (vect_recog_dot_prod_pattern): Don't check whether the statement
10063 is part of a group here.
10064 (vect_recog_sad_pattern): Likewise.
10065 (vect_recog_widen_sum_pattern): Likewise.
10067 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
10070 * final.c (rest_of_clean_state): Also look for calls inside sequences.
10072 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
10075 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
10077 (ix86_output_indirect_function_return): New prototype.
10078 (ix86_split_simple_return_pop_internal): Likewise.
10079 * config/i386/i386.c (indirect_return_via_cx): New.
10080 (indirect_return_via_cx_bnd): Likewise.
10081 (indirect_thunk_name): Handle return va CX_REG.
10082 (output_indirect_thunk_function): Create alias for
10083 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
10084 (ix86_output_indirect_jmp): Remove the bool argument.
10085 (ix86_output_indirect_function_return): New function.
10086 (ix86_split_simple_return_pop_internal): Likewise.
10087 * config/i386/i386.md (*indirect_jump): Don't pass false
10088 to ix86_output_indirect_jmp.
10089 (*tablejump_1): Likewise.
10090 (simple_return_pop_internal): Change it to define_insn_and_split.
10091 Call ix86_split_simple_return_pop_internal to split it for
10092 -mfunction-return=.
10093 (simple_return_indirect_internal): Call
10094 ix86_output_indirect_function_return instead of
10095 ix86_output_indirect_jmp.
10097 2018-02-26 Jakub Jelinek <jakub@redhat.com>
10100 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
10101 memset and value initialization afterwards.
10103 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10105 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10107 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10110 * common/config/aarch64/aarch64-common.c
10111 (aarch_option_optimization_table[]): Switch
10112 off fomit-frame-pointer
10114 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10115 Chung-Ju Wu <jasonwucj@gmail.com>
10117 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10119 (store_multiple): Ditto.
10121 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10123 * config.gcc: Add --with-cpu support for nds32 target.
10124 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10125 * config/nds32/nds32.opt: Add -mcpu= option.
10127 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10129 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10130 isel=yes): Warn for these deprecated options.
10132 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10134 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10135 ISA_2_5_MASKS_EMBEDDED.
10137 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10139 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10140 p->max as pointers rather than using iterative_hash_expr.
10142 2018-02-23 Carl Love <cel@us.ibm.com>
10144 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10145 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10147 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10148 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10149 P8V_BUILTIN_VEC_VUNSIGNED2.
10151 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10154 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10155 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10157 * lra-constraints.c (curr_insn_transform): Set up
10158 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10160 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10161 * lra-eliminations.c (spill_pseudos): Ditto.
10162 (process_insn_for_elimination): Ditto.
10163 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10164 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10165 LRA_NON_CLOBBERED_ALT.
10167 2018-02-22 Martin Sebor <msebor@redhat.com>
10169 PR tree-optimization/84480
10170 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10171 to maybe_diag_stxncpy_trunc. Call it.
10172 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10173 from gimple_fold_builtin_strcpy. Print inlining stack.
10174 (handle_builtin_stxncpy): Print inlining stack.
10175 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10177 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10180 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10181 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10182 and -fcheck-pointer-bounds are used together.
10183 (indirect_thunk_prefix): New enum.
10184 (indirect_thunk_need_prefix): New function.
10185 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10186 "_nt" instead of "_bnd" for NOTRACK prefix.
10187 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10188 (output_indirect_thunk_function): Likewise.
10190 (ix86_code_end): Update output_indirect_thunk_function calls.
10191 (ix86_output_indirect_branch_via_reg): Replace
10192 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10193 (ix86_output_indirect_branch_via_push): Likewise.
10194 (ix86_output_function_return): Likewise.
10195 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10196 incompatible with -fcf-protection=branch and
10197 -fcheck-pointer-bounds.
10199 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10202 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10203 Change gcc_assert call to output_operand_lossage.
10205 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10207 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10209 2018-02-22 DJ Delorie <dj@redhat.com>
10210 Sebastian Perta <sebastian.perta@renesas.com>
10211 Oleg Endo <olegendo@gcc.gnu.org>
10213 * config/rx/rx.c (rx_rtx_costs): New function.
10214 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10216 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10218 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10220 2018-02-22 Martin Liska <mliska@suse.cz>
10223 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10224 Add "native" as a possible value.
10226 2018-02-22 Martin Liska <mliska@suse.cz>
10229 * config/i386/i386.c (ix86_option_override_internal):
10230 Add "native" as a possible value for -march and -mtune.
10232 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10235 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10236 to all type variants.
10238 PR tree-optimization/84503
10239 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10240 width as info->bitpos + info->bitsize - start.
10241 (merged_store_group::merge_overlapping): Simplify width computation.
10242 (check_no_overlap): New function.
10243 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10244 start + width and last_order of the group, fail if check_no_overlap
10246 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10247 to group if check_no_overlap fails.
10249 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10251 * config/rs6000/altivec.md: Delete contraint arguments to
10252 define_expand, define_split, and define_peephole2, and in
10253 define_insn_and_split if always unused.
10254 * config/rs6000/darwin.md: Ditto.
10255 * config/rs6000/dfp.md: Ditto.
10256 * config/rs6000/rs6000.md: Ditto.
10257 * config/rs6000/sync.md: Ditto.
10258 * config/rs6000/vector.md: Ditto.
10259 * config/rs6000/vsx.md: Ditto.
10261 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10263 * config/rs6000/altivec.md: Write output control strings as braced
10264 blocks instead of double-quoted strings.
10265 * config/rs6000/darwin.md: Ditto.
10266 * config/rs6000/rs6000.md: Ditto.
10267 * config/rs6000/vector.md: Ditto.
10268 * config/rs6000/vsx.md: Ditto.
10270 2018-02-21 Jason Merrill <jason@redhat.com>
10272 PR c++/84314 - ICE with templates and fastcall attribute.
10273 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10275 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10277 * ipa-cp.c (determine_versionability): Fix comment typos.
10279 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10282 * ipa-cp.c (determine_versionability): Do not version functions caling
10285 2018-02-21 Martin Liska <mliska@suse.cz>
10288 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10289 Add "native" as a possible value.
10290 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10291 the macro when native cpu detection is available.
10293 2018-02-21 Martin Liska <mliska@suse.cz>
10296 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10297 Add "native" as a possible value.
10298 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10299 when native cpu detection is available.
10301 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10302 Martin Sebor <msebor@redhat.com>
10304 PR tree-optimization/84478
10305 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10307 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10308 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10309 support which is conservatively correct, for 2 only stay conservative
10310 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10311 argument to the 2 argument get_range_strlen, adjust 6 arg
10312 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10314 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10315 (gimple_fold_builtin_strlen): Pass true as last argument to
10318 2018-02-20 Martin Sebor <msebor@redhat.com>
10320 PR middle-end/84095
10321 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10322 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10323 (builtin_memref::builtin_memref): Factor out parts into
10324 set_base_and_offset and call it.
10326 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10328 PR middle-end/84406
10329 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10330 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10331 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10332 search at the associated MODE_INT.
10334 2018-02-20 Jeff Law <law@redhat.com>
10336 PR middle-end/82123
10337 PR tree-optimization/81592
10338 PR middle-end/79257
10339 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10340 for range data rather than using global data.
10341 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10342 range data rather than using global data.
10343 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10344 pass it to children as needed.
10345 (struct directive::fmtresult): Similarly.
10346 (struct directive::set_width): Similarly.
10347 (struct directive::set_precision): Similarly.
10348 (format_integer, format_directive, parse_directive): Similarly.
10349 (format_none): Accept unnamed vr_values parameter.
10350 (format_percent, format_floating, format_character): Similarly.
10351 (format_string, format_plain): Similarly.
10352 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10353 the EVRP range analyzer for range data rather than using global data.
10354 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10355 gimple-ssa-evrp-analyze.h
10356 (class sprintf_dom_walker): Add after_dom_children member function.
10357 Add evrp_range_analyzer member.
10358 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10359 range analyzer as needed.
10360 (sprintf_dom_walker::after_dom_children): New member function.
10361 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10363 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10364 (evrp_range_analyzer::pop_to_marker): Likewise.
10366 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10368 PR tree-optimization/84419
10369 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10370 with the required type if its current type is compatible but
10373 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10375 PR middle-end/82004
10376 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10377 after vectorization.
10379 2018-02-20 Martin Liska <mliska@suse.cz>
10382 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10383 possible values if we don't have a hint.
10385 2018-02-20 Martin Liska <mliska@suse.cz>
10389 * final.c (shorten_branches): Build align_tab array with one
10391 * opts.c (finish_options): Add alignment option limit check.
10392 (MAX_CODE_ALIGN): Likewise.
10393 (MAX_CODE_ALIGN_VALUE): Likewise.
10394 * doc/invoke.texi: Document maximum allowed option value for
10395 all -falign-* options.
10397 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10400 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10401 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10402 * var-tracking.c (emit_note_insn_var_location): Remove all references
10403 to NOTE_INSN_CALL_ARG_LOCATION.
10404 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10405 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10406 Use copy_rtx_if_shared.
10407 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10408 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10409 (dwarf2out_var_location): Remove handling of
10410 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10412 * final.c (final_scan_insn): Remove all references to
10413 NOTE_INSN_CALL_ARG_LOCATION.
10414 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10415 before dumping final insns.
10416 * except.c (emit_note_eh_region_end): Remove all references to
10417 NOTE_INSN_CALL_ARG_LOCATION.
10418 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10419 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10420 * config/arc/arc.c (hwloop_optimize): Likewise.
10421 * config/arm/arm.c (create_fix_barrier): Likewise.
10422 * config/s390/s390.c (s390_chunkify_start): Likewise.
10423 * config/sh/sh.c (find_barrier): Likewise.
10424 * config/i386/i386.c (rest_of_insert_endbranch,
10425 ix86_seh_fixup_eh_fallthru): Likewise.
10426 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10427 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10428 * config/frv/frv.c (frv_function_prologue): Likewise.
10429 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10431 (note_outside_basic_block_p): Remove all references to
10432 NOTE_INSN_CALL_ARG_LOCATION.
10433 * gengtype.c (adjust_field_rtx_def): Likewise.
10434 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10436 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10437 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10440 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10443 PR tree-optimization/84452
10444 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10445 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10448 2018-02-19 Martin Liska <mliska@suse.cz>
10451 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10453 2018-02-19 Martin Liska <mliska@suse.cz>
10454 Richard Sandiford <richard.sandiford@linaro.org>
10456 PR tree-optimization/82491
10457 * gimple-fold.c (get_base_constructor): Make earlier bail out
10460 2018-02-19 Carl Love <cel@us.ibm.com>
10462 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10463 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10465 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10466 P8V_BUILTIN_VEC_NEG.
10468 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10470 * config/rl78/rl78.md (movdf): New define expand.
10472 2018-02-19 Martin Liska <mliska@suse.cz>
10475 * doc/invoke.texi: Fix typo.
10476 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10478 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10480 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10481 handle rs6000_single_float and rs6000_double_float specially for
10484 2018-02-16 Jeff Law <law@redhat.com>
10486 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10487 (pop_regs): Use it.
10489 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10492 * ipa-inline.c (inline_small_functions): Fix a typo.
10494 2018-02-16 Nathan Sidwell <nathan@acm.org>
10496 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10498 2018-02-16 Carl Love <cel@us.ibm.com>
10500 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10501 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10502 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10503 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10504 expansion to P8V_BUILTIN_VEC_FLOAT2.
10506 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10508 PR rtl-optimization/70023
10509 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10510 src_regno into account.
10512 2018-02-16 Carl Love <cel@us.ibm.com>
10514 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10515 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10516 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10517 * config/rs6000/rs6000.c: Remove case statements for
10518 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10519 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10520 and P9V_BUILTIN_VEC_VINSERT4B.
10521 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10522 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10523 * config/rs6000/vsx.md:
10524 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10527 2018-02-16 Carl Love <cel@us.ibm.com>
10529 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10531 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10533 * config/rs6000/rs6000-c.c: Add the definitions for
10534 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10535 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10536 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10537 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10538 definition for insert4b and define insn *insert3b_internal.
10539 * doc/extend.texi: Add documentation for vec_extract4b.
10541 2018-02-16 Nathan Sidwell <nathan@acm.org>
10543 * doc/extend.texi (Backwards Compatibility): Mention friend
10544 injection. Note for-scope is deprecated.
10545 * doc/invoke.texi (-ffriend-injection): Deprecate.
10547 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10549 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10550 that moved to I2, also allow destinations that are a paradoxical
10551 subreg (instead of a normal reg).
10553 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10556 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10559 2018-02-16 Richard Biener <rguenther@suse.de>
10561 PR tree-optimization/84037
10562 PR tree-optimization/84016
10564 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10565 Adjust vec_construct for the fact we need additional higher latency
10566 128bit inserts for AVX256 and AVX512 vector builds.
10567 (ix86_add_stmt_cost): Scale vector construction cost for
10570 2018-02-16 Richard Biener <rguenther@suse.de>
10572 PR tree-optimization/84417
10573 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10574 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10575 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10577 2018-02-16 Martin Liska <mliska@suse.cz>
10580 * internal-fn.def (ASAN_CHECK): Set proper flags.
10581 (ASAN_MARK): Likewise.
10583 2018-02-16 Julia Koval <julia.koval@intel.com>
10585 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10586 from PTA_CANNONLAKE.
10588 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10591 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10592 Use ++iter rather than iter++ for std::list iterators.
10593 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10594 defer deleting them until all nodes in the forest are processed. Do
10595 free even leaf nodes. Change to_process into auto_vec.
10598 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10599 * vec.h (vec_default_construct): Use memset instead of placement new
10600 if BROKEN_VALUE_INITIALIZATION is defined.
10601 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10602 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10605 PR rtl-optimization/83723
10606 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10607 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10608 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10610 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10612 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10614 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10616 PR rtl-optimization/81443
10617 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10618 from inner REGs to paradoxical SUBREGs.
10620 2018-02-16 Richard Biener <rguenther@suse.de>
10622 PR tree-optimization/84399
10623 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10624 For operands we can analyze at their definition make sure we can
10625 analyze them at each use as well.
10627 2018-02-16 Richard Biener <rguenther@suse.de>
10629 PR tree-optimization/84190
10630 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10631 volatile accesses if the decl isn't volatile.
10633 2018-02-15 Jason Merrill <jason@redhat.com>
10635 PR c++/84314 - ICE with templates and fastcall attribute.
10636 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10637 TYPE_CANONICAL on an existing type.
10639 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10641 PR tree-optimization/84383
10642 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10643 dstoff nor call operand_equal_p if dstbase is NULL.
10645 PR tree-optimization/84334
10646 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10647 also a CONSTANT_CLASS_P, punt.
10649 2018-02-14 Jim Wilson <jimw@sifive.com>
10651 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10652 first SMALL_OPERAND check. New local min_second_step. Move assert
10653 to where locals are set. Add TARGET_RVC support.
10654 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10656 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10658 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10660 2018-02-14 Martin Sebor <msebor@redhat.com>
10662 PR tree-optimization/83698
10663 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10664 arrays constrain the offset range to their bounds.
10665 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10666 (builtin_access::overlap): Avoid setting the size of overlap if it's
10668 (maybe_diag_overlap): Also consider arrays when deciding what values
10669 of offsets to include in diagnostics.
10671 2018-02-14 Martin Sebor <msebor@redhat.com>
10674 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10675 that correspond to the kind of a declaration.
10677 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10680 * config/pa/pa.md: Load address of PIC label using the linkage table
10681 if the label is nonlocal.
10683 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10685 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10686 warning message if user requests -maltivec=be.
10687 * doc/invoke.texi: Document deprecation of -maltivec=be.
10689 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10692 * config/rs6000/rs6000-c.c: Update definitions for
10693 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10694 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10696 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10699 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10700 add _get_ssp intrinsics. Remove argument from
10701 __builtin_ia32_rdssp[d|q].
10702 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10703 * config/i386/i386-builtin.def: Remove argument from
10704 __builtin_ia32_rdssp[d|q].
10705 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10706 ix86_expand_special_args_builtin for _rdssp[d|q].
10707 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10708 Clear register before usage.
10709 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10710 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10712 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10714 PR tree-optimization/84357
10715 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10716 operand 1 of an ARRAY_REF too.
10718 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10721 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10722 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10724 (set_of_reg): New struct.
10725 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10726 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10727 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10729 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10730 Split into bitclr, bitset, bitinvert patterns if appropriate.
10731 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10732 use rx_fuse_in_memory_bitop.
10733 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10734 to named insn, correct maximum insn length.
10736 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10739 * machmode.def: Define a complex mode for PARTIAL_INT.
10740 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10742 * doc/rtl.texi: Document CSPImode.
10743 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10745 (msp430_hard_regno_nregs_with_padding): Likewise.
10747 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10750 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10752 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10754 PR rtl-optimization/84169
10755 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10756 we generated a parallel as new i3 and we split that to new i2 and i3
10757 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10758 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10759 those to i2, not i1. Partially rewrite this scan code.
10761 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10764 * stor-layout.c (place_field): For variable length fields, adjust
10765 offset_align afterwards not just based on the field's alignment,
10766 but also on the size.
10768 PR middle-end/84309
10769 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10770 of exps and logs in the use_exp2 case.
10772 2018-02-13 Jeff Law <law@redhat.com>
10774 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10775 entry for "vector".
10777 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10780 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10784 * common.opt (gas-loc-support, gas-locview-support): New.
10785 (ginline-points, ginternal-reset-location-views): New.
10786 * doc/invoke.texi: Document them. Use @itemx where intended.
10787 (gvariable-location-views): Adjust.
10788 * target.def (reset_location_view): New.
10789 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10790 (TARGET_RESET_LOCATION_VIEW): New.
10791 * doc/tm.texi: Rebuilt.
10792 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10793 (dwarf2out_default_as_locview_support): New.
10794 (output_asm_line_debug_info): Use option variables.
10795 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10796 (output_loc_list): Likewise.
10797 (add_high_low_attributes): Check option variables.
10798 Don't output entry view attribute in strict mode.
10799 (gen_inlined_subroutine_die): Check option variables.
10800 (dwarf2out_inline_entry): Likewise.
10801 (init_sections_and_labels): Likewise.
10802 (dwarf2out_early_finish): Likewise.
10803 (maybe_reset_location_view): New, from...
10804 (dwarf2out_var_location): ... here. Call it.
10805 * debug.h (dwarf2out_default_as_loc_support): Declare.
10806 (dwarf2out_default_as_locview_support): Declare.
10807 * hooks.c (hook_int_rtx_insn_0): New.
10808 * hooks.h (hook_int_rtx_insn_0): Declare.
10809 * toplev.c (process_options): Take -gas-loc-support and
10810 -gas-locview-support from dwarf2out. Enable
10811 -gvariable-location-views by default only with locview
10812 assembler support. Enable -ginternal-reset-location-views by
10813 default only if the target defines the corresponding hook.
10814 Enable -ginline-points by default if location views are
10815 enabled; force it disabled if statement frontiers are
10817 * tree-inline.c (expand_call_inline): Check option variables.
10818 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10820 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10822 PR tree-optimization/84321
10823 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10824 handling. Also check whether the anti-range contains any values
10825 that satisfy the mask; switch to a VR_RANGE if not.
10827 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10830 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10832 2018-02-13 Martin Jambor <mjambor@suse.cz>
10835 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10836 of call statements, also set location of a load to a temporary.
10838 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10840 * config/rl78/rl78.c (add_vector_labels): New function.
10841 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10842 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10843 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10844 which checks that no arguments are passed.
10845 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10846 * doc/extend.texi: Documentation for the new attribute.
10848 2018-02-13 Andreas Schwab <schwab@suse.de>
10850 * config/riscv/linux.h (CPP_SPEC): Define.
10852 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10855 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10856 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10857 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10858 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10859 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10860 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10861 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10862 temporarily for AES and PCLMUL builtins.
10864 PR tree-optimization/84339
10865 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10866 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10869 PR middle-end/84309
10870 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10871 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10872 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10874 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10876 * omp-simd-clone.h: New file.
10877 * omp-simd-clone.c: Include omp-simd-clone.h.
10878 (expand_simd_clones): No longer static.
10879 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10880 cgraph.h and omp-simd-clone.h.
10881 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10882 (vect_recog_widen_shift_pattern): Formatting fix.
10883 (vect_pattern_recog_1): Don't check optab for calls.
10886 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10887 operands[2] into a REG before using gen_lowpart on it.
10889 2018-02-12 Jeff Law <law@redhat.com>
10892 * config/sh/sh.c (find_barrier): Consider a sibling call
10895 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10896 successfully back substituting a reg.
10898 2018-02-12 Richard Biener <rguenther@suse.de>
10900 PR tree-optimization/84037
10901 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10902 parameter, move visited init to caller.
10903 (vect_slp_analyze_operations): Separate cost from validity
10904 check, initialize visited once for all instances.
10905 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10907 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10909 (vect_model_promotion_demotion_cost): Likewise.
10910 (vectorizable_bswap): Guard cost modeling with !slp_node
10911 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10913 (vectorizable_call): Likewise.
10914 (vectorizable_conversion): Likewise.
10915 (vectorizable_assignment): Likewise.
10916 (vectorizable_shift): Likewise.
10917 (vectorizable_operation): Likewise.
10918 (vectorizable_store): Likewise.
10919 (vectorizable_load): Likewise.
10920 (vectorizable_condition): Likewise.
10921 (vectorizable_comparison): Likewise.
10923 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10926 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10927 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10928 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10930 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10932 PR middle-end/83665
10933 * params.def (inline-min-speedup): Increase from 8 to 15.
10934 (max-inline-insns-auto): Decrease from 40 to 30.
10935 * ipa-split.c (consider_split): Add some buffer for function to
10936 be considered inlining candidate.
10937 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10940 2018-02-12 Richard Biener <rguenther@suse.de>
10942 PR tree-optimization/84037
10943 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10944 matched stmts if we cannot swap the non-matched ones.
10946 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10948 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10949 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10950 _mm_maskz_scalef_round_ss): New intrinsics.
10951 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10952 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10953 __builtin_ia32_scalefss_round): Remove.
10954 (__builtin_ia32_scalefsd_mask_round,
10955 __builtin_ia32_scalefss_mask_round): New intrinsics.
10956 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10957 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10958 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10959 "<round_constraint>")): Changed to ...
10960 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10961 "<round_scalar_constraint>")): ... this.
10962 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10963 %0, %1, %2<round_op3>}"): Changed to ...
10964 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10965 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10966 %2<round_scalar_mask_op3>}"): ... this.
10967 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10969 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10971 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10972 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10973 (_mm_maskz_sqrt_round_ss): New intrinsics.
10974 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10975 (__builtin_ia32_sqrtsd_mask_round)
10976 (__builtin_ia32_sqrtss_mask_round): New builtins.
10977 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10978 (__builtin_ia32_sqrtss_round): Remove.
10979 (__builtin_ia32_sqrtsd_mask_round)
10980 (__builtin_ia32_sqrtss_mask_round): New builtins.
10981 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10982 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10983 ((match_operand:VF_128 1 "vector_operand"
10984 "xBm,<round_constraint>")): Changed to ...
10985 ((match_operand:VF_128 1 "vector_operand"
10986 "xBm,<round_scalar_constraint>")): ... this.
10987 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10988 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10989 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10990 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10991 %<iptr>1<round_scalar_mask_op3>}): ... this.
10992 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10993 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10995 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10998 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10999 Cast vec_cmpeq result to correct type.
11000 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
11001 Cast vec_cmpgt result to correct type.
11003 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
11005 * final.c (final_scan_insn_1): Renamed from...
11006 (final_scan_insn): ... this. New wrapper, to recover
11007 seen from the outermost call in recursive ones.
11008 * config/sparc/sparc.c (output_return): Drop seen from call.
11009 (output_sibcall): Likewise.
11010 * config/visium/visium.c (output_branch): Likewise.
11012 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
11014 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
11017 2018-02-10 Alan Modra <amodra@gmail.com>
11020 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
11021 Specify LR as an input.
11023 2018-02-10 Jakub Jelinek <jakub@redhat.com>
11026 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
11027 remove_member_access_dummy_vars): New functions.
11028 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
11029 lower_omp_1, execute_lower_omp): Use them.
11031 PR rtl-optimization/84308
11032 * shrink-wrap.c (spread_components): Release todo vector.
11034 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
11036 PR rtl-optimization/57193
11037 * ira-color.c (struct allocno_color_data): Add member
11038 conflict_allocno_hard_prefs.
11039 (update_conflict_allocno_hard_prefs): New.
11040 (bucket_allocno_compare_func): Add a preference based on
11041 conflict_allocno_hard_prefs.
11042 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
11043 (color_allocnos): Remove a dead code. Initiate
11044 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
11046 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11049 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
11050 constraint from =wa to wa. Avoid a subreg on the output operand,
11051 instead use a pseudo and subreg it in a move.
11052 (p9_xxbrd_<mode>): Changed to ...
11053 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
11054 (p9_xxbrd_v2df): New expander.
11055 (p9_xxbrw_<mode>): Changed to ...
11056 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
11057 (p9_xxbrw_v4sf): New expander.
11059 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11061 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
11063 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
11066 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
11067 multiply in 32-bit mode.
11068 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
11069 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
11072 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
11074 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
11075 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
11076 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
11077 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
11079 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
11082 * dwarf2out.c (is_trivial_indirect_ref): New function.
11083 (dwarf2out_late_global_decl): Do not generate a location
11084 attribute for variables that have a non-trivial DECL_VALUE_EXPR
11085 and that are not defined in the current unit.
11087 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11089 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
11090 instead of a libcall for UNORDERED.
11092 2018-02-09 Tamar Christina <tamar.christina@arm.com>
11095 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
11096 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
11098 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11101 * config/s390/s390.c (s390_set_current_function): Invoke
11102 s390_indirect_branch_settings also if fndecl didn't change.
11104 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11106 * config/rs6000/rs6000.md (blockage): Set length to zero.
11108 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11110 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11112 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11115 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11116 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11120 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11121 PARALLEL incoming that failed vt_get_decl_and_offset check.
11123 PR middle-end/84237
11124 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11125 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11127 (get_variable_section): For decls in named .bss* sections pass true as
11128 second argument to bss_initializer_p.
11130 2018-02-09 Marek Polacek <polacek@redhat.com>
11131 Jakub Jelinek <jakub@redhat.com>
11134 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11135 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11136 Sync some changes from cxx_fold_indirect_ref.
11138 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11140 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11142 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11143 (BLOCK_INLINE_ENTRY_LABEL): New.
11144 (dwarf2out_var_location): Disregard inline entry markers.
11145 (inline_entry_data): New struct.
11146 (inline_entry_data_hasher): New hashtable type.
11147 (inline_entry_data_hasher::hash): New.
11148 (inline_entry_data_hasher::equal): New.
11149 (inline_entry_data_table): New variable.
11150 (add_high_low_attributes): Add DW_AT_entry_pc and
11151 DW_AT_GNU_entry_view attributes if a pending entry is found
11152 in inline_entry_data_table. Add old entry_pc attribute only
11153 if debug nonbinding markers are disabled.
11154 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11155 markers are enabled.
11156 (block_within_block_p, dwarf2out_inline_entry): New.
11157 (dwarf2out_finish): Check that no entries remained in
11158 inline_entry_data_table.
11159 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11160 (final_scan_insn, notice_source_line): Likewise.
11161 (rest_of_clean_state): Skip inline entry markers.
11162 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11164 * gimple.c (gimple_build_debug_inline_entry): New.
11165 * gimple.h (enum gimple_debug_subcode): Add
11166 GIMPLE_DEBUG_INLINE_ENTRY.
11167 (gimple_build_debug_inline_entry): Declare.
11168 (gimple_debug_inline_entry_p): New.
11169 (gimple_debug_nonbind_marker_p): Adjust.
11170 * insn-notes.def (INLINE_ENTRY): New.
11171 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11172 inline entry marker notes.
11173 (print_insn): Likewise.
11174 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11175 (INSN_DEBUG_MARKER_KIND): Likewise.
11176 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11177 * tree-inline.c (expand_call_inline): Build and insert
11178 debug_inline_entry stmt.
11179 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11180 inline entry blocks early, if nonbind markers are enabled.
11181 (dump_scope_block): Dump fragment info.
11182 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11183 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11184 (gimple_build_debug_inline_entry): New.
11185 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11186 Enable/disable inline entry points too.
11187 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11188 (DEBUG_INSN): Describe inline entry markers.
11190 * common.opt (gvariable-location-views): New.
11191 (gvariable-location-views=incompat5): New.
11192 * config.in: Rebuilt.
11193 * configure: Rebuilt.
11194 * configure.ac: Test assembler for view support.
11195 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11196 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11197 * dwarf2out.c (var_loc_view): New typedef.
11198 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11199 (dwarf2out_locviews_in_attribute): New.
11200 (dwarf2out_locviews_in_loclist): New.
11201 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11202 (enum dw_line_info_opcode): Add LI_adv_address.
11203 (struct dw_line_info_table): Add view.
11204 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11205 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11206 (zero_view_p): New variable.
11207 (ZERO_VIEW_P): New macro.
11208 (output_asm_line_debug_info): New.
11209 (struct var_loc_node): Add view.
11210 (add_AT_view_list, AT_loc_list): New.
11211 (add_var_loc_to_decl): Add view param. Test it against last.
11212 (new_loc_list): Add view params. Record them.
11213 (AT_loc_list_ptr): Handle loc and view lists.
11214 (view_list_to_loc_list_val_node): New.
11215 (print_dw_val): Handle dw_val_class_view_list.
11216 (size_of_die): Likewise.
11217 (value_format): Likewise.
11218 (loc_list_has_views): New.
11219 (gen_llsym): Set vl_symbol too.
11220 (maybe_gen_llsym, skip_loc_list_entry): New.
11221 (dwarf2out_maybe_output_loclist_view_pair): New.
11222 (output_loc_list): Output view list or entries too.
11223 (output_view_list_offset): New.
11224 (output_die): Handle dw_val_class_view_list.
11225 (output_dwarf_version): New.
11226 (output_compilation_unit_header): Use it.
11227 (output_skeleton_debug_sections): Likewise.
11228 (output_rnglists, output_line_info): Likewise.
11229 (output_pubnames, output_aranges): Update version comments.
11230 (output_one_line_info_table): Output view numbers in asm comments.
11231 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11232 Call maybe_gen_llsym.
11233 (loc_list_from_tree_1): Adjust.
11234 (add_AT_location_description): Create view list attribute if
11235 needed, check it's absent otherwise.
11236 (convert_cfa_to_fb_loc_list): Adjust.
11237 (maybe_emit_file): Call output_asm_line_debug_info for test.
11238 (dwarf2out_var_location): Reset views as needed. Precompute
11239 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11240 attribute. Set view.
11241 (new_line_info_table): Reset next view.
11242 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11243 (dwarf2out_source_line): Likewise. Output view resets and labels to
11244 the assembler, or select appropriate line info opcodes.
11245 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11246 (optimize_string_length): Catch it. Adjust.
11247 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11248 dw_val_class_view_list, and remove it if no longer needed.
11249 (hash_loc_list): Hash view numbers.
11250 (loc_list_hasher::equal): Compare them.
11251 (optimize_location_lists): Check whether a view list symbol is
11252 needed, and whether the locview attribute is present, and
11253 whether they match. Remove the locview attribute if no longer
11255 (index_location_lists): Call skip_loc_list_entry for test.
11256 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11257 Use output_dwarf_version.
11258 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11259 (struct dw_val_node): Add val_view_list.
11260 * final.c (SEEN_NEXT_VIEW): New.
11261 (set_next_view_needed): New.
11262 (clear_next_view_needed): New.
11263 (maybe_output_next_view): New.
11264 (final_start_function): Rename to...
11265 (final_start_function_1): ... this. Take pointer to FIRST,
11266 add SEEN parameter. Emit param bindings in the initial view.
11267 (final_start_function): Reintroduce SEEN-less interface.
11268 (final): Rename to...
11269 (final_1): ... this. Take SEEN parameter. Output final pending
11270 next view at the end.
11271 (final): Reintroduce seen-less interface.
11272 (final_scan_insn): Output pending next view before switching
11273 sections or ending a block. Mark the next view as needed when
11274 outputting variable locations. Notify debug backend of section
11275 changes, and of location view changes.
11276 (rest_of_handle_final): Adjust.
11277 * toplev.c (process_options): Autodetect value for debug variable
11278 location views option. Warn on incompat5 without -gdwarf-5.
11279 * doc/invoke.texi (gvariable-location-views): New.
11280 (gvariable-location-views=incompat5): New.
11281 (gno-variable-location-views): New.
11283 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11285 PR tree-optimization/84136
11286 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11287 that the result of find_edge is non-NULL.
11289 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11292 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11293 storing integer register in SImode. Fix cost of 256 and 512
11294 byte aligned SSE register store.
11296 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11298 * config/i386/i386.c (ix86_multiplication_cost): Fix
11299 multiplication cost for TARGET_AVX512DQ.
11301 2018-02-08 Marek Polacek <polacek@redhat.com>
11303 PR tree-optimization/84238
11304 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11307 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11309 PR tree-optimization/84265
11310 * tree-vect-stmts.c (vectorizable_store): Don't treat
11311 VMAT_CONTIGUOUS accesses as grouped.
11312 (vectorizable_load): Likewise.
11314 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11316 PR tree-optimization/81635
11317 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11318 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11319 (test_round_for_mask): New functions.
11320 (wide_int_cc_tests): Call test_round_for_mask.
11321 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11322 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11323 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11324 range returned by get_range_info.
11326 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11329 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11330 * symtab.c: Include builtins.h
11331 (symtab_node::output_to_lto_symbol_table_p): Move here
11332 from lto-streamer-out.c:output_symbol_p.
11333 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11334 (output_symbol_p): Move all logic to symtab.c
11335 (produce_symtab): Update.
11337 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11339 * config/s390/s390-opts.h (enum indirect_branch): Define.
11340 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11341 (s390_indirect_branch_via_thunk)
11342 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11343 (enum s390_indirect_branch_type): Define.
11344 * config/s390/s390.c (struct s390_frame_layout, struct
11345 machine_function): Remove.
11346 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11347 (indirect_branch_table_label_no, indirect_branch_table_name):
11349 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11350 (enum s390_indirect_branch_option): Define.
11351 (s390_return_addr_from_memory): New function.
11352 (s390_handle_string_attribute): New function.
11353 (s390_attribute_table): Add new attribute handler.
11354 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11355 (s390_indirect_branch_via_thunk): New function.
11356 (s390_indirect_branch_via_inline_thunk): New function.
11357 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11358 sibling call optimization for non z10 compiles.
11359 (s390_emit_call): Force indirect branch target to be a single
11360 register. Add r1 clobber for non-z10 compiles.
11361 (s390_emit_epilogue): Emit return jump via return_use expander.
11362 (s390_reorg): Handle JUMP_INSNs as execute targets.
11363 (s390_option_override_internal): Perform validity checks for the
11364 new command line options.
11365 (s390_indirect_branch_attrvalue): New function.
11366 (s390_indirect_branch_settings): New function.
11367 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11368 (s390_output_indirect_thunk_function): New function.
11369 (s390_code_end): Implement target hook.
11370 (s390_case_values_threshold): Implement target hook.
11371 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11373 * config/s390/s390.h (struct s390_frame_layout)
11374 (struct machine_function): Move here from s390.c.
11375 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11376 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11377 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11378 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11379 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11380 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11381 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11382 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11383 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11384 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11385 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11386 (mnemonic attribute): Add values which aren't recognized
11388 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11389 pattern for branch conversion. Fix mnemonic attribute.
11390 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11391 indirect branch via thunk if requested.
11392 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11393 ("*indirect_jump"): Disable for branch conversion using out of
11395 ("indirect_jump_via_thunk<mode>_z10")
11396 ("indirect_jump_via_thunk<mode>")
11397 ("indirect_jump_via_inlinethunk<mode>_z10")
11398 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11399 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11400 ("casesi_jump_via_inlinethunk<mode>_z10")
11401 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11402 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11403 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11404 ("*indirect2_jump"): Disable for branch conversion.
11405 ("casesi_jump"): Turn into expander and expand patterns for branch
11407 ("return_use"): New expander.
11408 ("*return"): Emit return via thunk and rename it to ...
11409 ("*return<mode>"): ... this one.
11410 * config/s390/s390.opt: Add new options and and enum for the
11413 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11415 * lra-constraints.c (match_reload): Unconditionally use
11416 gen_lowpart_SUBREG, rather than selecting between that
11417 and equivalent gen_rtx_SUBREG code.
11419 2018-02-08 Richard Biener <rguenther@suse.de>
11421 PR tree-optimization/84233
11422 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11423 changed flag instead of boguously re-using phi_inserted.
11425 2018-02-08 Martin Jambor <mjambor@suse.cz>
11427 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11428 static local variables.
11430 2018-02-08 Richard Biener <rguenther@suse.de>
11432 PR tree-optimization/84278
11433 * tree-vect-stmts.c (vectorizable_store): When looking for
11434 smaller vector types to perform grouped strided loads/stores
11435 make sure the mode is supported by the target.
11436 (vectorizable_load): Likewise.
11438 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11440 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11441 Increase LDP/STP opportunities by adding adjacent callee-saves.
11443 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11445 PR rtl-optimization/84068
11446 PR rtl-optimization/83459
11447 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11449 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11451 PR tree-optimization/84224
11452 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11453 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11454 non-zero arguments.
11456 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11459 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11460 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11461 count, remove one USE.
11463 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11465 * doc/install.texi (Configuration): Document the
11466 --with-long-double-format={ibm,ieee} PowerPC configuration
11470 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11471 Convert from define_expand to be define_insn_and_split. Rework
11472 float/double/_Float128 conversions to QI/HI/SImode to work with
11473 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11474 conversions to QI/HImode types did a store and then a load to
11475 truncate the value. For conversions to VSX registers, don't split
11476 the insn, instead emit the code directly. Use the code iterator
11477 any_fix to combine signed and unsigned conversions.
11478 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11479 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11480 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11481 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11482 (fix_<mode>di2_hw): Likewise.
11483 (fixuns_<mode>di2_hw): Likewise.
11484 (fix_<mode>si2_hw): Likewise.
11485 (fixuns_<mode>si2_hw): Likewise.
11486 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11487 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11488 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11489 fix<uns>_trunc<SFDF:mode>si2_p8.
11490 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11491 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11492 (fix<uns>_<mode>_mem): Likewise.
11493 (fctiw<u>z_<mode>_mem): Likewise.
11494 (fix<uns>_<mode>_mem): Likewise.
11495 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11496 the register allocator from doing a direct move to the GPRs to do
11497 a store, and instead use the ISA 3.0 store byte/half-word from
11498 vector register instruction. For IEEE 128-bit floating point,
11499 also optimize stores of 32-bit ints.
11500 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11502 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11504 * genextract.c (push_pathstr_operand): New function to support
11506 (walk_rtx): Call push_pathstr_operand.
11507 (print_path): Support [a-zA-Z].
11509 2018-02-07 Richard Biener <rguenther@suse.de>
11511 PR tree-optimization/84037
11512 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11513 (cse_and_gimplify_to_preheader): Declare.
11514 (vect_get_place_in_interleaving_chain): Likewise.
11515 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11517 (_loop_vec_info::~_loop_vec_info): Delete it.
11518 (cse_and_gimplify_to_preheader): New function.
11519 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11520 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11521 (vectorizable_load): Likewise. For grouped stores always base
11522 the IV on the first element.
11523 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11524 condition before gimplifying.
11526 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11528 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11529 *DIV_EXPR and *MOD_EXPR.
11531 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11534 * config/i386/i386.c (ix86_option_override_internal): Mask out
11535 the CF_SET bit when checking -fcf-protection.
11537 2018-02-07 Tom de Vries <tom@codesourcery.com>
11540 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11543 2018-02-07 Richard Biener <rguenther@suse.de>
11545 PR tree-optimization/84204
11546 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11549 PR tree-optimization/84205
11550 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11551 special-case isl_ast_op_zdiv_r.
11553 PR tree-optimization/84223
11554 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11555 Only add conditions from within the region.
11556 (gather_bbs::after_dom_children): Adjust.
11558 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11561 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11562 * config/avr/avr.md: Only post-reload split REG-REG moves if
11563 either register is GENERAL_REG_P.
11565 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11567 PR tree-optimization/84235
11568 * tree-ssa-scopedtables.c
11569 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11570 if the subtraction is performed in floating point type where NaNs are
11571 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11572 build 1. Formatting fix.
11574 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11577 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11578 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11579 and skip it regardless of bb boundaries. Use CALL_P macro,
11580 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11581 unnecessarily, formatting fix.
11583 2018-02-06 Michael Collison <michael.collison@arm.com>
11585 * config/arm/thumb2.md:
11586 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11587 (*thumb_mov_notscc): Ditto.
11589 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11592 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11593 unsigned_fix, not "s".
11595 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11597 * configure.ac (gcc_fn_eh_frame_ro): New function.
11598 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11599 correct .eh_frame permissions.
11600 * configure: Regenerate.
11602 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11604 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11605 irrelevant options.
11607 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11609 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11610 Display warning message for -mno-speculate-indirect-jumps.
11612 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11614 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11616 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11618 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11620 PR tree-optimization/84225
11621 * tree-eh.c (find_trapping_overflow): Only call
11622 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11624 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11627 * config/i386/i386.c: Reimplement the check of possible options
11628 -mibt/-mshstk conbination. Change error messages.
11629 * doc/invoke.texi: Fix a typo: remove extra '='.
11631 2018-02-06 Marek Polacek <polacek@redhat.com>
11633 PR tree-optimization/84228
11634 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11636 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11639 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11640 emitted arch directives.
11641 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11642 __ARM_FEATURE_COPROC before changing architectures.
11644 2018-02-06 Richard Biener <rguenther@suse.de>
11646 * config/i386/i386.c (print_reg): Fix typo.
11647 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11649 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11651 * configure: Regenerate.
11653 2018-02-05 Martin Sebor <msebor@redhat.com>
11655 PR tree-optimization/83369
11656 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11659 2018-02-05 Martin Liska <mliska@suse.cz>
11661 * doc/invoke.texi: Cherry-pick upstream r323995.
11663 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11665 * ira.c (ira_init_register_move_cost): Adjust comment.
11667 2018-02-05 Martin Liska <mliska@suse.cz>
11669 PR gcov-profile/84137
11670 * doc/gcov.texi: Fix typo in documentation.
11672 2018-02-05 Martin Liska <mliska@suse.cz>
11674 PR gcov-profile/83879
11675 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11676 using dlopen functionality.
11678 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11680 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11681 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11682 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11683 _mm_maskz_range_round_ss): New intrinsics.
11684 (__builtin_ia32_rangesd128_round)
11685 (__builtin_ia32_rangess128_round): Remove.
11686 (__builtin_ia32_rangesd128_mask_round,
11687 __builtin_ia32_rangess128_mask_round): New builtins.
11688 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11689 __builtin_ia32_rangess128_round): Remove.
11690 (__builtin_ia32_rangesd128_mask_round,
11691 __builtin_ia32_rangess128_mask_round): New builtins.
11692 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11693 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11694 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11695 "<round_saeonly_constraint>")): Changed to ...
11696 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11697 "<round_saeonly_scalar_constraint>")): ... this.
11698 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11699 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11700 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11701 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11702 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11704 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11706 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11708 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11709 Remove all values except native, 8540 and 8548.
11711 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11713 * config/i386/i386.c (ix86_output_function_return): Pass
11714 INVALID_REGNUM, instead of -1, as invalid register number to
11715 indirect_thunk_name and output_indirect_thunk.
11717 2018-02-02 Julia Koval <julia.koval@intel.com>
11719 * config.gcc: Add -march=icelake.
11720 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11721 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11722 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11723 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11724 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11725 (processor_target_table): Add icelake.
11726 (ix86_option_override_internal): Handle new PTAs.
11727 (get_builtin_code_for_version): Handle icelake.
11728 (M_INTEL_COREI7_ICELAKE): New.
11729 (fold_builtin_cpu): Handle icelake.
11730 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11731 * doc/invoke.texi: Add -march=icelake.
11733 2018-02-02 Julia Koval <julia.koval@intel.com>
11735 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11736 to wide_int_bitmask.
11737 * wide-int-bitmask.h: New.
11739 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11742 * config/i386/i386.md: Replace Pmode with word_mode in
11743 builtin_setjmp_setup and builtin_longjmp to support x32.
11745 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11749 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11751 (rs6000_supported_cpu_names): New static variable.
11752 (linux_cpu_translation_table): Likewise.
11753 (elf_platform) <cpu>: Define new static variable and use it.
11754 Translate kernel AT_PLATFORM name to canonical name if needed.
11755 Error if platform name is unknown.
11757 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11760 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11762 2018-02-01 Jeff Law <law@redhat.com>
11765 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11766 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11767 the scratch if RELEASE_VIA_POP is false.
11768 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11769 If we have to save a temporary register, decrement SIZE appropriately.
11770 Pass new arguments to release_scratch_register_on_entry.
11771 (ix86_adjust_stack_and_probe): Likewise.
11772 (ix86_emit_probe_stack_range): Pass new arguments to
11773 release_scratch_register_on_entry.
11775 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11777 PR rtl-optimization/84157
11778 * combine.c (change_zero_ext): Use REG_P predicate in
11779 front of HARD_REGISTER_P predicate.
11781 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11783 * config/avr/avr.c (avr_option_override): Move disabling of
11784 -fdelete-null-pointer-checks to...
11785 * common/config/avr/avr-common.c (avr_option_optimization_table):
11788 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11790 PR tree-optimization/81635
11791 * tree-data-ref.c (split_constant_offset_1): For types that
11792 wrap on overflow, try to use range info to prove that wrapping
11795 2018-02-01 Renlin Li <renlin.li@arm.com>
11798 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11799 TAILCALL_ADDR_REGS.
11800 (aarch64_register_move_cost): Likewise.
11801 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11802 TAILCALL_ADDR_REGS.
11803 (REG_CLASS_NAMES): Likewise.
11804 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11805 TAILCALL_ADDR_REGS. Remove IP registers.
11806 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11808 2018-02-01 Richard Biener <rguenther@suse.de>
11810 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11811 for specifying RPO order and allow NULL for that.
11812 * domwalk.c (dom_walker::dom_walker): Likewise.
11813 (dom_walker::walk): Handle NULL RPO order.
11814 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11816 (rewrite_update_dom_walker): Likewise.
11817 (mark_def_dom_walker): Likewise.
11819 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11821 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11822 (aarch64_maybe_expand_sve_subreg_move): Declare.
11823 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11824 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11826 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11827 that are semantically a reverse operation.
11828 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11829 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11830 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11832 (aarch64_can_change_mode_class): For big-endian, forbid changes
11833 between two SVE modes if they have different element sizes.
11835 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11837 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11838 the TImode handling for big-endian targets.
11840 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11842 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11843 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11845 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11846 Remove BSWAP handing for big-endian targets and use the form of
11847 LD1RQ appropariate for the mode.
11849 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11851 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11852 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11853 duplicated element.
11855 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11858 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11859 check for operands that need to go through aarch64_sve_reload_be.
11861 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11863 PR tree-optimization/81661
11864 PR tree-optimization/84117
11865 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11866 * tree-eh.c: Include gimplify.h.
11867 (find_trapping_overflow, replace_trapping_overflow,
11868 rewrite_to_non_trapping_overflow): New functions.
11869 * tree-vect-loop.c: Include tree-eh.h.
11870 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11871 * tree-data-ref.c: Include tree-eh.h.
11872 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11874 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11876 PR rtl-optimization/84123
11877 * combine.c (change_zero_ext): Check if hard register satisfies
11878 can_change_dest_mode before calling gen_lowpart_SUBREG.
11880 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11883 * ira.c (ira_init_register_move_cost): Remove assert.
11885 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11887 PR rtl-optimization/84071
11888 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11889 * doc/tm.texi: Regenerate.
11891 2018-01-31 Richard Biener <rguenther@suse.de>
11893 PR tree-optimization/84132
11894 * tree-data-ref.c (analyze_miv_subscript): Properly
11895 check whether evolution_function_is_affine_multivariate_p
11896 before calling gcd_of_steps_may_divide_p.
11898 2018-01-31 Julia Koval <julia.koval@intel.com>
11901 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11902 * config/i386/i386.md (rdpid_rex64) New.
11903 (rdpid): Make 32bit only.
11905 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11908 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11909 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11911 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11914 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11916 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11918 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11920 PR rtl-optimization/84071
11921 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11922 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11924 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11926 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11927 (arc_attribute_table): Add 'aux' attribute.
11928 (arc_in_small_data_p): Consider aux like variables.
11929 (arc_is_aux_reg_p): New function.
11930 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11931 (arc_get_aux_arg): New function.
11932 (prepare_move_operands): Handle aux-register access.
11933 (arc_handle_aux_attribute): New function.
11934 * doc/extend.texi (ARC Variable attributes): Add subsection.
11936 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11938 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11939 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11940 (arc_attribute_table): Add 'uncached' attribute.
11941 (arc_print_operand): Print '.di' flag for uncached memory
11943 (arc_in_small_data_p): Do not consider for small data the uncached
11945 (arc_is_uncached_mem_p): New function.
11946 * config/arc/predicates.md (compact_store_memory_operand): Check
11947 for uncached memory accesses.
11948 (nonvol_nonimm_operand): Likewise.
11949 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11951 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11954 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11955 falign-loops=): Add Optimization flag.
11957 2018-01-30 Jeff Law <law@redhat.com>
11960 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11961 INT_REGISTERS_SAVED. Check it prior to calling
11962 get_scratch_register_on_entry.
11963 (ix86_adjust_stack_and_probe): Similarly.
11964 (ix86_emit_probe_stack_range): Similarly.
11965 (ix86_expand_prologue): Corresponding changes.
11967 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11970 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11971 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11973 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11976 * lra-constraints.c (curr_insn_transform): Process AND in the
11979 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11981 PR rtl-optimization/83986
11982 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11983 dependence against last_pending_memory_flush in addition to
11984 pending_jump_insns.
11986 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11988 PR tree-optimization/81611
11989 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11992 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11995 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11998 2018-01-30 Richard Biener <rguenther@suse.de>
11999 Jakub Jelinek <jakub@redhat.com>
12001 PR tree-optimization/84111
12002 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
12003 inner loops added during recursion, as they don't have up-to-date
12006 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12009 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
12010 (can_inline_edge_by_limits_p): ... here.
12011 (can_early_inline_edge_p, check_callers,
12012 update_caller_keys, update_callee_keys, recursive_inlining,
12013 add_new_edges_to_heap, speculation_useful_p,
12014 inline_small_functions,
12015 inline_small_functions, flatten_function,
12016 inline_to_all_callers_1): Update.
12018 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
12020 * profile-count.c (profile_count::combine_with_ipa_count): Handle
12023 2018-01-30 Richard Biener <rguenther@suse.de>
12025 PR tree-optimization/83008
12026 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
12027 invariant and constant vector uses in stmts when they need
12028 more than one stmt.
12030 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12033 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
12034 * configure: Regenerate.
12036 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12038 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
12040 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
12041 Use gen_rtx_REG rather than gen_lowpart.
12043 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12045 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
12046 rather than 0 when creating partial subregs.
12048 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
12050 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
12053 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
12056 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
12057 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
12058 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
12059 flags. This restores the settings used before the 2017-07-24.
12060 Turning off pre increment/decrement/modify allows IVOPTS to
12061 optimize DF/SF loops where the index is an int.
12063 2018-01-29 Richard Biener <rguenther@suse.de>
12064 Kelvin Nilsen <kelvin@gcc.gnu.org>
12067 * tree-vect-stmts.c (vectorizable_call): Don't call
12068 targetm.vectorize_builtin_md_vectorized_function if callee is
12071 2018-01-22 Carl Love <cel@us.ibm.com>
12073 * doc/extend.tex: Fix typo in second arg in
12074 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
12076 2018-01-29 Richard Biener <rguenther@suse.de>
12078 PR tree-optimization/84086
12079 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
12080 (flush_ssaname_freelist): When SSA names were released reset
12081 the SCEV hash table.
12083 2018-01-29 Richard Biener <rguenther@suse.de>
12085 PR tree-optimization/84057
12086 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
12087 removed paths when removing edges.
12089 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
12091 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
12092 -mfunction-return=@var{choice}.
12094 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
12096 PR diagnostic/84034
12097 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
12098 Handle CR like TAB.
12099 (layout::print_source_line): Likewise.
12100 (test_get_line_width_without_trailing_whitespace): Add test cases.
12102 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12104 PR middle-end/84040
12105 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12108 2018-01-26 Jim Wilson <jimw@sifive.com>
12110 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12112 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12115 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12117 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12118 and CMP + SUB-immediate -> SUBS.
12120 2018-01-26 Martin Sebor <msebor@redhat.com>
12122 PR tree-optimization/83896
12123 * tree-ssa-strlen.c (get_string_len): Rename...
12124 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12125 Avoid assuming length is constant.
12126 (handle_char_store): Use HOST_WIDE_INT for string length.
12128 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12131 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12132 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12134 2018-01-26 Richard Biener <rguenther@suse.de>
12136 PR rtl-optimization/84003
12137 * dse.c (record_store): Only record redundant stores when
12138 the earlier store aliases at least all accesses the later one does.
12140 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12142 PR rtl-optimization/83985
12143 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12144 REG_CFA_RESTORE insns.
12145 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12146 REG_CFA_RESTORE insns here.
12149 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12150 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12152 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12154 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12155 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12156 (arc_init): Likewise.
12157 (arc_override_options): Likewise.
12158 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12160 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12162 * config/arc/arc.h (TARGET_DBNZ): Define.
12163 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12164 properly set the tune attribute.
12165 (dbnz): Use TARGET_DBNZ guard.
12166 * config/arc/arc.opt (mtune): Add core3 option.
12168 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12170 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12171 recognize new pic like addresses.
12172 (arc_delegitimize_address): Clean up.
12174 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12176 * config/arc/arc-arches.def: Option mrf16 valid for all
12178 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12179 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12180 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12181 * config/arc/arc-tables.opt: Regenerate.
12182 * config/arc/arc.c (arc_conditional_register_usage): Handle
12183 reduced register file case.
12184 (arc_file_start): Set must have build attributes.
12185 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12186 mrf16 option value.
12187 * config/arc/arc.opt (mrf16): Add new option.
12188 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12189 * config/arc/genmultilib.awk: Handle new mrf16 option.
12190 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12191 * config/arc/t-multilib: Regenerate.
12192 * doc/invoke.texi (ARC Options): Document mrf16 option.
12194 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12196 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12197 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12198 (arc_attribute_table): Add 'secure_call' attribute.
12199 (arc_print_operand): Print secure call operand.
12200 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12202 (arc_is_secure_call_p): New function. * config/arc/arc.md
12203 (call_i): Add support for sjli instruction.
12204 (call_value_i): Likewise.
12205 * config/arc/constraints.md (Csc): New constraint.
12207 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12208 John Eric Martin <John.Martin@emmicro-us.com>
12210 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12211 * config/arc/arc.c (_arc_jli_section): New struct.
12212 (arc_jli_section): New type.
12213 (rc_jli_sections): New static variable.
12214 (arc_handle_jli_attribute): New function.
12215 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12216 (arc_file_end): New function.
12217 (TARGET_ASM_FILE_END): Define.
12218 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12219 (arc_add_jli_section): New function.
12220 (jli_call_scan): Likewise.
12221 (arc_reorg): Call jli_call_scan.
12222 (arc_output_addsi): Remove 'S' from printing asm operand.
12223 (arc_is_jli_call_p): New function.
12224 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12226 (movhi_insn): Likewise.
12227 (movsi_insn): Likewise.
12228 (movsi_set_cc_insn): Likewise.
12229 (loadqi_update): Likewise.
12230 (load_zeroextendqisi_update): Likewise.
12231 (load_signextendqisi_update): Likewise.
12232 (loadhi_update): Likewise.
12233 (load_zeroextendhisi_update): Likewise.
12234 (load_signextendhisi_update): Likewise.
12235 (loadsi_update): Likewise.
12236 (loadsf_update): Likewise.
12237 (movsicc_insn): Likewise.
12238 (bset_insn): Likewise.
12239 (bxor_insn): Likewise.
12240 (bclr_insn): Likewise.
12241 (bmsk_insn): Likewise.
12242 (bicsi3_insn): Likewise.
12243 (cmpsi_cc_c_insn): Likewise.
12244 (movsi_ne): Likewise.
12245 (movsi_cond_exec): Likewise.
12246 (clrsbsi2): Likewise.
12247 (norm_f): Likewise.
12257 (call_i): Remove 'S' asm letter, add jli instruction.
12258 (call_value_i): Likewise.
12259 * config/arc/arc.op (mjli-always): New option.
12260 * config/arc/constraints.md (Cji): New constraint.
12261 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12263 (subsf3_fpx): Likewise.
12264 (mulsf3_fpx): Likewise.
12265 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12267 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12268 function attrbutes.
12269 * doc/invoke.texi (ARC): Document mjli-always option.
12271 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12273 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12274 avoid addition with 0 and use incw and decw where possible.
12276 2018-01-26 Richard Biener <rguenther@suse.de>
12278 PR tree-optimization/81082
12279 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12280 association if it requires casting to unsigned.
12281 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12282 from fold_plusminus_mult_expr to catch important cases late when
12283 range info is available.
12285 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12287 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12288 * configure.ac (hidden_linkonce): New test.
12289 * configure: Regenerate.
12290 * config.in: Regenerate.
12292 2018-01-26 Julia Koval <julia.koval@intel.com>
12294 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12295 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12296 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12297 _mm_mask_bitshuffle_epi64_mask): Fix type.
12298 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12299 USI_FTYPE_V4DI_V4DI_USI): Remove.
12300 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12301 __builtin_ia32_vpshufbitqmb256_mask,
12302 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12303 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12304 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12306 2018-01-26 Alan Modra <amodra@gmail.com>
12309 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12310 UNSPEC_VBPERMQ. Sort other unspecs.
12312 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12314 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12316 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12318 PR middle-end/83055
12319 * predict.c (drop_profile): Do not push/pop cfun; update also
12321 (handle_missing_profiles): Fix logic looking for zero profiles.
12323 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12325 PR middle-end/83977
12326 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12327 on functions with #pragma omp declare simd or functions with simd
12329 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12330 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12331 Remove trailing \n from warning_at calls.
12333 2018-01-25 Tom de Vries <tom@codesourcery.com>
12336 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12337 for neutered workers.
12339 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12342 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12343 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12345 2018-01-24 Jeff Law <law@redhat.com>
12348 * i386.c (get_probe_interval): Move to earlier point.
12349 (ix86_compute_frame_layout): If -fstack-clash-protection and
12350 the frame is larger than the probe interval, then use pushes
12351 to save registers rather than reg->mem moves.
12352 (ix86_expand_prologue): Remove conditional for int_registers_saved
12355 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12358 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12359 min/max for never referenced object.
12361 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12363 PR middle-end/83977
12364 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12366 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12367 attributes from DECL_ATTRIBUTES (decl) without affecting
12368 DECL_ATTRIBUTES (current_function_decl).
12369 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12370 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12372 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12374 PR tree-optimization/83979
12375 * fold-const.c (fold_comparison): Use constant_boolean_node
12376 instead of boolean_{true,false}_node.
12378 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12380 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12383 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12385 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12386 Simplify the clause that sets the length attribute.
12387 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12388 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12389 clause that sets the length attribute.
12390 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12392 2018-01-24 Tom de Vries <tom@codesourcery.com>
12395 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12396 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12397 Add strict parameter.
12398 (prevent_branch_around_nothing): Insert dummy insn between branch to
12399 label and label with no ptx insn inbetween.
12400 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12402 2018-01-24 Tom de Vries <tom@codesourcery.com>
12405 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12406 for neutered threads in warp.
12407 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12409 2018-01-24 Richard Biener <rguenther@suse.de>
12411 PR tree-optimization/83176
12412 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12415 2018-01-24 Richard Biener <rguenther@suse.de>
12417 PR tree-optimization/82819
12418 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12419 code generating pluses that are no-ops in the target precision.
12421 2018-01-24 Richard Biener <rguenther@suse.de>
12423 PR middle-end/84000
12424 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12426 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12428 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12429 to merge probabilities.
12430 * predict.c (probably_never_executed): Also mark as cold functions
12431 with global 0 profile and guessed local profile.
12432 * profile-count.c (profile_probability::combine_with_count): New
12434 * profile-count.h (profile_probability::operator*,
12435 profile_probability::operator*=, profile_probability::operator/,
12436 profile_probability::operator/=): Reduce precision to adjusted
12437 and set value to guessed on contradictory divisions.
12438 (profile_probability::combine_with_freq): Remove.
12439 (profile_probability::combine_wiht_count): Declare.
12440 (profile_count::force_nonzero):: Set to adjusted.
12441 (profile_count::probability_in):: Set quality to adjusted.
12442 * tree-ssa-tail-merge.c (replace_block_by): Use
12443 combine_with_count.
12445 2018-01-23 Andrew Waterman <andrew@sifive.com>
12446 Jim Wilson <jimw@sifive.com>
12448 * config/riscv/riscv.c (riscv_stack_boundary): New.
12449 (riscv_option_override): Set riscv_stack_boundary. Handle
12450 riscv_preferred_stack_boundary_arg.
12451 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12452 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12453 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12454 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12455 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12456 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12458 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12461 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12462 of struct ix86_frame.
12463 (ix86_expand_epilogue): Likewise. Add a local variable for
12464 the reg_save_offset field in struct ix86_frame.
12466 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12468 PR tree-optimization/82604
12469 * tree-loop-distribution.c (enum partition_kind): New enum item
12470 PKIND_PARTIAL_MEMSET.
12471 (partition_builtin_p): Support above new enum item.
12472 (generate_code_for_partition): Ditto.
12473 (compute_access_range): Differentiate cases that equality can be
12474 proven at all loops, the innermost loops or no loops.
12475 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12476 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12477 (finalize_partitions, distribute_loop): Don't fuse partition of
12478 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12479 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12480 parloop is enabled.
12482 2018-01-23 Martin Liska <mliska@suse.cz>
12484 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12485 order to ignore the predictor.
12486 (PRED_POLYMORPHIC_CALL): Likewise.
12487 (PRED_RECURSIVE_CALL): Likewise.
12489 2018-01-23 Martin Liska <mliska@suse.cz>
12491 * tree-profile.c (tree_profiling): Print function header to
12492 aware reader which function we are working on.
12493 * value-prof.c (gimple_find_values_to_profile): Do not print
12494 not interesting value histograms.
12496 2018-01-23 Martin Liska <mliska@suse.cz>
12498 * profile-count.h (enum profile_quality): Add
12499 profile_uninitialized as the first value. Do not number values
12500 as they are zero based.
12501 (profile_count::verify): Update sanity check.
12502 (profile_probability::verify): Likewise.
12504 2018-01-23 Nathan Sidwell <nathan@acm.org>
12506 * doc/invoke.texi (ffor-scope): Deprecate.
12508 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12510 PR tree-optimization/83510
12511 * domwalk.c (set_all_edges_as_executable): New function.
12512 (dom_walker::dom_walker): Convert bool param
12513 "skip_unreachable_blocks" to enum reachability. Move setup of
12514 edge flags to set_all_edges_as_executable and only do it when
12515 reachability is REACHABLE_BLOCKS.
12516 * domwalk.h (enum dom_walker::reachability): New enum.
12517 (dom_walker::dom_walker): Convert bool param
12518 "skip_unreachable_blocks" to enum reachability.
12519 (set_all_edges_as_executable): New decl.
12520 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12521 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12523 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12524 but converting true to REACHABLE_BLOCKS.
12525 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12527 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12528 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12529 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12531 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12532 if check_all_array_refs will be called.
12534 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12536 * tree.c (selftest::test_location_wrappers): Add more test
12539 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12541 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12542 (selftest::test_bit_in_range): Likewise.
12544 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12547 * doc/sourcebuild.texi (vect_float): Say that the selector
12548 only describes the situation when -funsafe-math-optimizations is on.
12549 (vect_float_strict): Document.
12551 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12553 PR tree-optimization/83965
12554 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12555 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12556 instead of checking only for a reduction.
12557 (vect_recog_widen_sum_pattern): Likewise.
12559 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12561 * predict.c (probably_never_executed): Only use precise profile info.
12562 (compute_function_frequency): Skip after inlining hack since we now
12563 have quality checking.
12565 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12567 * profile-count.h (profile_probability::very_unlikely,
12568 profile_probability::unlikely, profile_probability::even): Set
12569 precision to guessed.
12571 2018-01-23 Richard Biener <rguenther@suse.de>
12573 PR tree-optimization/83963
12574 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12575 Properly terminate dominator walk when crossing the exit edge not
12576 when visiting its source block.
12578 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12581 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12582 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12584 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12586 PR tree-optimization/83957
12587 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12588 semicolon after for body surrounded by braces.
12590 PR tree-optimization/83081
12591 * profile-count.h (profile_probability::split): New method.
12592 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12593 Use profile_probability::split.
12594 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12595 when splitting a single conditional jump into 2.
12597 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12599 PR tree-optimization/69452
12600 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12603 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12605 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12606 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12607 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12609 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12611 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12613 * config/rl78/rl78.md (movdi): New define_expand.
12614 * config/rl78/rl78.c (rl78_split_movdi): New function.
12616 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12619 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12621 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12622 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12623 128-bit to produce an UNSPEC move to get the double word with the
12624 signbit and then a shift directly to do signbit.
12625 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12626 implementation with a new version that just does either a direct
12627 move or a regular move. Move memory interface to separate insns.
12628 Move insns so they are next to the expander.
12629 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12630 with signbit move. Split big and little endian case.
12631 (signbit<mode>2_dm_mem_le): Likewise.
12632 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12633 (signbit<mode>2_dm2): Likewise.
12635 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12637 * config/rl78/rl78.md (anddi3): New define_expand.
12639 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12641 * config/rl78/rl78.md (umindi3): New define_expand.
12643 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12645 * config/rl78/rl78.md (smindi3): New define_expand.
12647 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12649 * config/rl78/rl78.md (smaxdi3): New define_expand.
12651 2018-01-22 Carl Love <cel@us.ibm.com>
12653 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12654 LVX_V1TI): Add macro expansion.
12655 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12656 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12657 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12658 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12659 Change check to determine if the instruction is a byte reversing
12660 entry. Fix typo in comment.
12661 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12662 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12663 Add def_builtin calls for new builtins.
12664 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12665 Add define_insn expansion.
12667 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12669 * config/rl78/rl78.md (umaxdi3): New define_expand.
12671 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12673 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12674 for non-QImode registers.
12676 2018-01-22 Richard Biener <rguenther@suse.de>
12678 PR tree-optimization/83963
12679 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12680 including the loop exit block.
12681 (scop_detection::merge_sese): Likewise.
12682 (scop_detection::add_scop): Do it here instead.
12684 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12686 * doc/sourcebuild.texi (arm_softfloat): Document.
12688 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12691 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12692 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12693 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12695 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12696 David Edelsohn <dje.gcc@gmail.com>
12699 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12700 Change "crset eq" to "crset 2".
12701 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12702 (*call_indirect_aix<mode>_nospec): Likewise.
12703 (*call_value_indirect_aix<mode>_nospec): Likewise.
12704 (*call_indirect_elfv2<mode>_nospec): Likewise.
12705 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12706 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12707 change assembly output from . to $.
12708 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12709 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12710 (*tablejump<mode>_internal1_nospec): Likewise.
12712 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12715 * config/sh/sh_optimize_sett_clrt.cc:
12716 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12718 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12720 PR tree-optimization/83940
12721 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12722 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12723 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12724 use it to pass back the definition type.
12725 (vect_check_store_rhs): Likewise rhs_dt_out.
12726 (vect_build_gather_load_calls): Add a mask_dt argument and use
12727 it instead of a call to vect_is_simple_use.
12728 (vectorizable_store): Update calls to vect_check_load_store_mask
12729 and vect_check_store_rhs. Use the dt returned by the latter instead
12730 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12731 instead of calls to vect_is_simple_use. Pass the scalar rather
12732 than the vector operand to vect_is_simple_use when handling
12733 second and subsequent copies of an rhs value.
12734 (vectorizable_load): Update calls to vect_check_load_store_mask
12735 and vect_build_gather_load_calls. Use the cached mask_dt and
12736 gs_info.offset_dt instead of calls to vect_is_simple_use.
12738 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12740 PR middle-end/83945
12741 * tree-emutls.c: Include gimplify.h.
12742 (lower_emutls_2): New function.
12743 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12744 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12745 it before further processing.
12748 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12749 UINTVAL (trueop1) instead of INTVAL (op1).
12751 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12755 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12756 INCOMING_FRAME_SP_OFFSET if not defined.
12757 (scan_trace): Add ENTRY argument. If true and
12758 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12759 emit a note to adjust the CFA offset.
12760 (create_cfi_notes): Adjust scan_trace callers.
12761 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12762 INCOMING_FRAME_SP_OFFSET in the CIE.
12763 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12764 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12766 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12767 * doc/tm.texi: Regenerated.
12769 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12771 PR rtl-optimization/83147
12772 * lra-constraints.c (remove_inheritance_pseudos): Use
12773 lra_substitute_pseudo_within_insn.
12775 2018-01-19 Tom de Vries <tom@codesourcery.com>
12776 Cesar Philippidis <cesar@codesourcery.com>
12779 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12781 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12784 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12785 spaces for function labels.
12787 2018-01-19 Martin Liska <mliska@suse.cz>
12789 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12790 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12791 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12792 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12793 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12794 (PRED_CONST_RETURN): Change from 69 to 65.
12795 (PRED_NULL_RETURN): Change from 91 to 71.
12796 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12797 (PRED_LOOP_GUARD): Change from 66 to 73.
12799 2018-01-19 Martin Liska <mliska@suse.cz>
12801 * predict.c (predict_insn_def): Add new assert.
12802 (struct branch_predictor): Change type to signed integer.
12803 (test_prediction_value_range): Amend test to cover
12804 PROB_UNINITIALIZED.
12805 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12806 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12807 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12808 (PRED_LOOP_IV_COMPARE): Likewise.
12809 * predict.h (PROB_UNINITIALIZED): Define new constant.
12811 2018-01-19 Martin Liska <mliska@suse.cz>
12813 * predict.c (dump_prediction): Add new format for
12814 analyze_brprob.py script which is enabled with -details
12816 * profile-count.h (precise_p): New function.
12818 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12820 PR tree-optimization/83922
12821 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12822 there are no statements that need masking.
12823 (vect_active_double_reduction_p): New function.
12824 (vect_analyze_loop_operations): Use it when handling phis that
12825 are not in the loop header.
12827 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12829 PR tree-optimization/83914
12830 * tree-vect-loop.c (vectorizable_induction): Don't convert
12831 init_expr or apply the peeling adjustment for inductions
12832 that are nested within the vectorized loop.
12834 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12836 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12839 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12843 * function.h (gimplify_parameters): Add gimple_seq * argument.
12844 * function.c: Include gimple.h and options.h.
12845 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12846 for the added local temporaries if needed.
12847 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12848 if there are any parameter cleanups, wrap whole body into a
12849 try/finally with the cleanups.
12851 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12854 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12855 Use GET_MODE_CLASS for scalar floating point.
12857 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12861 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12862 Fix call of call_cgraph_insertion_hooks.
12864 2018-01-18 Martin Sebor <msebor@redhat.com>
12866 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12868 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12871 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12874 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12877 * common.opt: (-ffile-prefix-map): New option.
12878 * opts.c (common_handle_option): Defer it.
12879 * opts-global.c (handle_common_deferred_options): Handle it.
12880 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12881 * file-prefix-map.h: New file.
12882 (remap_debug_filename, add_debug_prefix_map): ...here.
12883 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12884 * final.c (debug_prefix_map, add_debug_prefix_map
12885 remap_debug_filename): Move to...
12886 * file-prefix-map.c: New file.
12887 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12888 generalize, get rid of alloca(), use strrchr() instead of strchr().
12889 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12890 Implement in terms of add_prefix_map().
12891 (remap_macro_filename, remap_debug_filename): Implement in term of
12893 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12894 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12895 * dbxout.c: Include file-prefix-map.h.
12896 * varasm.c: Likewise.
12897 * vmsdbgout.c: Likewise.
12898 * xcoffout.c: Likewise.
12899 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12900 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12901 * doc/invoke.texi (-ffile-prefix-map): Document.
12902 (-fdebug-prefix-map): Update description.
12904 2018-01-18 Martin Liska <mliska@suse.cz>
12906 * config/i386/i386.c (indirect_thunk_name): Document that also
12908 (output_indirect_thunk): Document why both instructions
12909 (pause and lfence) are generated.
12911 2018-01-18 Richard Biener <rguenther@suse.de>
12913 PR tree-optimization/83887
12914 * graphite-scop-detection.c
12915 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12916 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12917 (scop_detection::merge_sese): Re-implement with a flood-fill
12918 algorithm that properly finds a SESE region if it exists.
12920 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12923 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12924 pointer_diff optimizations use view_convert instead of convert.
12926 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12928 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12929 Generate different code for -mno-speculate-indirect-jumps.
12930 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12931 (*call_indirect_aix<mode>): Disable for
12932 -mno-speculate-indirect-jumps.
12933 (*call_indirect_aix<mode>_nospec): New define_insn.
12934 (*call_value_indirect_aix<mode>): Disable for
12935 -mno-speculate-indirect-jumps.
12936 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12937 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12938 -mno-speculate-indirect-jumps.
12939 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12941 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12943 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12944 long double type, set the flags for noting the default long double
12945 type, even if we don't pass or return a long double type.
12947 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12950 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12953 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12955 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12956 support for merge[hl].
12957 (fold_mergehl_helper): New helper function.
12958 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12959 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12960 (altivec_vmrglw_direct): Add xxmrglw insn.
12962 2018-01-17 Andrew Waterman <andrew@sifive.com>
12964 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12965 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12967 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12970 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12971 call the lto_location_cache before reading the
12972 DECL_SOURCE_LOCATION of the types.
12974 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12975 Richard Sandiford <richard.sandiford@linaro.org>
12977 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12978 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12979 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12980 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12981 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12983 * config/aarch64/constraints.md (aarch64_movti_operand):
12985 * config/aarch64/predicates.md (Uti): Add new constraint.
12987 2018-01-17 Carl Love <cel@us.ibm.com>
12989 * config/rs6000/vsx.md (define_expand xl_len_r,
12990 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12991 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12993 (define_expand, define_insn): Move the shift left from the
12994 define_insn to the define_expand for lxvl and stxvl instructions.
12995 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12996 and XL_LEN_R definitions to PURE.
12998 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
13000 * config/i386/i386.c (indirect_thunk_name): Declare regno
13001 as unsigned int. Compare regno with INVALID_REGNUM.
13002 (output_indirect_thunk): Ditto.
13003 (output_indirect_thunk_function): Ditto.
13004 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
13005 in the call to output_indirect_thunk_function.
13007 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
13009 PR middle-end/83884
13010 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
13011 rather than the size of inner_type to determine the stack slot size
13012 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
13014 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
13017 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
13020 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
13022 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
13023 endian Linux systems to optionally enable multilibs for selecting
13024 the long double type if the user configured an explicit type.
13025 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
13026 have no long double multilibs if not defined.
13027 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
13028 warn if the user used -mabi={ieee,ibm}longdouble and we built
13029 multilibs for long double.
13030 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
13031 appropriate multilib option.
13032 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
13034 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
13035 for building long double multilibs.
13036 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
13038 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
13040 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
13043 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
13045 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
13048 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
13051 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
13054 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
13057 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
13059 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
13061 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
13062 different rtl trees depending on TARGET_64BIT.
13063 (rs6000_gen_lvx): Likewise.
13065 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
13067 * config/visium/visium.md (nop): Tweak comment.
13068 (hazard_nop): Likewise.
13070 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13072 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
13073 -mspeculate-indirect-jumps.
13074 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
13075 for -mno-speculate-indirect-jumps.
13076 (*call_indirect_elfv2<mode>_nospec): New define_insn.
13077 (*call_value_indirect_elfv2<mode>): Disable for
13078 -mno-speculate-indirect-jumps.
13079 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
13080 (indirect_jump): Emit different RTL for
13081 -mno-speculate-indirect-jumps.
13082 (*indirect_jump<mode>): Disable for
13083 -mno-speculate-indirect-jumps.
13084 (*indirect_jump<mode>_nospec): New define_insn.
13085 (tablejump): Emit different RTL for
13086 -mno-speculate-indirect-jumps.
13087 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
13088 (tablejumpsi_nospec): New define_expand.
13089 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
13090 (tablejumpdi_nospec): New define_expand.
13091 (*tablejump<mode>_internal1): Disable for
13092 -mno-speculate-indirect-jumps.
13093 (*tablejump<mode>_internal1_nospec): New define_insn.
13094 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
13097 2018-01-16 Artyom Skrobov tyomitch@gmail.com
13099 * caller-save.c (insert_save): Drop unnecessary parameter. All
13102 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13103 Richard Biener <rguenth@suse.de>
13106 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13107 return early, inline manually is_gimple_sizepos. Make sure if we
13108 call gimplify_expr we don't end up with a gimple constant.
13109 * tree.c (variably_modified_type_p): Don't return true for
13110 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13111 * gimplify.h (is_gimple_sizepos): Remove.
13113 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13115 PR tree-optimization/83857
13116 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13117 vectorizable_live_operation for pure SLP statements.
13118 (vectorizable_live_operation): Handle PHIs.
13120 2018-01-16 Richard Biener <rguenther@suse.de>
13122 PR tree-optimization/83867
13123 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13124 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13126 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13129 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13130 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13131 If off is not INTEGER_CST, issue a may not be aligned warning
13132 rather than isn't aligned. Use isn%'t rather than isn't.
13133 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13135 <case MULT_EXPR>: Improve the case when bottom and one of the
13136 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13137 operand, in that case check if the other operand is multiple of
13138 bottom divided by the INTEGER_CST operand.
13140 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13143 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13144 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13145 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13146 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13147 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13148 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13149 (pa_function_arg_size): New function.
13151 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13153 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13154 in a separate statement.
13156 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13158 PR tree-optimization/83847
13159 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13160 group gathers and scatters.
13162 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13164 PR rtl-optimization/86620
13165 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13167 PR rtl-optimization/83213
13168 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13169 to last if both are JUMP_INSNs.
13171 PR tree-optimization/83843
13172 * gimple-ssa-store-merging.c
13173 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13174 store_immediate_info for bswap/nop orig_stores.
13176 2018-01-15 Andrew Waterman <andrew@sifive.com>
13178 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13180 <UDIV>: Increase cost if !TARGET_DIV.
13182 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13184 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13185 (define_attr "cr_logical_3op"): New.
13186 (cceq_ior_compare): Adjust.
13187 (cceq_ior_compare_complement): Adjust.
13188 (*cceq_rev_compare): Adjust.
13189 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13190 (is_cracked_insn): Adjust.
13191 (insn_must_be_first_in_group): Adjust.
13192 * config/rs6000/40x.md: Adjust.
13193 * config/rs6000/440.md: Adjust.
13194 * config/rs6000/476.md: Adjust.
13195 * config/rs6000/601.md: Adjust.
13196 * config/rs6000/603.md: Adjust.
13197 * config/rs6000/6xx.md: Adjust.
13198 * config/rs6000/7450.md: Adjust.
13199 * config/rs6000/7xx.md: Adjust.
13200 * config/rs6000/8540.md: Adjust.
13201 * config/rs6000/cell.md: Adjust.
13202 * config/rs6000/e300c2c3.md: Adjust.
13203 * config/rs6000/e500mc.md: Adjust.
13204 * config/rs6000/e500mc64.md: Adjust.
13205 * config/rs6000/e5500.md: Adjust.
13206 * config/rs6000/e6500.md: Adjust.
13207 * config/rs6000/mpc.md: Adjust.
13208 * config/rs6000/power4.md: Adjust.
13209 * config/rs6000/power5.md: Adjust.
13210 * config/rs6000/power6.md: Adjust.
13211 * config/rs6000/power7.md: Adjust.
13212 * config/rs6000/power8.md: Adjust.
13213 * config/rs6000/power9.md: Adjust.
13214 * config/rs6000/rs64.md: Adjust.
13215 * config/rs6000/titan.md: Adjust.
13217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13219 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13220 ix86_indirect_branch_register logic.
13222 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13224 * config/i386/constraints.md (Bs): Update
13225 ix86_indirect_branch_register check. Don't check
13226 ix86_indirect_branch_register with GOT_memory_operand.
13228 * config/i386/predicates.md (GOT_memory_operand): Don't check
13229 ix86_indirect_branch_register here.
13230 (GOT32_symbol_operand): Likewise.
13232 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13234 * config/i386/predicates.md (constant_call_address_operand):
13235 Rewrite ix86_indirect_branch_register logic.
13236 (sibcall_insn_operand): Likewise.
13238 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13240 * config/i386/constraints.md (Bs): Replace
13241 ix86_indirect_branch_thunk_register with
13242 ix86_indirect_branch_register.
13244 * config/i386/i386.md (indirect_jump): Likewise.
13245 (tablejump): Likewise.
13246 (*sibcall_memory): Likewise.
13247 (*sibcall_value_memory): Likewise.
13248 Peepholes of indirect call and jump via memory: Likewise.
13249 * config/i386/i386.opt: Likewise.
13250 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13251 (GOT_memory_operand): Likewise.
13252 (call_insn_operand): Likewise.
13253 (sibcall_insn_operand): Likewise.
13254 (GOT32_symbol_operand): Likewise.
13256 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13258 PR middle-end/83837
13259 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13260 type rather than type addr's type points to.
13261 (expand_omp_atomic_mutex): Likewise.
13262 (expand_omp_atomic): Likewise.
13264 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13267 * config/i386/i386.c (output_indirect_thunk_function): Use
13268 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13269 for __x86_return_thunk.
13271 2018-01-15 Richard Biener <rguenther@suse.de>
13273 PR middle-end/83850
13274 * expmed.c (extract_bit_field_1): Fix typo.
13276 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13279 * config/arm/iterators.md (VF): New mode iterator.
13280 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13281 Remove integer-related logic from pattern.
13282 (neon_vabd<mode>_3): Likewise.
13284 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13286 PR middle-end/82694
13287 * common.opt (fstrict-overflow): No longer an alias.
13288 (fwrapv-pointer): New option.
13289 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13290 also for pointer types based on flag_wrapv_pointer.
13291 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13292 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13293 opts->x_flag_wrapv got set.
13294 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13295 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13296 POINTER_TYPE_OVERFLOW_UNDEFINED.
13297 * match.pd: Likewise in address comparison pattern.
13298 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13300 2018-01-15 Richard Biener <rguenther@suse.de>
13303 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13304 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13305 Reset type names to their identifier if their TYPE_DECL doesn't
13306 have linkage (and thus is used for ODR and devirt).
13307 (save_debug_info_for_decl): Remove.
13308 (save_debug_info_for_type): Likewise.
13309 (add_tree_to_fld_list): Adjust.
13310 * tree-pretty-print.c (dump_generic_node): Make dumping of
13311 type names more robust.
13313 2018-01-15 Richard Biener <rguenther@suse.de>
13315 * BASE-VER: Bump to 8.0.1.
13317 2018-01-14 Martin Sebor <msebor@redhat.com>
13320 * builtins.c (check_access): Avoid warning when the no-warning bit
13323 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13325 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13326 * ira-color (allocno_hard_regs_compare): Likewise.
13328 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13331 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13332 Use .pushsection/.popsection.
13334 2018-01-14 Martin Sebor <msebor@redhat.com>
13337 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13339 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13341 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13342 entry from extra_headers.
13343 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13344 extra_headers, make the list bitwise identical to the i?86-*-* one.
13346 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13348 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13349 -mcmodel=large with -mindirect-branch=thunk,
13350 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13351 -mfunction-return=thunk-extern.
13352 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13353 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13354 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13356 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13358 * config/i386/i386.c (print_reg): Print the name of the full
13359 integer register without '%'.
13360 (ix86_print_operand): Handle 'V'.
13361 * doc/extend.texi: Document 'V' modifier.
13363 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13365 * config/i386/constraints.md (Bs): Disallow memory operand for
13366 -mindirect-branch-register.
13368 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13369 (GOT_memory_operand): Likewise.
13370 (call_insn_operand): Likewise.
13371 (sibcall_insn_operand): Likewise.
13372 (GOT32_symbol_operand): Likewise.
13373 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13374 for -mindirect-branch-register.
13375 (tablejump): Likewise.
13376 (*sibcall_memory): Likewise.
13377 (*sibcall_value_memory): Likewise.
13378 Disallow peepholes of indirect call and jump via memory for
13379 -mindirect-branch-register.
13380 (*call_pop): Replace m with Bw.
13381 (*call_value_pop): Likewise.
13382 (*sibcall_pop_memory): Replace m with Bs.
13383 * config/i386/i386.opt (mindirect-branch-register): New option.
13384 * doc/invoke.texi: Document -mindirect-branch-register option.
13386 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13388 * config/i386/i386-protos.h (ix86_output_function_return): New.
13389 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13390 set function_return_type.
13391 (indirect_thunk_name): Add ret_p to indicate thunk for function
13393 (output_indirect_thunk_function): Pass false to
13394 indirect_thunk_name.
13395 (ix86_output_indirect_branch_via_reg): Likewise.
13396 (ix86_output_indirect_branch_via_push): Likewise.
13397 (output_indirect_thunk_function): Create alias for function
13398 return thunk if regno < 0.
13399 (ix86_output_function_return): New function.
13400 (ix86_handle_fndecl_attribute): Handle function_return.
13401 (ix86_attribute_table): Add function_return.
13402 * config/i386/i386.h (machine_function): Add
13403 function_return_type.
13404 * config/i386/i386.md (simple_return_internal): Use
13405 ix86_output_function_return.
13406 (simple_return_internal_long): Likewise.
13407 * config/i386/i386.opt (mfunction-return=): New option.
13408 (indirect_branch): Mention -mfunction-return=.
13409 * doc/extend.texi: Document function_return function attribute.
13410 * doc/invoke.texi: Document -mfunction-return= option.
13412 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13414 * config/i386/i386-opts.h (indirect_branch): New.
13415 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13416 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13417 with local indirect jump when converting indirect call and jump.
13418 (ix86_set_indirect_branch_type): New.
13419 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13420 (indirectlabelno): New.
13421 (indirect_thunk_needed): Likewise.
13422 (indirect_thunk_bnd_needed): Likewise.
13423 (indirect_thunks_used): Likewise.
13424 (indirect_thunks_bnd_used): Likewise.
13425 (INDIRECT_LABEL): Likewise.
13426 (indirect_thunk_name): Likewise.
13427 (output_indirect_thunk): Likewise.
13428 (output_indirect_thunk_function): Likewise.
13429 (ix86_output_indirect_branch_via_reg): Likewise.
13430 (ix86_output_indirect_branch_via_push): Likewise.
13431 (ix86_output_indirect_branch): Likewise.
13432 (ix86_output_indirect_jmp): Likewise.
13433 (ix86_code_end): Call output_indirect_thunk_function if needed.
13434 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13436 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13437 (ix86_attribute_table): Add indirect_branch.
13438 * config/i386/i386.h (machine_function): Add indirect_branch_type
13439 and has_local_indirect_jump.
13440 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13442 (tablejump): Likewise.
13443 (*indirect_jump): Use ix86_output_indirect_jmp.
13444 (*tablejump_1): Likewise.
13445 (simple_return_indirect_internal): Likewise.
13446 * config/i386/i386.opt (mindirect-branch=): New option.
13447 (indirect_branch): New.
13450 (thunk-inline): Likewise.
13451 (thunk-extern): Likewise.
13452 * doc/extend.texi: Document indirect_branch function attribute.
13453 * doc/invoke.texi: Document -mindirect-branch= option.
13455 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13458 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13460 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13462 * ipa-inline.c (want_inline_small_function_p): Return false if
13463 inlining has already failed with CIF_FINAL_ERROR.
13464 (update_caller_keys): Call want_inline_small_function_p before
13466 (update_callee_keys): Likewise.
13468 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13470 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13472 (rs6000_quadword_masked_address_p): Likewise.
13473 (quad_aligned_load_p): Likewise.
13474 (quad_aligned_store_p): Likewise.
13475 (const_load_sequence_p): Add comment to describe the outer-most loop.
13476 (mimic_memory_attributes_and_flags): New function.
13477 (rs6000_gen_stvx): Likewise.
13478 (replace_swapped_aligned_store): Likewise.
13479 (rs6000_gen_lvx): Likewise.
13480 (replace_swapped_aligned_load): Likewise.
13481 (replace_swapped_load_constant): Capitalize argument name in
13482 comment describing this function.
13483 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13484 and stores that access quad-word aligned addresses and replace
13485 with stvx or lvx instructions when appropriate.
13486 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13487 New function prototype.
13488 (rs6000_quadword_masked_address_p): Likewise.
13489 (rs6000_gen_lvx): Likewise.
13490 (rs6000_gen_stvx): Likewise.
13491 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13492 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13493 when memory address is aligned.
13494 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13495 this split to select lvx instruction when memory address is aligned.
13496 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13497 instruction when memory address is aligned.
13498 (*vsx_le_perm_load_v16qi): Likewise.
13499 (four unnamed splitters): Modify to select the stvx instruction
13500 when memory is aligned.
13502 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13504 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13505 which appears in the queue multiple times.
13507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13508 Alan Hayward <alan.hayward@arm.com>
13509 David Sherwood <david.sherwood@arm.com>
13511 * tree-vectorizer.h (vec_lower_bound): New structure.
13512 (_loop_vec_info): Add check_nonzero and lower_bounds.
13513 (LOOP_VINFO_CHECK_NONZERO): New macro.
13514 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13515 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13516 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13517 fields. Make seg_len the distance travelled, not including the
13519 (dr_direction_indicator): Declare.
13520 (dr_zero_step_indicator): Likewise.
13521 (dr_known_forward_stride_p): Likewise.
13522 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13524 (runtime_alias_check_p): Allow runtime alias checks with
13526 (operator ==): Compare access_size and align.
13527 (prune_runtime_alias_test_list): Rework for new distinction between
13528 the access_size and seg_len.
13529 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13531 (get_segment_min_max): New function.
13532 (create_intersect_range_checks): Use it.
13533 (dr_step_indicator): New function.
13534 (dr_direction_indicator): Likewise.
13535 (dr_zero_step_indicator): Likewise.
13536 (dr_known_forward_stride_p): Likewise.
13537 * tree-loop-distribution.c (data_ref_segment_size): Return
13538 DR_STEP * (niters - 1).
13539 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13541 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13542 (vect_preserves_scalar_order_p): New function, split out from...
13543 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13544 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13545 (vect_vfa_access_size): New function.
13546 (vect_vfa_align): Likewise.
13547 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13548 (dump_lower_bound): New function.
13549 (vect_check_lower_bound): Likewise.
13550 (vect_small_gap_p): Likewise.
13551 (vectorizable_with_step_bound_p): Likewise.
13552 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13553 depencies if the vectorization factor is 1. Convert the checks
13554 for nonzero steps into checks on the bounds of DR_STEP. Try using
13555 a bunds check for variable steps if the minimum required step is
13556 relatively small. Update calls to the dr_with_seg_len
13557 constructor and to vect_compile_time_alias.
13558 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13560 (vect_loop_versioning): Call it.
13561 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13563 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13566 Alan Hayward <alan.hayward@arm.com>
13567 David Sherwood <david.sherwood@arm.com>
13569 * doc/sourcebuild.texi (vect_scatter_store): Document.
13570 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13572 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13574 * genopinit.c (main): Add supports_vec_scatter_store and
13575 supports_vec_scatter_store_cached to target_optabs.
13576 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13577 IFN_MASK_SCATTER_STORE.
13578 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13580 * internal-fn.h (internal_store_fn_p): Declare.
13581 (internal_fn_stored_value_index): Likewise.
13582 * internal-fn.c (scatter_store_direct): New macro.
13583 (expand_scatter_store_optab_fn): New function.
13584 (direct_scatter_store_optab_supported_p): New macro.
13585 (internal_store_fn_p): New function.
13586 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13587 IFN_MASK_SCATTER_STORE.
13588 (internal_fn_mask_index): Likewise.
13589 (internal_fn_stored_value_index): New function.
13590 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13591 for scatter stores.
13592 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13593 * optabs-query.c (supports_vec_scatter_store_p): New function.
13594 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13595 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13596 true for scatter stores.
13597 (vect_gather_scatter_fn_p): Handle scatter stores too.
13598 (vect_check_gather_scatter): Consider using scatter stores if
13599 supports_vec_scatter_store_p.
13600 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13601 scatter stores too.
13602 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13603 internal_fn_stored_value_index.
13604 (check_load_store_masking): Handle scatter stores too.
13605 (vect_get_store_rhs): Make public.
13606 (vectorizable_call): Use internal_store_fn_p.
13607 (vectorizable_store): Handle scatter store internal functions.
13608 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13609 when deciding whether the end of the group has been reached.
13610 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13611 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13612 (mask_scatter_store<mode>): New insns.
13614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13615 Alan Hayward <alan.hayward@arm.com>
13616 David Sherwood <david.sherwood@arm.com>
13618 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13619 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13620 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13622 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13623 Use vect_truncate_gather_scatter_offset if we can't treat the
13624 operation as a normal gather load or scatter store.
13625 (get_group_load_store_type): Take the gather_scatter_info
13626 as argument. Try using a gather load or scatter store for
13627 single-element groups.
13628 (get_load_store_type): Update calls to get_group_load_store_type
13629 and vect_use_strided_gather_scatters_p.
13631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13632 Alan Hayward <alan.hayward@arm.com>
13633 David Sherwood <david.sherwood@arm.com>
13635 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13636 optional tree argument.
13637 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13639 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13640 but continue to use the current value as a fallback.
13641 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13642 to compare the updates.
13643 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13644 (get_load_store_type): Use it when handling a strided access.
13645 (vect_get_strided_load_store_ops): New function.
13646 (vect_get_data_ptr_increment): Likewise.
13647 (vectorizable_load): Handle strided gather loads. Always pass
13648 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13650 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13651 Alan Hayward <alan.hayward@arm.com>
13652 David Sherwood <david.sherwood@arm.com>
13654 * doc/md.texi (gather_load@var{m}): Document.
13655 (mask_gather_load@var{m}): Likewise.
13656 * genopinit.c (main): Add supports_vec_gather_load and
13657 supports_vec_gather_load_cached to target_optabs.
13658 * optabs-tree.c (init_tree_optimization_optabs): Use
13659 ggc_cleared_alloc to allocate target_optabs.
13660 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13661 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13663 * internal-fn.h (internal_load_fn_p): Declare.
13664 (internal_gather_scatter_fn_p): Likewise.
13665 (internal_fn_mask_index): Likewise.
13666 (internal_gather_scatter_fn_supported_p): Likewise.
13667 * internal-fn.c (gather_load_direct): New macro.
13668 (expand_gather_load_optab_fn): New function.
13669 (direct_gather_load_optab_supported_p): New macro.
13670 (direct_internal_fn_optab): New function.
13671 (internal_load_fn_p): Likewise.
13672 (internal_gather_scatter_fn_p): Likewise.
13673 (internal_fn_mask_index): Likewise.
13674 (internal_gather_scatter_fn_supported_p): Likewise.
13675 * optabs-query.c (supports_at_least_one_mode_p): New function.
13676 (supports_vec_gather_load_p): Likewise.
13677 * optabs-query.h (supports_vec_gather_load_p): Declare.
13678 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13679 and memory_type field.
13680 (NUM_PATTERNS): Bump to 15.
13681 * tree-vect-data-refs.c: Include internal-fn.h.
13682 (vect_gather_scatter_fn_p): New function.
13683 (vect_describe_gather_scatter_call): Likewise.
13684 (vect_check_gather_scatter): Try using internal functions for
13685 gather loads. Recognize existing calls to a gather load function.
13686 (vect_analyze_data_refs): Consider using gather loads if
13687 supports_vec_gather_load_p.
13688 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13689 (vect_get_gather_scatter_offset_type): Likewise.
13690 (vect_convert_mask_for_vectype): Likewise.
13691 (vect_add_conversion_to_patterm): Likewise.
13692 (vect_try_gather_scatter_pattern): Likewise.
13693 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13694 (vect_vect_recog_func_ptrs): Add it.
13695 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13696 internal_fn_mask_index and internal_gather_scatter_fn_p.
13697 (check_load_store_masking): Take the gather_scatter_info as an
13698 argument and handle gather loads.
13699 (vect_get_gather_scatter_ops): New function.
13700 (vectorizable_call): Check internal_load_fn_p.
13701 (vectorizable_load): Likewise. Handle gather load internal
13703 (vectorizable_store): Update call to check_load_store_masking.
13704 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13705 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13706 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13707 (aarch64_gather_scale_operand_d): New predicates.
13708 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13709 (mask_gather_load<mode>): New insns.
13711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13712 Alan Hayward <alan.hayward@arm.com>
13713 David Sherwood <david.sherwood@arm.com>
13715 * optabs.def (fold_left_plus_optab): New optab.
13716 * doc/md.texi (fold_left_plus_@var{m}): Document.
13717 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13718 * internal-fn.c (fold_left_direct): Define.
13719 (expand_fold_left_optab_fn): Likewise.
13720 (direct_fold_left_optab_supported_p): Likewise.
13721 * fold-const-call.c (fold_const_fold_left): New function.
13722 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13723 * tree-parloops.c (valid_reduction_p): New function.
13724 (gather_scalar_reductions): Use it.
13725 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13726 (vect_finish_replace_stmt): Declare.
13727 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13728 (needs_fold_left_reduction_p): New function, split out from...
13729 (vect_is_simple_reduction): ...here. Accept reductions that
13730 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13731 (vect_force_simple_reduction): Also store the reduction type in
13732 the assignment's STMT_VINFO_REDUC_TYPE.
13733 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13734 (merge_with_identity): New function.
13735 (vect_expand_fold_left): Likewise.
13736 (vectorize_fold_left_reduction): Likewise.
13737 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13738 scalar phi in place for it. Check for target support and reject
13739 cases that would reassociate the operation. Defer the transform
13740 phase to vectorize_fold_left_reduction.
13741 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13742 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13743 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13747 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13748 call to ifc_temp_var.
13750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13751 Alan Hayward <alan.hayward@arm.com>
13752 David Sherwood <david.sherwood@arm.com>
13754 * target.def (legitimize_address_displacement): Take the original
13755 offset as a poly_int.
13756 * targhooks.h (default_legitimize_address_displacement): Update
13758 * targhooks.c (default_legitimize_address_displacement): Likewise.
13759 * doc/tm.texi: Regenerate.
13760 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13761 as an argument, moving assert of ad->disp == ad->disp_term to...
13762 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13763 Try calling targetm.legitimize_address_displacement before expanding
13764 the address rather than afterwards, and adjust for the new interface.
13765 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13766 Match the new hook interface. Handle SVE addresses.
13767 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13768 new hook interface.
13770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13772 * Makefile.in (OBJS): Add early-remat.o.
13773 * target.def (select_early_remat_modes): New hook.
13774 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13775 * doc/tm.texi: Regenerate.
13776 * targhooks.h (default_select_early_remat_modes): Declare.
13777 * targhooks.c (default_select_early_remat_modes): New function.
13778 * timevar.def (TV_EARLY_REMAT): New timevar.
13779 * passes.def (pass_early_remat): New pass.
13780 * tree-pass.h (make_pass_early_remat): Declare.
13781 * early-remat.c: New file.
13782 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13784 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13787 Alan Hayward <alan.hayward@arm.com>
13788 David Sherwood <david.sherwood@arm.com>
13790 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13791 vfm1 with a bound_epilog parameter.
13792 (vect_do_peeling): Update calls accordingly, and move the prologue
13793 call earlier in the function. Treat the base bound_epilog as 0 for
13794 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13795 this base when peeling for gaps.
13796 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13797 with fully-masked loops.
13798 (vect_estimate_min_profitable_iters): Handle the single peeled
13799 iteration in that case.
13801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13802 Alan Hayward <alan.hayward@arm.com>
13803 David Sherwood <david.sherwood@arm.com>
13805 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13806 single-element interleaving even if the size is not a power of 2.
13807 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13808 accesses for single-element interleaving if the group size is
13811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13812 Alan Hayward <alan.hayward@arm.com>
13813 David Sherwood <david.sherwood@arm.com>
13815 * doc/md.texi (fold_extract_last_@var{m}): Document.
13816 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13817 * optabs.def (fold_extract_last_optab): New optab.
13818 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13819 * internal-fn.c (fold_extract_direct): New macro.
13820 (expand_fold_extract_optab_fn): Likewise.
13821 (direct_fold_extract_optab_supported_p): Likewise.
13822 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13823 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13824 EXTRACT_LAST_REDUCTION.
13825 (get_initial_def_for_reduction): Do not create an initial vector
13826 for EXTRACT_LAST_REDUCTION reductions.
13827 (vectorizable_reduction): Leave the scalar phi in place for
13828 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13829 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13830 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13831 transform phase to vectorizable_condition.
13832 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13834 (vect_finish_stmt_generation): ...here.
13835 (vect_finish_replace_stmt): New function.
13836 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13837 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13839 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13842 Alan Hayward <alan.hayward@arm.com>
13843 David Sherwood <david.sherwood@arm.com>
13845 * doc/md.texi (extract_last_@var{m}): Document.
13846 * optabs.def (extract_last_optab): New optab.
13847 * internal-fn.def (EXTRACT_LAST): New internal function.
13848 * internal-fn.c (cond_unary_direct): New macro.
13849 (expand_cond_unary_optab_fn): Likewise.
13850 (direct_cond_unary_optab_supported_p): Likewise.
13851 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13852 loops using EXTRACT_LAST.
13853 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13854 (extract_last_<mode>): ...this optab.
13855 (vec_extract<mode><Vel>): Update accordingly.
13857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13858 Alan Hayward <alan.hayward@arm.com>
13859 David Sherwood <david.sherwood@arm.com>
13861 * target.def (empty_mask_is_expensive): New hook.
13862 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13863 * doc/tm.texi: Regenerate.
13864 * targhooks.h (default_empty_mask_is_expensive): Declare.
13865 * targhooks.c (default_empty_mask_is_expensive): New function.
13866 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13867 if the target says that empty masks are expensive.
13868 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13870 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13873 Alan Hayward <alan.hayward@arm.com>
13874 David Sherwood <david.sherwood@arm.com>
13876 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13877 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13878 (vect_use_loop_mask_for_alignment_p): New function.
13879 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13880 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13881 niters_skip argument. Make sure that the first niters_skip elements
13882 of the first iteration are inactive.
13883 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13884 Update call to vect_set_loop_masks_directly.
13885 (get_misalign_in_elems): New function, split out from...
13886 (vect_gen_prolog_loop_niters): ...here.
13887 (vect_update_init_of_dr): Take a code argument that specifies whether
13888 the adjustment should be added or subtracted.
13889 (vect_update_init_of_drs): Likewise.
13890 (vect_prepare_for_masked_peels): New function.
13891 (vect_do_peeling): Skip prologue peeling if we're using a mask
13892 instead. Update call to vect_update_inits_of_drs.
13893 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13895 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13896 alignment. Do not include the number of peeled iterations in
13897 the minimum threshold in that case.
13898 (vectorizable_induction): Adjust the start value down by
13899 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13900 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13901 Take the number of skipped iterations into account when calculating
13903 * tree-vect-stmts.c (vect_gen_while_not): New function.
13905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13906 Alan Hayward <alan.hayward@arm.com>
13907 David Sherwood <david.sherwood@arm.com>
13909 * doc/sourcebuild.texi (vect_fully_masked): Document.
13910 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13911 default value to 0.
13912 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13914 (vect_analyze_loop_2): ...here. Don't check the vectorization
13915 factor against the number of loop iterations if the loop is
13918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13919 Alan Hayward <alan.hayward@arm.com>
13920 David Sherwood <david.sherwood@arm.com>
13922 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13923 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13924 (dump_groups): Update accordingly.
13925 (iv_use::mem_type): New member variable.
13926 (address_p): New function.
13927 (record_use): Add a mem_type argument and initialize the new
13929 (record_group_use): Add a mem_type argument. Use address_p.
13930 Remove obsolete null checks of base_object. Update call to record_use.
13931 (find_interesting_uses_op): Update call to record_group_use.
13932 (find_interesting_uses_cond): Likewise.
13933 (find_interesting_uses_address): Likewise.
13934 (get_mem_type_for_internal_fn): New function.
13935 (find_address_like_use): Likewise.
13936 (find_interesting_uses_stmt): Try find_address_like_use before
13937 calling find_interesting_uses_op.
13938 (addr_offset_valid_p): Use the iv mem_type field as the type
13939 of the addressed memory.
13940 (add_autoinc_candidates): Likewise.
13941 (get_address_cost): Likewise.
13942 (split_small_address_groups_p): Use address_p.
13943 (split_address_groups): Likewise.
13944 (add_iv_candidate_for_use): Likewise.
13945 (autoinc_possible_for_pair): Likewise.
13946 (rewrite_groups): Likewise.
13947 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13948 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13949 (get_alias_ptr_type_for_ptr_address): New function.
13950 (rewrite_use_address): Rewrite address uses in calls that were
13951 identified by find_address_like_use.
13953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13954 Alan Hayward <alan.hayward@arm.com>
13955 David Sherwood <david.sherwood@arm.com>
13957 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13959 * gimple-expr.h (is_gimple_addressable: Likewise.
13960 * gimple-expr.c (is_gimple_address): Likewise.
13961 * internal-fn.c (expand_call_mem_ref): New function.
13962 (expand_mask_load_optab_fn): Use it.
13963 (expand_mask_store_optab_fn): Likewise.
13965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13966 Alan Hayward <alan.hayward@arm.com>
13967 David Sherwood <david.sherwood@arm.com>
13969 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13970 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13971 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13972 (cond_umax@var{mode}): Document.
13973 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13974 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13975 (cond_umin_optab, cond_umax_optab): New optabs.
13976 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13977 (COND_IOR, COND_XOR): New internal functions.
13978 * internal-fn.h (get_conditional_internal_fn): Declare.
13979 * internal-fn.c (cond_binary_direct): New macro.
13980 (expand_cond_binary_optab_fn): Likewise.
13981 (direct_cond_binary_optab_supported_p): Likewise.
13982 (get_conditional_internal_fn): New function.
13983 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13984 Cope with reduction statements that are vectorized as calls rather
13986 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13987 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13988 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13989 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13990 (UNSPEC_COND_EOR): New unspecs.
13991 (optab): Add mappings for them.
13992 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13993 (sve_int_op, sve_fp_op): New int attributes.
13995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13996 Alan Hayward <alan.hayward@arm.com>
13997 David Sherwood <david.sherwood@arm.com>
13999 * optabs.def (while_ult_optab): New optab.
14000 * doc/md.texi (while_ult@var{m}@var{n}): Document.
14001 * internal-fn.def (WHILE_ULT): New internal function.
14002 * internal-fn.h (direct_internal_fn_supported_p): New override
14003 that takes two types as argument.
14004 * internal-fn.c (while_direct): New macro.
14005 (expand_while_optab_fn): New function.
14006 (convert_optab_supported_p): Likewise.
14007 (direct_while_optab_supported_p): New macro.
14008 * wide-int.h (wi::udiv_ceil): New function.
14009 * tree-vectorizer.h (rgroup_masks): New structure.
14010 (vec_loop_masks): New typedef.
14011 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
14012 and fully_masked_p.
14013 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
14014 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
14015 (vect_max_vf): New function.
14016 (slpeel_make_loop_iterate_ntimes): Delete.
14017 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
14018 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
14019 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
14020 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
14021 internal-fn.h, stor-layout.h and optabs-query.h.
14022 (vect_set_loop_mask): New function.
14023 (add_preheader_seq): Likewise.
14024 (add_header_seq): Likewise.
14025 (interleave_supported_p): Likewise.
14026 (vect_maybe_permute_loop_masks): Likewise.
14027 (vect_set_loop_masks_directly): Likewise.
14028 (vect_set_loop_condition_masked): Likewise.
14029 (vect_set_loop_condition_unmasked): New function, split out from
14030 slpeel_make_loop_iterate_ntimes.
14031 (slpeel_make_loop_iterate_ntimes): Rename to..
14032 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
14033 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
14034 (vect_do_peeling): Update call accordingly.
14035 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
14037 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
14038 mask_compare_type, can_fully_mask_p and fully_masked_p.
14039 (release_vec_loop_masks): New function.
14040 (_loop_vec_info): Use it to free the loop masks.
14041 (can_produce_all_loop_masks_p): New function.
14042 (vect_get_max_nscalars_per_iter): Likewise.
14043 (vect_verify_full_masking): Likewise.
14044 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
14045 retries, and free the mask rgroups before retrying. Check loop-wide
14046 reasons for disallowing fully-masked loops. Make the final decision
14047 about whether use a fully-masked loop or not.
14048 (vect_estimate_min_profitable_iters): Do not assume that peeling
14049 for the number of iterations will be needed for fully-masked loops.
14050 (vectorizable_reduction): Disable fully-masked loops.
14051 (vectorizable_live_operation): Likewise.
14052 (vect_halve_mask_nunits): New function.
14053 (vect_double_mask_nunits): Likewise.
14054 (vect_record_loop_mask): Likewise.
14055 (vect_get_loop_mask): Likewise.
14056 (vect_transform_loop): Handle the case in which the final loop
14057 iteration might handle a partial vector. Call vect_set_loop_condition
14058 instead of slpeel_make_loop_iterate_ntimes.
14059 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
14060 (check_load_store_masking): New function.
14061 (prepare_load_store_mask): Likewise.
14062 (vectorizable_store): Handle fully-masked loops.
14063 (vectorizable_load): Likewise.
14064 (supportable_widening_operation): Use vect_halve_mask_nunits for
14066 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
14067 (vect_gen_while): New function.
14068 * config/aarch64/aarch64.md (umax<mode>3): New expander.
14069 (aarch64_uqdec<mode>): New insn.
14071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14072 Alan Hayward <alan.hayward@arm.com>
14073 David Sherwood <david.sherwood@arm.com>
14075 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
14076 (reduc_xor_scal_optab): New optabs.
14077 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
14078 (reduc_xor_scal_@var{m}): Document.
14079 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
14080 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
14081 internal functions.
14082 * fold-const-call.c (fold_const_call): Handle them.
14083 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
14084 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
14085 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
14086 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
14087 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
14088 (UNSPEC_XORV): New unspecs.
14089 (optab): Add entries for them.
14090 (BITWISEV): New int iterator.
14091 (bit_reduc_op): New int attributes.
14093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14094 Alan Hayward <alan.hayward@arm.com>
14095 David Sherwood <david.sherwood@arm.com>
14097 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
14098 * internal-fn.def (VEC_SHL_INSERT): New internal function.
14099 * optabs.def (vec_shl_insert_optab): New optab.
14100 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
14101 (duplicate_and_interleave): Likewise.
14102 * tree-vect-loop.c: Include internal-fn.h.
14103 (neutral_op_for_slp_reduction): New function, split out from
14104 get_initial_defs_for_reduction.
14105 (get_initial_def_for_reduction): Handle option 2 for variable-length
14106 vectors by loading the neutral value into a vector and then shifting
14107 the initial value into element 0.
14108 (get_initial_defs_for_reduction): Replace the code argument with
14109 the neutral value calculated by neutral_op_for_slp_reduction.
14110 Use gimple_build_vector for constant-length vectors.
14111 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14112 but the first group_size elements have a neutral value.
14113 Use duplicate_and_interleave otherwise.
14114 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14115 Update call to get_initial_defs_for_reduction. Handle SLP
14116 reductions for variable-length vectors by creating one vector
14117 result for each scalar result, with the elements associated
14118 with other scalar results stubbed out with the neutral value.
14119 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14120 Require IFN_VEC_SHL_INSERT for double reductions on
14121 variable-length vectors, or SLP reductions that have
14122 a neutral value. Require can_duplicate_and_interleave_p
14123 support for variable-length unchained SLP reductions if there
14124 is no neutral value, such as for MIN/MAX reductions. Also require
14125 the number of vector elements to be a multiple of the number of
14126 SLP statements when doing variable-length unchained SLP reductions.
14127 Update call to vect_create_epilog_for_reduction.
14128 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14129 and remove initial values.
14130 (duplicate_and_interleave): Make public.
14131 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14132 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14135 Alan Hayward <alan.hayward@arm.com>
14136 David Sherwood <david.sherwood@arm.com>
14138 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14139 (can_duplicate_and_interleave_p): New function.
14140 (vect_get_and_check_slp_defs): Take the vector of statements
14141 rather than just the current one. Remove excess parentheses.
14142 Restriction rejectinon of vect_constant_def and vect_external_def
14143 for variable-length vectors to boolean types, or types for which
14144 can_duplicate_and_interleave_p is false.
14145 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14146 (duplicate_and_interleave): New function.
14147 (vect_get_constant_vectors): Use gimple_build_vector for
14148 constant-length vectors and suitable variable-length constant
14149 vectors. Use duplicate_and_interleave for other variable-length
14150 vectors. Don't defer the update when inserting new statements.
14152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14153 Alan Hayward <alan.hayward@arm.com>
14154 David Sherwood <david.sherwood@arm.com>
14156 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14157 min_profitable_iters doesn't go negative.
14159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14160 Alan Hayward <alan.hayward@arm.com>
14161 David Sherwood <david.sherwood@arm.com>
14163 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14164 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14165 * optabs.def (vec_mask_load_lanes_optab): New optab.
14166 (vec_mask_store_lanes_optab): Likewise.
14167 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14168 (MASK_STORE_LANES): Likewise.
14169 * internal-fn.c (mask_load_lanes_direct): New macro.
14170 (mask_store_lanes_direct): Likewise.
14171 (expand_mask_load_optab_fn): Handle masked operations.
14172 (expand_mask_load_lanes_optab_fn): New macro.
14173 (expand_mask_store_optab_fn): Handle masked operations.
14174 (expand_mask_store_lanes_optab_fn): New macro.
14175 (direct_mask_load_lanes_optab_supported_p): Likewise.
14176 (direct_mask_store_lanes_optab_supported_p): Likewise.
14177 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14179 (vect_load_lanes_supported): Likewise.
14180 * tree-vect-data-refs.c (strip_conversion): New function.
14181 (can_group_stmts_p): Likewise.
14182 (vect_analyze_data_ref_accesses): Use it instead of checking
14183 for a pair of assignments.
14184 (vect_store_lanes_supported): Take a masked_p parameter.
14185 (vect_load_lanes_supported): Likewise.
14186 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14187 vect_store_lanes_supported and vect_load_lanes_supported.
14188 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14189 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14190 parameter. Don't allow gaps for masked accesses.
14191 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14192 and vect_load_lanes_supported.
14193 (get_load_store_type): Take a masked_p parameter and update
14194 call to get_group_load_store_type.
14195 (vectorizable_store): Update call to get_load_store_type.
14196 Handle IFN_MASK_STORE_LANES.
14197 (vectorizable_load): Update call to get_load_store_type.
14198 Handle IFN_MASK_LOAD_LANES.
14200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14201 Alan Hayward <alan.hayward@arm.com>
14202 David Sherwood <david.sherwood@arm.com>
14204 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14206 * config/aarch64/aarch64-protos.h
14207 (aarch64_sve_struct_memory_operand_p): Declare.
14208 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14209 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14210 (VPRED, vpred): Handle SVE structure modes.
14211 * config/aarch64/constraints.md (Utx): New constraint.
14212 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14213 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14214 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14215 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14216 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14217 structure modes. Split into pieces after RA.
14218 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14219 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14221 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14222 SVE structure modes.
14223 (aarch64_classify_address): Likewise.
14224 (sizetochar): Move earlier in file.
14225 (aarch64_print_operand): Handle SVE register lists.
14226 (aarch64_array_mode): New function.
14227 (aarch64_sve_struct_memory_operand_p): Likewise.
14228 (TARGET_ARRAY_MODE): Redefine.
14230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14231 Alan Hayward <alan.hayward@arm.com>
14232 David Sherwood <david.sherwood@arm.com>
14234 * target.def (array_mode): New target hook.
14235 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14236 * doc/tm.texi: Regenerate.
14237 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14238 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14239 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14240 targetm.array_mode.
14241 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14245 Alan Hayward <alan.hayward@arm.com>
14246 David Sherwood <david.sherwood@arm.com>
14248 * fold-const.c (fold_binary_loc): Check the argument types
14249 rather than the result type when testing for a vector operation.
14251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14253 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14254 * doc/tm.texi: Regenerate.
14256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14257 Alan Hayward <alan.hayward@arm.com>
14258 David Sherwood <david.sherwood@arm.com>
14260 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14261 (sve): Document new AArch64 extension.
14262 * doc/md.texi (w): Extend the description of the AArch64
14263 constraint to include SVE vectors.
14264 (Upl, Upa): Document new AArch64 predicate constraints.
14265 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14267 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14268 (msve-vector-bits=): New option.
14269 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14270 SVE when these are disabled.
14271 (sve): New extension.
14272 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14273 modes. Adjust their number of units based on aarch64_sve_vg.
14274 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14275 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14276 aarch64_addr_query_type.
14277 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14278 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14279 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14280 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14281 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14282 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14283 (aarch64_simd_imm_zero_p): Delete.
14284 (aarch64_check_zero_based_sve_index_immediate): Declare.
14285 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14286 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14287 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14288 (aarch64_sve_float_mul_immediate_p): Likewise.
14289 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14290 rather than an rtx.
14291 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14292 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14293 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14294 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14295 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14296 (aarch64_regmode_natural_size): Likewise.
14297 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14298 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14300 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14301 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14302 for VG and the SVE predicate registers.
14303 (V_ALIASES): Add a "z"-prefixed alias.
14304 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14305 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14306 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14307 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14308 (REG_CLASS_NAMES): Add entries for them.
14309 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14310 and the predicate registers.
14311 (aarch64_sve_vg): Declare.
14312 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14313 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14314 (REGMODE_NATURAL_SIZE): Define.
14315 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14317 * config/aarch64/aarch64.c: Include cfgrtl.h.
14318 (simd_immediate_info): Add a constructor for series vectors,
14319 and an associated step field.
14320 (aarch64_sve_vg): New variable.
14321 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14322 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14323 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14324 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14325 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14326 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14327 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14328 (aarch64_get_mask_mode): New functions.
14329 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14330 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14331 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14332 predicate modes and predicate registers. Explicitly restrict
14333 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14334 to store a vector mode if it is recognized by
14335 aarch64_classify_vector_mode.
14336 (aarch64_regmode_natural_size): New function.
14337 (aarch64_hard_regno_caller_save_mode): Return the original mode
14339 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14340 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14341 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14342 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14344 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14345 does not overlap dest if the function is frame-related. Handle
14347 (aarch64_split_add_offset): New function.
14348 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14349 them aarch64_add_offset.
14350 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14351 and update call to aarch64_sub_sp.
14352 (aarch64_add_cfa_expression): New function.
14353 (aarch64_expand_prologue): Pass extra temporary registers to the
14354 functions above. Handle the case in which we need to emit new
14355 DW_CFA_expressions for registers that were originally saved
14356 relative to the stack pointer, but now have to be expressed
14357 relative to the frame pointer.
14358 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14360 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14361 IP0 and IP1 values for SVE frames.
14362 (aarch64_expand_vec_series): New function.
14363 (aarch64_expand_sve_widened_duplicate): Likewise.
14364 (aarch64_expand_sve_const_vector): Likewise.
14365 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14366 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14367 into the register, rather than emitting a SET directly.
14368 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14369 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14370 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14371 (offset_9bit_signed_scaled_p): New functions.
14372 (aarch64_replicate_bitmask_imm): New function.
14373 (aarch64_bitmask_imm): Use it.
14374 (aarch64_cannot_force_const_mem): Reject expressions involving
14375 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14376 (aarch64_classify_index): Handle SVE indices, by requiring
14377 a plain register index with a scale that matches the element size.
14378 (aarch64_classify_address): Handle SVE addresses. Assert that
14379 the mode of the address is VOIDmode or an integer mode.
14380 Update call to aarch64_classify_symbol.
14381 (aarch64_classify_symbolic_expression): Update call to
14382 aarch64_classify_symbol.
14383 (aarch64_const_vec_all_in_range_p): New function.
14384 (aarch64_print_vector_float_operand): Likewise.
14385 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14386 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14387 and the FP immediates 1.0 and 0.5.
14388 (aarch64_print_address_internal): Handle SVE addresses.
14389 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14390 (aarch64_regno_regclass): Handle predicate registers.
14391 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14393 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14394 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14395 (aarch64_convert_sve_vector_bits): New function.
14396 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14397 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14398 rather than an rtx.
14399 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14400 Handle SVE vector and predicate modes. Accept VL-based constants
14401 that need only one temporary register, and VL offsets that require
14402 no temporary registers.
14403 (aarch64_conditional_register_usage): Mark the predicate registers
14404 as fixed if SVE isn't available.
14405 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14406 Return true for SVE vector and predicate modes.
14407 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14408 rather than an unsigned int. Handle SVE modes.
14409 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14411 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14413 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14414 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14415 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14416 (aarch64_sve_float_mul_immediate_p): New functions.
14417 (aarch64_sve_valid_immediate): New function.
14418 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14419 Explicitly reject structure modes. Check for INDEX constants.
14420 Handle PTRUE and PFALSE constants.
14421 (aarch64_check_zero_based_sve_index_immediate): New function.
14422 (aarch64_simd_imm_zero_p): Delete.
14423 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14424 vector modes. Accept constants in the range of CNT[BHWD].
14425 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14426 ask for an Advanced SIMD mode.
14427 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14428 (aarch64_simd_vector_alignment): Handle SVE predicates.
14429 (aarch64_vectorize_preferred_vector_alignment): New function.
14430 (aarch64_simd_vector_alignment_reachable): Use it instead of
14432 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14433 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14435 (MAX_VECT_LEN): Delete.
14436 (expand_vec_perm_d): Add a vec_flags field.
14437 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14438 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14439 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14441 (aarch64_evpc_rev): Rename to...
14442 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14443 (aarch64_evpc_rev_global): New function.
14444 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14445 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14447 (aarch64_evpc_sve_tbl): New function.
14448 (aarch64_expand_vec_perm_const_1): Update after rename of
14449 aarch64_evpc_rev. Handle SVE permutes too, trying
14450 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14451 than aarch64_evpc_tbl.
14452 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14453 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14454 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14455 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14456 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14457 (aarch64_expand_sve_vcond): New functions.
14458 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14459 of aarch64_vector_mode_p.
14460 (aarch64_dwarf_poly_indeterminate_value): New function.
14461 (aarch64_compute_pressure_classes): Likewise.
14462 (aarch64_can_change_mode_class): Likewise.
14463 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14464 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14465 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14466 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14467 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14468 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14469 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14470 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14472 (Dn, Dl, Dr): Accept const as well as const_vector.
14473 (Dz): Likewise. Compare against CONST0_RTX.
14474 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14475 of "vector" where appropriate.
14476 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14477 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14478 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14479 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14480 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14481 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14482 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14483 (v_int_equiv): Extend to SVE modes.
14484 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14486 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14487 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14488 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14489 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14490 (SVE_COND_FP_CMP): New int iterators.
14491 (perm_hilo): Handle the new unpack unspecs.
14492 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14494 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14495 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14496 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14497 (aarch64_equality_operator, aarch64_constant_vector_operand)
14498 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14499 (aarch64_sve_nonimmediate_operand): Likewise.
14500 (aarch64_sve_general_operand): Likewise.
14501 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14502 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14503 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14504 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14505 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14506 (aarch64_sve_float_arith_immediate): Likewise.
14507 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14508 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14509 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14510 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14511 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14512 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14513 (aarch64_sve_float_arith_operand): Likewise.
14514 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14515 (aarch64_sve_float_mul_operand): Likewise.
14516 (aarch64_sve_vec_perm_operand): Likewise.
14517 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14518 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14519 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14520 as well as const_vector.
14521 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14522 in file. Use CONST0_RTX and CONSTM1_RTX.
14523 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14524 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14525 Use aarch64_simd_imm_zero.
14526 * config/aarch64/aarch64-sve.md: New file.
14527 * config/aarch64/aarch64.md: Include it.
14528 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14529 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14530 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14531 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14532 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14533 (sve): New attribute.
14534 (enabled): Disable instructions with the sve attribute unless
14536 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14537 aarch64_expand_mov_immediate.
14538 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14539 CNT[BHSD] immediates.
14540 (movti): Split CONST_POLY_INT moves into two halves.
14541 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14542 Split additions that need a temporary here if the destination
14543 is the stack pointer.
14544 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14545 (*add<mode>3_poly_1): New instruction.
14546 (set_clobber_cc): New expander.
14548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14550 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14551 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14552 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14553 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14554 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14555 Change innermode from fixed_mode_size to machine_mode.
14556 (simplify_subreg): Update call accordingly. Handle a constant-sized
14557 subreg of a variable-length CONST_VECTOR.
14559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14560 Alan Hayward <alan.hayward@arm.com>
14561 David Sherwood <david.sherwood@arm.com>
14563 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14564 (add_offset_to_base): New function, split out from...
14565 (create_mem_ref): ...here. When handling a scale other than 1,
14566 check first whether the address is valid without the offset.
14567 Add it into the base if so, leaving the index and scale as-is.
14569 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14572 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14573 fold_for_warn before checking if arg2 is INTEGER_CST.
14575 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14577 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14578 (store_multiple_operation): Delete.
14579 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14580 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14581 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14582 guarded by TARGET_STRING.
14583 (rs6000_output_load_multiple): Delete.
14584 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14585 OPTION_MASK_STRING / TARGET_STRING handling.
14586 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14587 (const rs6000_opt_masks) <"string">: Change mask to 0.
14588 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14589 (MASK_STRING): Delete.
14590 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14592 (load_multiple): Delete.
14599 (store_multiple): Delete.
14606 (movmemsi_8reg): Delete.
14607 (corresponding unnamed define_insn): Delete.
14608 (movmemsi_6reg): Delete.
14609 (corresponding unnamed define_insn): Delete.
14610 (movmemsi_4reg): Delete.
14611 (corresponding unnamed define_insn): Delete.
14612 (movmemsi_2reg): Delete.
14613 (corresponding unnamed define_insn): Delete.
14614 (movmemsi_1reg): Delete.
14615 (corresponding unnamed define_insn): Delete.
14616 * config/rs6000/rs6000.opt (mno-string): New.
14617 (mstring): Replace by deprecation warning stub.
14618 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14620 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14622 * regrename.c (regrename_do_replace): If replacing the same
14623 reg multiple times, try to reuse last created gen_raw_REG.
14626 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14627 main to workaround a bug in GDB.
14629 2018-01-12 Tom de Vries <tom@codesourcery.com>
14632 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14634 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14636 PR rtl-optimization/80481
14637 * ira-color.c (get_cap_member): New function.
14638 (allocnos_conflict_by_live_ranges_p): Use it.
14639 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14640 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14642 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14645 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14646 (*saddl_se_1): Ditto.
14647 (*ssubsi_1): Ditto.
14648 (*ssubl_se_1): Ditto.
14650 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14652 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14653 rather than wi::to_widest for DR_INITs.
14654 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14655 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14656 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14658 (vect_analyze_group_access_1): Note that here.
14660 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14662 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14663 polynomial type sizes.
14665 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14667 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14668 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14669 (gimple_add_tmp_var): Likewise.
14671 2018-01-12 Martin Liska <mliska@suse.cz>
14673 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14674 (gimple_alloc_sizes): Likewise.
14675 (dump_gimple_statistics): Use PRIu64 in printf format.
14676 * gimple.h: Change uint64_t to int.
14678 2018-01-12 Martin Liska <mliska@suse.cz>
14680 * tree-core.h: Use uint64_t instead of int.
14681 * tree.c (tree_node_counts): Likewise.
14682 (tree_node_sizes): Likewise.
14683 (dump_tree_statistics): Use PRIu64 in printf format.
14685 2018-01-12 Martin Liska <mliska@suse.cz>
14687 * Makefile.in: As qsort_chk is implemented in vec.c, add
14688 vec.o to linkage of gencfn-macros.
14689 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14690 passing the info to record_node_allocation_statistics.
14691 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14693 * ggc-common.c (struct ggc_usage): Add operator== and use
14694 it in operator< and compare function.
14695 * mem-stats.h (struct mem_usage): Likewise.
14696 * vec.c (struct vec_usage): Remove operator< and compare
14697 function. Can be simply inherited.
14699 2018-01-12 Martin Jambor <mjambor@suse.cz>
14702 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14703 * tree-ssa-math-opts.c: Include domwalk.h.
14704 (convert_mult_to_fma_1): New function.
14705 (fma_transformation_info): New type.
14706 (fma_deferring_state): Likewise.
14707 (cancel_fma_deferring): New function.
14708 (result_of_phi): Likewise.
14709 (last_fma_candidate_feeds_initial_phi): Likewise.
14710 (convert_mult_to_fma): Added deferring logic, split actual
14711 transformation to convert_mult_to_fma_1.
14712 (math_opts_dom_walker): New type.
14713 (math_opts_dom_walker::after_dom_children): New method, body moved
14714 here from pass_optimize_widening_mul::execute, added deferring logic
14716 (pass_optimize_widening_mul::execute): Moved most of code to
14717 math_opts_dom_walker::after_dom_children.
14718 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14719 * config/i386/i386.c (ix86_option_override_internal): Added
14720 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14722 2018-01-12 Richard Biener <rguenther@suse.de>
14725 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14726 inline instance vars.
14728 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14731 * config/rx/rx.c (rx_is_restricted_memory_address):
14732 Handle SUBREG case.
14734 2018-01-12 Richard Biener <rguenther@suse.de>
14736 PR tree-optimization/80846
14737 * target.def (split_reduction): New target hook.
14738 * targhooks.c (default_split_reduction): New function.
14739 * targhooks.h (default_split_reduction): Declare.
14740 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14741 target requests first reduce vectors by combining low and high
14743 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14744 (get_vectype_for_scalar_type_and_size): Export.
14745 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14746 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14747 * doc/tm.texi: Regenerate.
14748 * config/i386/i386.c (ix86_split_reduction): Implement
14749 TARGET_VECTORIZE_SPLIT_REDUCTION.
14751 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14754 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14755 in PIC mode except for TARGET_VXWORKS_RTP.
14756 * config/sparc/sparc.c: Include cfgrtl.h.
14757 (TARGET_INIT_PIC_REG): Define.
14758 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14759 (sparc_pic_register_p): New predicate.
14760 (sparc_legitimate_address_p): Use it.
14761 (sparc_legitimize_pic_address): Likewise.
14762 (sparc_delegitimize_address): Likewise.
14763 (sparc_mode_dependent_address_p): Likewise.
14764 (gen_load_pcrel_sym): Remove 4th parameter.
14765 (load_got_register): Adjust call to above. Remove obsolete stuff.
14766 (sparc_expand_prologue): Do not call load_got_register here.
14767 (sparc_flat_expand_prologue): Likewise.
14768 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14769 (sparc_use_pseudo_pic_reg): New function.
14770 (sparc_init_pic_reg): Likewise.
14771 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14772 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14774 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14776 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14777 Add item for branch_cost.
14779 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14781 PR rtl-optimization/83565
14782 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14783 not extend the result to a larger mode for rotate operations.
14784 (num_sign_bit_copies1): Likewise.
14786 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14789 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14791 Use values-Xc.o for -pedantic.
14792 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14794 2018-01-12 Martin Liska <mliska@suse.cz>
14797 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14799 (possible_polymorphic_call_targets): Use it.
14800 (ipa_devirt): Likewise.
14802 2018-01-12 Martin Liska <mliska@suse.cz>
14804 * profile-count.h (enum profile_quality): Use 0 as invalid
14805 enum value of profile_quality.
14807 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14809 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14810 -mext-string options.
14812 2018-01-12 Richard Biener <rguenther@suse.de>
14814 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14815 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14816 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14818 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14820 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14822 * configure.ac (--with-long-double-format): Add support for the
14823 configuration option to change the default long double format on
14825 * config.gcc (powerpc*-linux*-*): Likewise.
14826 * configure: Regenerate.
14827 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14828 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14829 used without modification.
14831 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14833 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14834 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14835 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14836 MISC_BUILTIN_SPEC_BARRIER.
14837 (rs6000_init_builtins): Likewise.
14838 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14840 (speculation_barrier): New define_insn.
14841 * doc/extend.texi: Document __builtin_speculation_barrier.
14843 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14846 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14847 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14848 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14850 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14851 integral modes instead of "ss" and "sd".
14852 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14853 vectors with 32-bit and 64-bit elements.
14854 (vecdupssescalarmodesuffix): New mode attribute.
14855 (vec_dup<mode>): Use it.
14857 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14860 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14861 frame if argument is passed on stack.
14863 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14866 * ree.c (combine_reaching_defs): Optimize also
14867 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14868 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14870 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14872 PR middle-end/83189
14873 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14875 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14877 PR middle-end/83718
14878 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14879 after they are computed.
14881 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14883 PR tree-optimization/83695
14884 * gimple-loop-linterchange.cc
14885 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14886 reset cached scev information after interchange.
14887 (pass_linterchange::execute): Remove call to scev_reset_htab.
14889 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14891 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14892 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14893 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14894 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14895 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14896 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14897 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14898 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14899 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14900 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14901 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14902 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14903 (V_lane_reg): Likewise.
14904 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14906 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14907 (vfmal_lane_low<mode>_intrinsic,
14908 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14909 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14910 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14911 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14912 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14913 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14915 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14917 * config/arm/arm-cpus.in (fp16fml): New feature.
14918 (ALL_SIMD): Add fp16fml.
14919 (armv8.2-a): Add fp16fml as an option.
14920 (armv8.3-a): Likewise.
14921 (armv8.4-a): Add fp16fml as part of fp16.
14922 * config/arm/arm.h (TARGET_FP16FML): Define.
14923 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14925 * config/arm/arm-modes.def (V2HF): Define.
14926 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14927 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14928 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14929 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14930 vfmsl_low, vfmsl_high): New set of builtins.
14931 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14932 (vfml_op): New code attribute.
14933 (VFMLHALVES): New int iterator.
14934 (VFML, VFMLSEL): New mode attributes.
14935 (V_reg): Define mapping for V2HF.
14936 (V_hi, V_lo): New mode attributes.
14937 (VF_constraint): Likewise.
14938 (vfml_half, vfml_half_selector): New int attributes.
14939 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14941 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14942 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14944 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14945 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14946 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14947 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14949 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14950 Document new effective target and option set.
14952 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14954 * config/arm/arm-cpus.in (armv8_4): New feature.
14955 (ARMv8_4a): New fgroup.
14956 (armv8.4-a): New arch.
14957 * config/arm/arm-tables.opt: Regenerate.
14958 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14959 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14960 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14961 Add matching rules for -march=armv8.4-a and extensions.
14962 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14964 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14967 * config/rx/rx.md (BW): New mode attribute.
14968 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14970 2018-01-11 Richard Biener <rguenther@suse.de>
14972 PR tree-optimization/83435
14973 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14974 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14975 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14977 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14978 Alan Hayward <alan.hayward@arm.com>
14979 David Sherwood <david.sherwood@arm.com>
14981 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14983 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14984 (aarch64_print_address_internal): Use it to check for a zero offset.
14986 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14987 Alan Hayward <alan.hayward@arm.com>
14988 David Sherwood <david.sherwood@arm.com>
14990 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14991 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14992 Return a poly_int64 rather than a HOST_WIDE_INT.
14993 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14994 rather than a HOST_WIDE_INT.
14995 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14996 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14997 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14998 final_offset from HOST_WIDE_INT to poly_int64.
14999 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
15000 to_constant when getting the number of units in an Advanced SIMD
15002 (aarch64_builtin_vectorized_function): Check for a constant number
15004 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
15006 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
15007 attribute instead of GET_MODE_NUNITS.
15008 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
15009 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
15010 GET_MODE_SIZE for fixed-size registers.
15011 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
15012 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
15013 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
15014 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
15015 (aarch64_print_operand, aarch64_print_address_internal)
15016 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
15017 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
15018 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
15019 Handle polynomial GET_MODE_SIZE.
15020 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
15021 wider than SImode without modification.
15022 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
15023 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
15024 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
15025 passing and returning SVE modes.
15026 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
15027 rather than GEN_INT.
15028 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
15029 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
15030 (aarch64_allocate_and_probe_stack_space): Likewise.
15031 (aarch64_layout_frame): Cope with polynomial offsets.
15032 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
15033 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
15034 polynomial offsets.
15035 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
15036 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
15037 poly_int64 rather than a HOST_WIDE_INT.
15038 (aarch64_get_separate_components, aarch64_process_components)
15039 (aarch64_expand_prologue, aarch64_expand_epilogue)
15040 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
15041 (aarch64_anchor_offset): New function, split out from...
15042 (aarch64_legitimize_address): ...here.
15043 (aarch64_builtin_vectorization_cost): Handle polynomial
15044 TYPE_VECTOR_SUBPARTS.
15045 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
15047 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
15048 number of elements from the PARALLEL rather than the mode.
15049 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
15050 rather than GET_MODE_BITSIZE.
15051 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
15052 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
15053 (aarch64_expand_vec_perm_const_1): Handle polynomial
15054 d->perm.length () and d->perm elements.
15055 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
15056 Apply to_constant to d->perm elements.
15057 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
15058 polynomial CONST_VECTOR_NUNITS.
15059 (aarch64_move_pointer): Take amount as a poly_int64 rather
15061 (aarch64_progress_pointer): Avoid temporary variable.
15062 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
15063 the mode attribute instead of GET_MODE.
15065 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15066 Alan Hayward <alan.hayward@arm.com>
15067 David Sherwood <david.sherwood@arm.com>
15069 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
15070 x exists before using it.
15071 (aarch64_add_constant_internal): Rename to...
15072 (aarch64_add_offset_1): ...this. Replace regnum with separate
15073 src and dest rtxes. Handle the case in which they're different,
15074 including when the offset is zero. Replace scratchreg with an rtx.
15075 Use 2 additions if there is no spare register into which we can
15076 move a 16-bit constant.
15077 (aarch64_add_constant): Delete.
15078 (aarch64_add_offset): Replace reg with separate src and dest
15079 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
15080 Use aarch64_add_offset_1.
15081 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
15082 an rtx rather than an int. Take the delta as a poly_int64
15083 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
15084 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
15085 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
15086 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
15087 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
15088 and aarch64_add_sp.
15089 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
15090 aarch64_add_constant.
15092 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15094 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
15095 Use scalar_float_mode.
15097 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15099 * config/aarch64/aarch64-simd.md
15100 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
15101 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15102 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15103 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15104 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15105 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15106 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15107 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15108 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15109 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15111 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15114 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15115 targ_options->x_arm_arch_string is non NULL.
15117 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15119 * config/aarch64/aarch64.h
15120 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15122 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15125 * expmed.c (emit_store_flag_force): Swap if const op0
15126 and change VOIDmode to mode of op0.
15128 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15130 PR rtl-optimization/83761
15131 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15132 than bytes to mode_for_size.
15134 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15136 PR middle-end/83189
15137 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15138 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15141 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15143 PR middle-end/83575
15144 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15145 when in layout mode.
15146 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15147 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15150 2018-01-10 Michael Collison <michael.collison@arm.com>
15152 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15153 * config/aarch64/aarch64-option-extension.def: Add
15154 AARCH64_OPT_EXTENSION of 'fp16fml'.
15155 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15156 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15157 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15158 * config/aarch64/constraints.md (Ui7): New constraint.
15159 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15160 (VFMLA_SEL_W): Ditto.
15163 (VFMLA16_LOW): New int iterator.
15164 (VFMLA16_HIGH): Ditto.
15165 (UNSPEC_FMLAL): New unspec.
15166 (UNSPEC_FMLSL): Ditto.
15167 (UNSPEC_FMLAL2): Ditto.
15168 (UNSPEC_FMLSL2): Ditto.
15169 (f16mac): New code attribute.
15170 * config/aarch64/aarch64-simd-builtins.def
15171 (aarch64_fmlal_lowv2sf): Ditto.
15172 (aarch64_fmlsl_lowv2sf): Ditto.
15173 (aarch64_fmlalq_lowv4sf): Ditto.
15174 (aarch64_fmlslq_lowv4sf): Ditto.
15175 (aarch64_fmlal_highv2sf): Ditto.
15176 (aarch64_fmlsl_highv2sf): Ditto.
15177 (aarch64_fmlalq_highv4sf): Ditto.
15178 (aarch64_fmlslq_highv4sf): Ditto.
15179 (aarch64_fmlal_lane_lowv2sf): Ditto.
15180 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15181 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15182 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15183 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15184 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15185 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15186 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15187 (aarch64_fmlal_lane_highv2sf): Ditto.
15188 (aarch64_fmlsl_lane_highv2sf): Ditto.
15189 (aarch64_fmlal_laneq_highv2sf): Ditto.
15190 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15191 (aarch64_fmlalq_lane_highv4sf): Ditto.
15192 (aarch64_fmlsl_lane_highv4sf): Ditto.
15193 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15194 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15195 * config/aarch64/aarch64-simd.md:
15196 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15197 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15198 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15199 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15200 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15201 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15202 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15203 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15204 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15205 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15206 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15207 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15208 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15209 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15210 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15211 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15212 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15213 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15214 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15215 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15216 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15217 (vfmlsl_low_u32): Ditto.
15218 (vfmlalq_low_u32): Ditto.
15219 (vfmlslq_low_u32): Ditto.
15220 (vfmlal_high_u32): Ditto.
15221 (vfmlsl_high_u32): Ditto.
15222 (vfmlalq_high_u32): Ditto.
15223 (vfmlslq_high_u32): Ditto.
15224 (vfmlal_lane_low_u32): Ditto.
15225 (vfmlsl_lane_low_u32): Ditto.
15226 (vfmlal_laneq_low_u32): Ditto.
15227 (vfmlsl_laneq_low_u32): Ditto.
15228 (vfmlalq_lane_low_u32): Ditto.
15229 (vfmlslq_lane_low_u32): Ditto.
15230 (vfmlalq_laneq_low_u32): Ditto.
15231 (vfmlslq_laneq_low_u32): Ditto.
15232 (vfmlal_lane_high_u32): Ditto.
15233 (vfmlsl_lane_high_u32): Ditto.
15234 (vfmlal_laneq_high_u32): Ditto.
15235 (vfmlsl_laneq_high_u32): Ditto.
15236 (vfmlalq_lane_high_u32): Ditto.
15237 (vfmlslq_lane_high_u32): Ditto.
15238 (vfmlalq_laneq_high_u32): Ditto.
15239 (vfmlslq_laneq_high_u32): Ditto.
15240 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15241 (AARCH64_FL_FOR_ARCH8_4): New.
15242 (AARCH64_ISA_F16FML): New ISA flag.
15243 (TARGET_F16FML): New feature flag for fp16fml.
15244 (doc/invoke.texi): Document new fp16fml option.
15246 2018-01-10 Michael Collison <michael.collison@arm.com>
15248 * config/aarch64/aarch64-builtins.c:
15249 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15250 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15251 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15252 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15253 (AARCH64_ISA_SHA3): New ISA flag.
15254 (TARGET_SHA3): New feature flag for sha3.
15255 * config/aarch64/iterators.md (sha512_op): New int attribute.
15256 (CRYPTO_SHA512): New int iterator.
15257 (UNSPEC_SHA512H): New unspec.
15258 (UNSPEC_SHA512H2): Ditto.
15259 (UNSPEC_SHA512SU0): Ditto.
15260 (UNSPEC_SHA512SU1): Ditto.
15261 * config/aarch64/aarch64-simd-builtins.def
15262 (aarch64_crypto_sha512hqv2di): New builtin.
15263 (aarch64_crypto_sha512h2qv2di): Ditto.
15264 (aarch64_crypto_sha512su0qv2di): Ditto.
15265 (aarch64_crypto_sha512su1qv2di): Ditto.
15266 (aarch64_eor3qv8hi): Ditto.
15267 (aarch64_rax1qv2di): Ditto.
15268 (aarch64_xarqv2di): Ditto.
15269 (aarch64_bcaxqv8hi): Ditto.
15270 * config/aarch64/aarch64-simd.md:
15271 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15272 (aarch64_crypto_sha512su0qv2di): Ditto.
15273 (aarch64_crypto_sha512su1qv2di): Ditto.
15274 (aarch64_eor3qv8hi): Ditto.
15275 (aarch64_rax1qv2di): Ditto.
15276 (aarch64_xarqv2di): Ditto.
15277 (aarch64_bcaxqv8hi): Ditto.
15278 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15279 (vsha512h2q_u64): Ditto.
15280 (vsha512su0q_u64): Ditto.
15281 (vsha512su1q_u64): Ditto.
15282 (veor3q_u16): Ditto.
15283 (vrax1q_u64): Ditto.
15284 (vxarq_u64): Ditto.
15285 (vbcaxq_u16): Ditto.
15286 * config/arm/types.md (crypto_sha512): New type attribute.
15287 (crypto_sha3): Ditto.
15288 (doc/invoke.texi): Document new sha3 option.
15290 2018-01-10 Michael Collison <michael.collison@arm.com>
15292 * config/aarch64/aarch64-builtins.c:
15293 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15294 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15295 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15296 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15297 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15298 (AARCH64_ISA_SM4): New ISA flag.
15299 (TARGET_SM4): New feature flag for sm4.
15300 * config/aarch64/aarch64-simd-builtins.def
15301 (aarch64_sm3ss1qv4si): Ditto.
15302 (aarch64_sm3tt1aq4si): Ditto.
15303 (aarch64_sm3tt1bq4si): Ditto.
15304 (aarch64_sm3tt2aq4si): Ditto.
15305 (aarch64_sm3tt2bq4si): Ditto.
15306 (aarch64_sm3partw1qv4si): Ditto.
15307 (aarch64_sm3partw2qv4si): Ditto.
15308 (aarch64_sm4eqv4si): Ditto.
15309 (aarch64_sm4ekeyqv4si): Ditto.
15310 * config/aarch64/aarch64-simd.md:
15311 (aarch64_sm3ss1qv4si): Ditto.
15312 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15313 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15314 (aarch64_sm4eqv4si): Ditto.
15315 (aarch64_sm4ekeyqv4si): Ditto.
15316 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15317 (sm3part_op): Ditto.
15318 (CRYPTO_SM3TT): Ditto.
15319 (CRYPTO_SM3PART): Ditto.
15320 (UNSPEC_SM3SS1): New unspec.
15321 (UNSPEC_SM3TT1A): Ditto.
15322 (UNSPEC_SM3TT1B): Ditto.
15323 (UNSPEC_SM3TT2A): Ditto.
15324 (UNSPEC_SM3TT2B): Ditto.
15325 (UNSPEC_SM3PARTW1): Ditto.
15326 (UNSPEC_SM3PARTW2): Ditto.
15327 (UNSPEC_SM4E): Ditto.
15328 (UNSPEC_SM4EKEY): Ditto.
15329 * config/aarch64/constraints.md (Ui2): New constraint.
15330 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15331 * config/arm/types.md (crypto_sm3): New type attribute.
15332 (crypto_sm4): Ditto.
15333 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15334 (vsm3tt1aq_u32): Ditto.
15335 (vsm3tt1bq_u32): Ditto.
15336 (vsm3tt2aq_u32): Ditto.
15337 (vsm3tt2bq_u32): Ditto.
15338 (vsm3partw1q_u32): Ditto.
15339 (vsm3partw2q_u32): Ditto.
15340 (vsm4eq_u32): Ditto.
15341 (vsm4ekeyq_u32): Ditto.
15342 (doc/invoke.texi): Document new sm4 option.
15344 2018-01-10 Michael Collison <michael.collison@arm.com>
15346 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15347 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15348 (AARCH64_FL_FOR_ARCH8_4): New.
15349 (AARCH64_FL_V8_4): New flag.
15350 (doc/invoke.texi): Document new armv8.4-a option.
15352 2018-01-10 Michael Collison <michael.collison@arm.com>
15354 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15355 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15356 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15357 * config/aarch64/aarch64-option-extension.def: Add
15358 AARCH64_OPT_EXTENSION of 'sha2'.
15359 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15360 (crypto): Disable sha2 and aes if crypto disabled.
15361 (crypto): Enable aes and sha2 if enabled.
15362 (simd): Disable sha2 and aes if simd disabled.
15363 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15365 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15366 (TARGET_SHA2): New feature flag for sha2.
15367 (TARGET_AES): New feature flag for aes.
15368 * config/aarch64/aarch64-simd.md:
15369 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15370 conditional on TARGET_AES.
15371 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15372 (aarch64_crypto_sha1hsi): Make pattern conditional
15374 (aarch64_crypto_sha1hv4si): Ditto.
15375 (aarch64_be_crypto_sha1hv4si): Ditto.
15376 (aarch64_crypto_sha1su1v4si): Ditto.
15377 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15378 (aarch64_crypto_sha1su0v4si): Ditto.
15379 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15380 (aarch64_crypto_sha256su0v4si): Ditto.
15381 (aarch64_crypto_sha256su1v4si): Ditto.
15382 (doc/invoke.texi): Document new aes and sha2 options.
15384 2018-01-10 Martin Sebor <msebor@redhat.com>
15386 PR tree-optimization/83781
15387 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15390 2018-01-11 Martin Sebor <msebor@gmail.com>
15391 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15393 PR tree-optimization/83501
15394 PR tree-optimization/81703
15396 * tree-ssa-strlen.c (get_string_cst): Rename...
15397 (get_string_len): ...to this. Handle global constants.
15398 (handle_char_store): Adjust.
15400 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15401 Jim Wilson <jimw@sifive.com>
15403 * config/riscv/riscv-protos.h (riscv_output_return): New.
15404 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15405 (riscv_attribute_table, riscv_output_return),
15406 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15407 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15408 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15410 (riscv_expand_prologue): Add early return for naked function.
15411 (riscv_expand_epilogue): Likewise.
15412 (riscv_function_ok_for_sibcall): Return false for naked function.
15413 (riscv_set_current_function): New.
15414 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15415 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15416 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15417 * doc/extend.texi (RISC-V Function Attributes): New.
15419 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15421 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15422 check for 128-bit long double before checking TCmode.
15423 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15424 128-bit long doubles before checking TFmode or TCmode.
15425 (FLOAT128_IBM_P): Likewise.
15427 2018-01-10 Martin Sebor <msebor@redhat.com>
15429 PR tree-optimization/83671
15430 * builtins.c (c_strlen): Unconditionally return zero for the empty
15432 Use -Warray-bounds for warnings.
15433 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15434 for non-constant array indices with COMPONENT_REF, arrays of
15435 arrays, and pointers to arrays.
15436 (gimple_fold_builtin_strlen): Determine and set length range for
15437 non-constant character arrays.
15439 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15441 PR middle-end/81897
15442 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15445 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15447 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15449 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15452 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15453 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15454 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15455 indexed_or_indirect_operand predicate.
15456 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15457 (*vsx_le_perm_load_v8hi): Likewise.
15458 (*vsx_le_perm_load_v16qi): Likewise.
15459 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15460 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15461 (*vsx_le_perm_store_v8hi): Likewise.
15462 (*vsx_le_perm_store_v16qi): Likewise.
15463 (eight unnamed splitters): Likewise.
15465 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15467 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15468 * config/rs6000/emmintrin.h: Likewise.
15469 * config/rs6000/mmintrin.h: Likewise.
15470 * config/rs6000/xmmintrin.h: Likewise.
15472 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15475 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15477 * tree.c (tree_nop_conversion): Return true for location wrapper
15479 (maybe_wrap_with_location): New function.
15480 (selftest::check_strip_nops): New function.
15481 (selftest::test_location_wrappers): New function.
15482 (selftest::tree_c_tests): Call it.
15483 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15484 (maybe_wrap_with_location): New decl.
15485 (EXPR_LOCATION_WRAPPER_P): New macro.
15486 (location_wrapper_p): New inline function.
15487 (tree_strip_any_location_wrapper): New inline function.
15489 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15492 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15493 stack_realign_offset for the largest alignment of stack slot
15495 (ix86_find_max_used_stack_alignment): New function.
15496 (ix86_finalize_stack_frame_flags): Use it. Set
15497 max_used_stack_alignment if we don't realign stack.
15498 * config/i386/i386.h (machine_function): Add
15499 max_used_stack_alignment.
15501 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15503 * config/arm/arm.opt (-mbranch-cost): New option.
15504 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15507 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15510 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15511 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15513 2018-01-10 Richard Biener <rguenther@suse.de>
15516 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15517 early out so it also covers the case where we have a non-NULL
15520 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15522 PR tree-optimization/83753
15523 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15524 for non-strided grouped accesses if the number of elements is 1.
15526 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15529 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15530 * i386.h (TARGET_USE_GATHER): Define.
15531 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15533 2018-01-10 Martin Liska <mliska@suse.cz>
15536 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15537 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15539 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15540 CLEANUP_NO_PARTITIONING is not set.
15542 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15544 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15545 for vectors, as a partial revert of r254296.
15546 * rtl.h (const_vec_p): Delete.
15547 (const_vec_duplicate_p): Don't test for vector CONSTs.
15548 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15549 * expmed.c (make_tree): Likewise.
15552 * common.md (E, F): Use CONSTANT_P instead of checking for
15554 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15555 checking for CONST_VECTOR.
15557 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15559 PR middle-end/83575
15560 * predict.c (force_edge_cold): Handle in more sane way edges
15561 with no prediction.
15563 2018-01-09 Carl Love <cel@us.ibm.com>
15565 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15567 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15568 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15569 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15570 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15571 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15572 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15573 * config/rs6000/rs6000-protos.h: Add extern defition for
15574 rs6000_generate_float2_double_code.
15575 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15577 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15578 (float2_v2df): Add define_expand.
15580 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15583 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15584 op_mode in the force_to_mode call.
15586 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15588 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15589 instead of checking each element individually.
15590 (aarch64_evpc_uzp): Likewise.
15591 (aarch64_evpc_zip): Likewise.
15592 (aarch64_evpc_ext): Likewise.
15593 (aarch64_evpc_rev): Likewise.
15594 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15595 instead of checking each element individually. Return true without
15597 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15598 whether all selected elements come from the same input, instead of
15599 checking each element individually. Remove calls to gen_rtx_REG,
15600 start_sequence and end_sequence and instead assert that no rtl is
15603 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15605 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15606 order of HIGH and CONST checks.
15608 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15610 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15611 if the destination isn't an SSA_NAME.
15613 2018-01-09 Richard Biener <rguenther@suse.de>
15615 PR tree-optimization/83668
15616 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15618 (canonicalize_loop_form): ... here, renamed from ...
15619 (canonicalize_loop_closed_ssa_form): ... this and amended to
15620 swap successor edges for loop exit blocks to make us use
15621 the RPO order we need for initial schedule generation.
15623 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15625 PR tree-optimization/64811
15626 * match.pd: When optimizing comparisons with Inf, avoid
15627 introducing or losing exceptions from comparisons with NaN.
15629 2018-01-09 Martin Liska <mliska@suse.cz>
15632 * asan.c (shadow_mem_size): Add gcc_assert.
15634 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15636 Don't save registers in main().
15639 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15640 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15641 * config/avr/avr.c (avr_set_current_function): Don't error if
15642 naked, OS_task or OS_main are specified at the same time.
15643 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15645 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15647 * common/config/avr/avr-common.c (avr_option_optimization_table):
15648 Switch on -mmain-is-OS_task for optimizing compilations.
15650 2018-01-09 Richard Biener <rguenther@suse.de>
15652 PR tree-optimization/83572
15653 * graphite.c: Include cfganal.h.
15654 (graphite_transform_loops): Connect infinite loops to exit
15655 and remove fake edges at the end.
15657 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15659 * ipa-inline.c (edge_badness): Revert accidental checkin.
15661 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15664 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15665 symbols; not inline clones.
15667 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15670 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15671 hard registers. Formatting fixes.
15673 PR preprocessor/83722
15674 * gcc.c (try_generate_repro): Pass
15675 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15676 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15679 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15680 Kito Cheng <kito.cheng@gmail.com>
15682 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15683 (riscv_leaf_function_p): Delete.
15684 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15686 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15688 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15690 (do_ifelse): New function.
15691 (do_isel): New function.
15692 (do_sub3): New function.
15693 (do_add3): New function.
15694 (do_load_mask_compare): New function.
15695 (do_overlap_load_compare): New function.
15696 (expand_compare_loop): New function.
15697 (expand_block_compare): Call expand_compare_loop() when appropriate.
15698 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15699 option description.
15700 (-mblock-compare-inline-loop-limit): New option.
15702 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15705 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15706 Reverse order of second and third operands in first alternative.
15707 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15708 of first and second elements in UNSPEC_VPERMR vector.
15709 (altivec_expand_vec_perm_le): Likewise.
15711 2018-01-08 Jeff Law <law@redhat.com>
15713 PR rtl-optimizatin/81308
15714 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15715 (process_switch): If group_case_labels makes a change, then set
15717 (pass_convert_switch::execute): If a switch is converted, then
15718 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15720 PR rtl-optimization/81308
15721 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15724 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15726 PR target/83663 - Revert r255946
15727 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15728 generation for cases where splatting a value is not useful.
15729 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15730 across a vec_duplicate and a paradoxical subreg forming a vector
15731 mode to a vec_concat.
15733 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15735 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15736 -march=armv8.3-a variants.
15737 * config/arm/t-multilib: Likewise.
15738 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15740 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15742 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15744 (cceq_ior_compare_complement): Give it a name so I can use it, and
15745 change boolean_or_operator predicate to boolean_operator so it can
15746 be used to generate a crand.
15747 (eqne): New code iterator.
15748 (bd/bd_neg): New code_attrs.
15749 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15750 a single define_insn.
15751 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15752 decrement (bdnzt/bdnzf/bdzt/bdzf).
15753 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15754 with the new names of the branch decrement patterns, and added the
15755 names of the branch decrement conditional patterns.
15757 2018-01-08 Richard Biener <rguenther@suse.de>
15759 PR tree-optimization/83563
15760 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15763 2018-01-08 Richard Biener <rguenther@suse.de>
15765 PR middle-end/83713
15766 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15768 2018-01-08 Richard Biener <rguenther@suse.de>
15770 PR tree-optimization/83685
15771 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15772 references to abnormals.
15774 2018-01-08 Richard Biener <rguenther@suse.de>
15777 * dwarf2out.c (output_indirect_strings): Handle empty
15778 skeleton_debug_str_hash.
15779 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15781 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15783 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15784 (emit_store_direct): Likewise.
15785 (arc_trampoline_adjust_address): Likewise.
15786 (arc_asm_trampoline_template): New function.
15787 (arc_initialize_trampoline): Use asm_trampoline_template.
15788 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15789 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15790 * config/arc/arc.md (flush_icache): Delete pattern.
15792 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15794 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15795 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15798 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15801 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15802 by not USED_FOR_TARGET.
15803 (make_pass_resolve_sw_modes): Likewise.
15805 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15807 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15810 2018-01-08 Richard Biener <rguenther@suse.de>
15812 PR middle-end/83580
15813 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15815 2018-01-08 Richard Biener <rguenther@suse.de>
15817 PR middle-end/83517
15818 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15820 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15822 PR middle-end/81897
15823 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15824 basic blocks with a small number of successors.
15825 (convert_control_dep_chain_into_preds): Improve handling of
15827 (dump_predicates): Split apart into...
15828 (dump_pred_chain): ...here...
15829 (dump_pred_info): ...and here.
15830 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15831 (can_chain_union_be_invalidated_p): Improve check for invalidation
15833 (uninit_uses_cannot_happen): Avoid unnecessary if
15834 convert_control_dep_chain_into_preds yielded nothing.
15836 2018-01-06 Martin Sebor <msebor@redhat.com>
15838 PR tree-optimization/83640
15839 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15840 subtracting negative offset from size.
15841 (builtin_access::overlap): Adjust offset bounds of the access to fall
15842 within the size of the object if possible.
15844 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15846 PR rtl-optimization/83699
15847 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15848 extract_bit_field_as_subreg to cases in which the extracted
15849 value is also a vector.
15851 * lra-constraints.c (process_alt_operands): Test for the equivalence
15852 substitutions when detecting a possible reload cycle.
15854 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15857 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15858 by default if flag_selective_schedling{,2}. Formatting fixes.
15860 PR rtl-optimization/83682
15861 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15862 if it has non-VECTOR_MODE element mode.
15863 (vec_duplicate_p): Likewise.
15865 PR middle-end/83694
15866 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15867 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15869 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15872 * config/i386/i386-builtin.def
15873 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15874 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15875 Require also OPTION_MASK_ISA_AVX512F in addition to
15876 OPTION_MASK_ISA_GFNI.
15877 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15878 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15879 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15880 to OPTION_MASK_ISA_GFNI.
15881 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15882 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15883 OPTION_MASK_ISA_AVX512BW.
15884 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15885 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15886 addition to OPTION_MASK_ISA_GFNI.
15887 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15888 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15889 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15890 to OPTION_MASK_ISA_GFNI.
15891 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15892 a requirement for all ISAs rather than any of them with a few
15894 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15896 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15897 bitmasks to be enabled with 3 exceptions, instead of requiring any
15898 enabled ISA with lots of exceptions.
15899 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15900 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15901 Change avx512bw in isa attribute to avx512f.
15902 * config/i386/sgxintrin.h: Add license boilerplate.
15903 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15904 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15905 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15906 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15908 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15909 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15910 temporarily sse2 rather than sse if not enabled already.
15913 * config/i386/sse.md (VI248_VLBW): Rename to ...
15914 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15915 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15916 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15917 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15918 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15919 mode iterator instead of VI248_VLBW.
15921 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15923 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15924 (record_modified): Skip clobbers; add debug output.
15925 (param_change_prob): Use sreal frequencies.
15927 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15929 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15930 punt for user-aligned variables.
15932 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15934 * tree-chrec.c (chrec_contains_symbols): Return true for
15937 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15940 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15941 of (x|y) == x for BICS pattern.
15943 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15945 PR tree-optimization/83605
15946 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15947 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15950 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15952 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15953 * config/epiphany/rtems.h: New file.
15955 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15956 Uros Bizjak <ubizjak@gmail.com>
15959 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15960 QIreg_operand instead of register_operand predicate.
15961 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15962 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15963 comments instead of -fmitigate[-_]rop.
15965 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15968 * cgraphunit.c (symbol_table::compile): Switch to text_section
15969 before calling assembly_start debug hook.
15970 * run-rtl-passes.c (run_rtl_passes): Likewise.
15973 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15975 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15976 range_int_cst_p rather than !symbolic_range_p before calling
15977 extract_range_from_multiplicative_op_1.
15979 2018-01-04 Jeff Law <law@redhat.com>
15981 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15982 redundant test in assertion.
15984 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15986 * doc/rtl.texi: Document machine_mode wrapper classes.
15988 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15990 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15991 using tree_to_uhwi.
15993 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15995 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15996 the VEC_PERM_EXPR fold to fail.
15998 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16001 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
16002 to switched_sections.
16004 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16007 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
16008 test for d.testing.
16010 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
16013 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
16014 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
16016 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16019 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
16020 is BLKmode and bitpos not zero or mode change is needed.
16022 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
16025 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
16028 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
16031 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
16032 instead of MULT rtx. Update all corresponding splitters.
16033 (*saddl_se): Ditto.
16034 (*ssub<modesuffix>): Ditto.
16035 (*ssubl_se): Ditto.
16036 (*cmp_sadd_di): Update split patterns.
16037 (*cmp_sadd_si): Ditto.
16038 (*cmp_sadd_sidi): Ditto.
16039 (*cmp_ssub_di): Ditto.
16040 (*cmp_ssub_si): Ditto.
16041 (*cmp_ssub_sidi): Ditto.
16042 * config/alpha/predicates.md (const23_operand): New predicate.
16043 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
16044 Look for ASHIFT, not MULT inner operand.
16045 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
16047 2018-01-04 Martin Liska <mliska@suse.cz>
16049 PR gcov-profile/83669
16050 * gcov.c (output_intermediate_file): Add version to intermediate
16052 * doc/gcov.texi: Document new field 'version' in intermediate
16053 file format. Fix location of '-k' option of gcov command.
16055 2018-01-04 Martin Liska <mliska@suse.cz>
16058 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
16060 2018-01-04 Jakub Jelinek <jakub@redhat.com>
16062 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
16064 2018-01-03 Martin Sebor <msebor@redhat.com>
16066 PR tree-optimization/83655
16067 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
16068 checking calls with invalid arguments.
16070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16072 * tree-vect-stmts.c (vect_get_store_rhs): New function.
16073 (vectorizable_mask_load_store): Delete.
16074 (vectorizable_call): Return false for masked loads and stores.
16075 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
16076 instead of gimple_assign_rhs1.
16077 (vectorizable_load): Handle IFN_MASK_LOAD.
16078 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
16080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16082 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
16084 (vectorizable_mask_load_store): ...here.
16085 (vectorizable_load): ...and here.
16087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16089 * tree-vect-stmts.c (vect_build_all_ones_mask)
16090 (vect_build_zero_merge_argument): New functions, split out from...
16091 (vectorizable_load): ...here.
16093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16095 * tree-vect-stmts.c (vect_check_store_rhs): New function,
16097 (vectorizable_mask_load_store): ...here.
16098 (vectorizable_store): ...and here.
16100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16102 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16104 (vectorizable_mask_load_store): ...here.
16106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16108 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16109 (vect_model_store_cost): Take a vec_load_store_type instead of a
16111 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16112 (vect_model_store_cost): Take a vec_load_store_type instead of a
16114 (vectorizable_mask_load_store): Update accordingly.
16115 (vectorizable_store): Likewise.
16116 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16120 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16121 IFN_MASK_LOAD calls here rather than...
16122 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16125 Alan Hayward <alan.hayward@arm.com>
16126 David Sherwood <david.sherwood@arm.com>
16128 * expmed.c (extract_bit_field_1): For vector extracts,
16129 fall back to extract_bit_field_as_subreg if vec_extract
16132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16133 Alan Hayward <alan.hayward@arm.com>
16134 David Sherwood <david.sherwood@arm.com>
16136 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16137 they are variable or constant sized.
16138 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16139 slots for constant-sized data.
16141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16142 Alan Hayward <alan.hayward@arm.com>
16143 David Sherwood <david.sherwood@arm.com>
16145 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16146 handling COND_EXPRs with boolean comparisons, try to find a better
16147 basis for the mask type than the boolean itself.
16149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16151 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16152 is calculated and how it can be overridden.
16153 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16154 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16156 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16160 Alan Hayward <alan.hayward@arm.com>
16161 David Sherwood <david.sherwood@arm.com>
16163 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16164 Remove the mode argument.
16165 (aarch64_simd_valid_immediate): Remove the mode and inverse
16167 * config/aarch64/iterators.md (bitsize): New iterator.
16168 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16169 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16170 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16171 aarch64_simd_valid_immediate.
16172 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16173 (aarch64_reg_or_bic_imm): Likewise.
16174 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16175 with an insn_type enum and msl with a modifier_type enum.
16176 Replace element_width with a scalar_mode. Change the shift
16177 to unsigned int. Add constructors for scalar_float_mode and
16178 scalar_int_mode elements.
16179 (aarch64_vect_float_const_representable_p): Delete.
16180 (aarch64_can_const_movi_rtx_p)
16181 (aarch64_simd_scalar_immediate_valid_for_move)
16182 (aarch64_simd_make_constant): Update call to
16183 aarch64_simd_valid_immediate.
16184 (aarch64_advsimd_valid_immediate_hs): New function.
16185 (aarch64_advsimd_valid_immediate): Likewise.
16186 (aarch64_simd_valid_immediate): Remove mode and inverse
16187 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16188 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16189 and aarch64_float_const_representable_p on the result.
16190 (aarch64_output_simd_mov_immediate): Remove mode argument.
16191 Update call to aarch64_simd_valid_immediate and use of
16192 simd_immediate_info.
16193 (aarch64_output_scalar_simd_mov_immediate): Update call
16196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16197 Alan Hayward <alan.hayward@arm.com>
16198 David Sherwood <david.sherwood@arm.com>
16200 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16201 (mode_nunits): Likewise CONST_MODE_NUNITS.
16202 * machmode.def (ADJUST_NUNITS): Document.
16203 * genmodes.c (mode_data::need_nunits_adj): New field.
16204 (blank_mode): Update accordingly.
16205 (adj_nunits): New variable.
16206 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16208 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16209 listed in adj_nunits.
16210 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16211 listed in adj_nunits. Don't emit case statements for such modes.
16212 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16213 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16214 nothing if adj_nunits is nonnull.
16215 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16216 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16217 (emit_mode_fbit): Update use of print_maybe_const_decl.
16218 (emit_move_size): Likewise. Treat the array as non-const
16220 (emit_mode_adjustments): Handle adj_nunits.
16222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16224 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16225 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16226 (VECTOR_MODES): Use it.
16227 (make_vector_modes): Take the prefix as an argument.
16229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16230 Alan Hayward <alan.hayward@arm.com>
16231 David Sherwood <david.sherwood@arm.com>
16233 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16234 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16235 for MODE_VECTOR_BOOL.
16236 * machmode.def (VECTOR_BOOL_MODE): Document.
16237 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16238 (make_vector_bool_mode): New function.
16239 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16241 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16242 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16244 * stor-layout.c (int_mode_for_mode): Likewise.
16245 * tree.c (build_vector_type_for_mode): Likewise.
16246 * varasm.c (output_constant_pool_2): Likewise.
16247 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16248 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16249 for MODE_VECTOR_BOOL.
16250 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16251 of mode class checks.
16252 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16253 instead of a list of mode class checks.
16254 (expand_vector_scalar_condition): Likewise.
16255 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16258 Alan Hayward <alan.hayward@arm.com>
16259 David Sherwood <david.sherwood@arm.com>
16261 * machmode.h (mode_size): Change from unsigned short to
16263 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16264 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16265 or if measurement_type is not polynomial.
16266 (fixed_size_mode::includes_p): Check for constant-sized modes.
16267 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16268 return a poly_uint16 rather than an unsigned short.
16269 (emit_mode_size): Change the type of mode_size from unsigned short
16270 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16271 (emit_mode_adjustments): Cope with polynomial vector sizes.
16272 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16274 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16276 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16277 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16278 * caller-save.c (setup_save_areas): Likewise.
16279 (replace_reg_with_saved_mem): Likewise.
16280 * calls.c (emit_library_call_value_1): Likewise.
16281 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16282 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16283 (gen_lowpart_for_combine): Likewise.
16284 * convert.c (convert_to_integer_1): Likewise.
16285 * cse.c (equiv_constant, cse_insn): Likewise.
16286 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16287 (cselib_subst_to_values): Likewise.
16288 * dce.c (word_dce_process_block): Likewise.
16289 * df-problems.c (df_word_lr_mark_ref): Likewise.
16290 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16291 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16292 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16293 (rtl_for_decl_location): Likewise.
16294 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16295 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16296 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16297 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16298 (expand_expr_real_1): Likewise.
16299 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16300 (pad_below): Likewise.
16301 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16302 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16303 * ira.c (get_subreg_tracking_sizes): Likewise.
16304 * ira-build.c (ira_create_allocno_objects): Likewise.
16305 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16306 (ira_sort_regnos_for_alter_reg): Likewise.
16307 * ira-costs.c (record_operand_costs): Likewise.
16308 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16309 (resolve_simple_move): Likewise.
16310 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16311 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16312 (lra_constraints): Likewise.
16313 (CONST_POOL_OK_P): Reject variable-sized modes.
16314 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16315 (add_pseudo_to_slot, lra_spill): Likewise.
16316 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16317 * optabs-query.c (get_best_extraction_insn): Likewise.
16318 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16319 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16320 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16321 * recog.c (offsettable_address_addr_space_p): Likewise.
16322 * regcprop.c (maybe_mode_change): Likewise.
16323 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16324 * regrename.c (build_def_use): Likewise.
16325 * regstat.c (dump_reg_info): Likewise.
16326 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16327 (find_reloads, find_reloads_subreg_address): Likewise.
16328 * reload1.c (eliminate_regs_1): Likewise.
16329 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16330 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16331 (simplify_binary_operation_1, simplify_subreg): Likewise.
16332 * targhooks.c (default_function_arg_padding): Likewise.
16333 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16334 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16335 (verify_gimple_assign_ternary): Likewise.
16336 * tree-inline.c (estimate_move_cost): Likewise.
16337 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16338 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16339 (get_address_cost_ainc): Likewise.
16340 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16341 (vect_supportable_dr_alignment): Likewise.
16342 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16343 (vectorizable_reduction): Likewise.
16344 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16345 (vectorizable_operation, vectorizable_load): Likewise.
16346 * tree.c (build_same_sized_truth_vector_type): Likewise.
16347 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16348 * var-tracking.c (emit_note_insn_var_location): Likewise.
16349 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16350 (ADDR_VEC_ALIGN): Likewise.
16352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16353 Alan Hayward <alan.hayward@arm.com>
16354 David Sherwood <david.sherwood@arm.com>
16356 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16358 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16359 or if measurement_type is polynomial.
16360 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16361 * combine.c (make_extraction): Likewise.
16362 * dse.c (find_shift_sequence): Likewise.
16363 * dwarf2out.c (mem_loc_descriptor): Likewise.
16364 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16365 (extract_bit_field, extract_low_bits): Likewise.
16366 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16367 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16368 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16369 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16370 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16371 * reload.c (find_reloads): Likewise.
16372 * reload1.c (alter_reg): Likewise.
16373 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16374 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16375 * tree-if-conv.c (predicate_mem_writes): Likewise.
16376 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16377 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16378 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16379 * valtrack.c (dead_debug_insert_temp): Likewise.
16380 * varasm.c (mergeable_constant_section): Likewise.
16381 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16384 Alan Hayward <alan.hayward@arm.com>
16385 David Sherwood <david.sherwood@arm.com>
16387 * expr.c (expand_assignment): Cope with polynomial mode sizes
16388 when assigning to a CONCAT.
16390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16391 Alan Hayward <alan.hayward@arm.com>
16392 David Sherwood <david.sherwood@arm.com>
16394 * machmode.h (mode_precision): Change from unsigned short to
16396 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16398 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16399 or if measurement_type is not polynomial.
16400 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16401 in which the mode is already known to be a scalar_int_mode.
16402 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16403 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16405 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16406 for GET_MODE_PRECISION.
16407 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16408 for GET_MODE_PRECISION.
16409 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16411 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16412 (expand_field_assignment, make_extraction): Likewise.
16413 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16414 (get_last_value): Likewise.
16415 * convert.c (convert_to_integer_1): Likewise.
16416 * cse.c (cse_insn): Likewise.
16417 * expr.c (expand_expr_real_1): Likewise.
16418 * lra-constraints.c (simplify_operand_subreg): Likewise.
16419 * optabs-query.c (can_atomic_load_p): Likewise.
16420 * optabs.c (expand_atomic_load): Likewise.
16421 (expand_atomic_store): Likewise.
16422 * ree.c (combine_reaching_defs): Likewise.
16423 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16424 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16425 * tree.h (type_has_mode_precision_p): Likewise.
16426 * ubsan.c (instrument_si_overflow): Likewise.
16428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16429 Alan Hayward <alan.hayward@arm.com>
16430 David Sherwood <david.sherwood@arm.com>
16432 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16433 polynomial numbers of units.
16434 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16435 (valid_vector_subparts_p): New function.
16436 (build_vector_type): Remove temporary shim and take the number
16437 of units as a poly_uint64 rather than an int.
16438 (build_opaque_vector_type): Take the number of units as a
16439 poly_uint64 rather than an int.
16440 * tree.c (build_vector_from_ctor): Handle polynomial
16441 TYPE_VECTOR_SUBPARTS.
16442 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16443 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16444 (build_vector_from_val): If the number of units is variable,
16445 use build_vec_duplicate_cst for constant operands and
16446 VEC_DUPLICATE_EXPR otherwise.
16447 (make_vector_type): Remove temporary is_constant ().
16448 (build_vector_type, build_opaque_vector_type): Take the number of
16449 units as a poly_uint64 rather than an int.
16450 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16452 * cfgexpand.c (expand_debug_expr): Likewise.
16453 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16454 (store_constructor, expand_expr_real_1): Likewise.
16455 (const_scalar_mask_from_tree): Likewise.
16456 * fold-const-call.c (fold_const_reduction): Likewise.
16457 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16458 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16459 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16460 (fold_relational_const): Likewise.
16461 (native_interpret_vector): Likewise. Change the size from an
16462 int to an unsigned int.
16463 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16464 TYPE_VECTOR_SUBPARTS.
16465 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16466 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16467 duplicating a non-constant operand into a variable-length vector.
16468 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16469 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16470 * ipa-icf.c (sem_variable::equals): Likewise.
16471 * match.pd: Likewise.
16472 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16473 * print-tree.c (print_node): Likewise.
16474 * stor-layout.c (layout_type): Likewise.
16475 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16476 * tree-cfg.c (verify_gimple_comparison): Likewise.
16477 (verify_gimple_assign_binary): Likewise.
16478 (verify_gimple_assign_ternary): Likewise.
16479 (verify_gimple_assign_single): Likewise.
16480 * tree-pretty-print.c (dump_generic_node): Likewise.
16481 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16482 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16483 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16484 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16485 (vect_shift_permute_load_chain): Likewise.
16486 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16487 (expand_vector_condition, optimize_vector_constructor): Likewise.
16488 (lower_vec_perm, get_compute_type): Likewise.
16489 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16490 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16491 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16492 (vect_recog_mask_conversion_pattern): Likewise.
16493 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16494 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16495 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16496 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16497 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16498 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16499 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16500 (supportable_widening_operation): Likewise.
16501 (supportable_narrowing_operation): Likewise.
16502 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16504 * varasm.c (output_constant): Likewise.
16506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16507 Alan Hayward <alan.hayward@arm.com>
16508 David Sherwood <david.sherwood@arm.com>
16510 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16511 so that both the length == 3 and length != 3 cases set up their
16512 own permute vectors. Add comments explaining why we know the
16513 number of elements is constant.
16514 (vect_permute_load_chain): Likewise.
16516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16517 Alan Hayward <alan.hayward@arm.com>
16518 David Sherwood <david.sherwood@arm.com>
16520 * machmode.h (mode_nunits): Change from unsigned char to
16522 (ONLY_FIXED_SIZE_MODES): New macro.
16523 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16524 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16525 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16527 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16528 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16529 or if measurement_type is not polynomial.
16530 * genmodes.c (ZERO_COEFFS): New macro.
16531 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16533 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16534 Use ZERO_COEFFS when emitting initializers.
16535 * data-streamer.h (bp_pack_poly_value): New function.
16536 (bp_unpack_poly_value): Likewise.
16537 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16538 for GET_MODE_NUNITS.
16539 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16540 for GET_MODE_NUNITS.
16541 * tree.c (make_vector_type): Remove temporary shim and make
16542 the real function take the number of units as a poly_uint64
16543 rather than an int.
16544 (build_vector_type_for_mode): Handle polynomial nunits.
16545 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16546 * emit-rtl.c (const_vec_series_p_1): Likewise.
16547 (gen_rtx_CONST_VECTOR): Likewise.
16548 * fold-const.c (test_vec_duplicate_folding): Likewise.
16549 * genrecog.c (validate_pattern): Likewise.
16550 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16551 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16552 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16553 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16554 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16555 * rtlanal.c (subreg_get_info): Likewise.
16556 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16557 (vect_grouped_load_supported): Likewise.
16558 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16559 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16560 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16561 (simplify_const_unary_operation, simplify_binary_operation_1)
16562 (simplify_const_binary_operation, simplify_ternary_operation)
16563 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16564 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16565 instead of CONST_VECTOR_NUNITS.
16566 * varasm.c (output_constant_pool_2): Likewise.
16567 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16568 explicit-encoded elements in the XVEC for variable-length vectors.
16570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16572 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16575 Alan Hayward <alan.hayward@arm.com>
16576 David Sherwood <david.sherwood@arm.com>
16578 * coretypes.h (fixed_size_mode): Declare.
16579 (fixed_size_mode_pod): New typedef.
16580 * builtins.h (target_builtins::x_apply_args_mode)
16581 (target_builtins::x_apply_result_mode): Change type to
16582 fixed_size_mode_pod.
16583 * builtins.c (apply_args_size, apply_result_size, result_vector)
16584 (expand_builtin_apply_args_1, expand_builtin_apply)
16585 (expand_builtin_return): Update accordingly.
16587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16589 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16590 * cselib.c (cselib_hash_rtx): Likewise.
16591 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16592 CONST_VECTOR encoding.
16594 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16595 Jeff Law <law@redhat.com>
16598 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16599 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16600 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16601 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16604 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16605 explicitly probe *sp in a noreturn function if there were any callee
16606 register saves or frame pointer is needed.
16608 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16611 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16612 BLKmode for ternary, binary or unary expressions.
16615 * var-tracking.c (delete_vta_debug_insn): New inline function.
16616 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16617 insns from get_insns () to NULL instead of each bb separately.
16618 Use delete_vta_debug_insn. No longer static.
16619 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16620 delete_vta_debug_insns callers.
16621 * rtl.h (delete_vta_debug_insns): Declare.
16622 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16623 instead of variable_tracking_main.
16625 2018-01-03 Martin Sebor <msebor@redhat.com>
16627 PR tree-optimization/83603
16628 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16629 arguments past the endof the argument list in functions declared
16630 without a prototype.
16631 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16632 Avoid checking when arguments are null.
16634 2018-01-03 Martin Sebor <msebor@redhat.com>
16637 * doc/extend.texi (attribute const): Fix a typo.
16638 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16639 issuing -Wsuggest-attribute for void functions.
16641 2018-01-03 Martin Sebor <msebor@redhat.com>
16643 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16644 offset_int::from instead of wide_int::to_shwi.
16645 (maybe_diag_overlap): Remove assertion.
16646 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16647 * gimple-ssa-sprintf.c (format_directive): Same.
16648 (parse_directive): Same.
16649 (sprintf_dom_walker::compute_format_length): Same.
16650 (try_substitute_return_value): Same.
16652 2018-01-03 Jeff Law <law@redhat.com>
16654 PR middle-end/83654
16655 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16656 non-constant residual for zero at runtime and avoid probing in
16657 that case. Reorganize code for trailing problem to mirror handling
16660 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16662 PR tree-optimization/83501
16663 * tree-ssa-strlen.c (get_string_cst): New.
16664 (handle_char_store): Call get_string_cst.
16666 2018-01-03 Martin Liska <mliska@suse.cz>
16668 PR tree-optimization/83593
16669 * tree-ssa-strlen.c: Include tree-cfg.h.
16670 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16671 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16672 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16674 (strlen_dom_walker::before_dom_children): Call
16675 gimple_purge_dead_eh_edges. Dump tranformation with details
16677 (strlen_dom_walker::before_dom_children): Update call by adding
16678 new argument cleanup_eh.
16679 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16681 2018-01-03 Martin Liska <mliska@suse.cz>
16684 * cif-code.def (VARIADIC_THUNK): New enum value.
16685 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16688 2018-01-03 Jan Beulich <jbeulich@suse.com>
16690 * sse.md (mov<mode>_internal): Tighten condition for when to use
16691 vmovdqu<ssescalarsize> for TI and OI modes.
16693 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16695 Update copyright years.
16697 2018-01-03 Martin Liska <mliska@suse.cz>
16700 * ipa-visibility.c (function_and_variable_visibility): Skip
16701 functions with noipa attribure.
16703 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16705 * gcc.c (process_command): Update copyright notice dates.
16706 * gcov-dump.c (print_version): Ditto.
16707 * gcov.c (print_version): Ditto.
16708 * gcov-tool.c (print_version): Ditto.
16709 * gengtype.c (create_file): Ditto.
16710 * doc/cpp.texi: Bump @copying's copyright year.
16711 * doc/cppinternals.texi: Ditto.
16712 * doc/gcc.texi: Ditto.
16713 * doc/gccint.texi: Ditto.
16714 * doc/gcov.texi: Ditto.
16715 * doc/install.texi: Ditto.
16716 * doc/invoke.texi: Ditto.
16718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16720 * vector-builder.h (vector_builder::m_full_nelts): Change from
16721 unsigned int to poly_uint64.
16722 (vector_builder::full_nelts): Update prototype accordingly.
16723 (vector_builder::new_vector): Likewise.
16724 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16725 (vector_builder::operator ==): Likewise.
16726 (vector_builder::finalize): Likewise.
16727 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16728 Take the number of elements as a poly_uint64 rather than an
16730 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16731 from unsigned int to poly_uint64.
16732 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16733 (vec_perm_indices::new_vector): Likewise.
16734 (vec_perm_indices::length): Likewise.
16735 (vec_perm_indices::nelts_per_input): Likewise.
16736 (vec_perm_indices::input_nelts): Likewise.
16737 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16738 number of elements per input as a poly_uint64 rather than an
16739 unsigned int. Use the original encoding for variable-length
16740 vectors, rather than clamping each individual element.
16741 For the second and subsequent elements in each pattern,
16742 clamp the step and base before clamping their sum.
16743 (vec_perm_indices::series_p): Handle polynomial element counts.
16744 (vec_perm_indices::all_in_range_p): Likewise.
16745 (vec_perm_indices_to_tree): Likewise.
16746 (vec_perm_indices_to_rtx): Likewise.
16747 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16748 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16749 (tree_vector_builder::new_binary_operation): Handle polynomial
16750 element counts. Return false if we need to know the number
16751 of elements at compile time.
16752 * fold-const.c (fold_vec_perm): Punt if the number of elements
16753 isn't known at compile time.
16755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16757 * vec-perm-indices.h (vec_perm_builder): Change element type
16758 from HOST_WIDE_INT to poly_int64.
16759 (vec_perm_indices::element_type): Update accordingly.
16760 (vec_perm_indices::clamp): Handle polynomial element_types.
16761 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16762 (vec_perm_indices::all_in_range_p): Likewise.
16763 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16765 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16766 polynomial vec_perm_indices element types.
16767 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16768 * fold-const.c (fold_vec_perm): Likewise.
16769 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16770 * tree-vect-generic.c (lower_vec_perm): Likewise.
16771 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16772 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16773 element type to HOST_WIDE_INT.
16775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16776 Alan Hayward <alan.hayward@arm.com>
16777 David Sherwood <david.sherwood@arm.com>
16779 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16780 rather than an int. Use plus_constant.
16781 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16782 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16785 Alan Hayward <alan.hayward@arm.com>
16786 David Sherwood <david.sherwood@arm.com>
16788 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16789 a HOST_WIDE_INT to a poly_int64.
16791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16792 Alan Hayward <alan.hayward@arm.com>
16793 David Sherwood <david.sherwood@arm.com>
16795 * calls.c (load_register_parameters): Cope with polynomial
16796 mode sizes. Require a constant size for BLKmode parameters
16797 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16798 forces a parameter to be padded at the lsb end in order to
16799 fill a complete number of words, require the parameter size
16800 to be ordered wrt UNITS_PER_WORD.
16802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16803 Alan Hayward <alan.hayward@arm.com>
16804 David Sherwood <david.sherwood@arm.com>
16806 * reload1.c (spill_stack_slot_width): Change element type
16807 from unsigned int to poly_uint64_pod.
16808 (alter_reg): Treat mode sizes as polynomial.
16810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16811 Alan Hayward <alan.hayward@arm.com>
16812 David Sherwood <david.sherwood@arm.com>
16814 * reload.c (complex_word_subreg_p): New function.
16815 (reload_inner_reg_of_subreg, push_reload): Use it.
16817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16818 Alan Hayward <alan.hayward@arm.com>
16819 David Sherwood <david.sherwood@arm.com>
16821 * lra-constraints.c (process_alt_operands): Reject matched
16822 operands whose sizes aren't ordered.
16823 (match_reload): Refer to this check here.
16825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16826 Alan Hayward <alan.hayward@arm.com>
16827 David Sherwood <david.sherwood@arm.com>
16829 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16830 that the mode size is in the set {1, 2, 4, 8, 16}.
16832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16833 Alan Hayward <alan.hayward@arm.com>
16834 David Sherwood <david.sherwood@arm.com>
16836 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16837 Use plus_constant instead of gen_rtx_PLUS.
16839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16840 Alan Hayward <alan.hayward@arm.com>
16841 David Sherwood <david.sherwood@arm.com>
16843 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16844 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16845 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16846 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16847 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16848 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16849 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16850 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16851 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16852 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16854 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16855 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16856 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16857 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16858 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16859 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16860 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16861 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16862 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16863 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16865 * expr.c (emit_move_resolve_push): Treat the input and result
16866 of PUSH_ROUNDING as a poly_int64.
16867 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16868 (emit_push_insn): Likewise.
16869 * lra-eliminations.c (mark_not_eliminable): Likewise.
16870 * recog.c (push_operand): Likewise.
16871 * reload1.c (elimination_effects): Likewise.
16872 * rtlanal.c (nonzero_bits1): Likewise.
16873 * calls.c (store_one_arg): Likewise. Require the padding to be
16874 known at compile time.
16876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16877 Alan Hayward <alan.hayward@arm.com>
16878 David Sherwood <david.sherwood@arm.com>
16880 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16881 Use plus_constant instead of gen_rtx_PLUS.
16883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16884 Alan Hayward <alan.hayward@arm.com>
16885 David Sherwood <david.sherwood@arm.com>
16887 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16888 rather than an int.
16890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16891 Alan Hayward <alan.hayward@arm.com>
16892 David Sherwood <david.sherwood@arm.com>
16894 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16895 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16896 via stack temporaries. Treat the mode size as polynomial too.
16898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16899 Alan Hayward <alan.hayward@arm.com>
16900 David Sherwood <david.sherwood@arm.com>
16902 * expr.c (expand_expr_real_2): When handling conversions involving
16903 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16904 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16905 as a poly_uint64 too.
16907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16908 Alan Hayward <alan.hayward@arm.com>
16909 David Sherwood <david.sherwood@arm.com>
16911 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16914 Alan Hayward <alan.hayward@arm.com>
16915 David Sherwood <david.sherwood@arm.com>
16917 * combine.c (can_change_dest_mode): Handle polynomial
16918 REGMODE_NATURAL_SIZE.
16919 * expmed.c (store_bit_field_1): Likewise.
16920 * expr.c (store_constructor): Likewise.
16921 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16922 and polynomial REGMODE_NATURAL_SIZE.
16923 (gen_lowpart_common): Likewise.
16924 * reginfo.c (record_subregs_of_mode): Likewise.
16925 * rtlanal.c (read_modify_subreg_p): Likewise.
16927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16928 Alan Hayward <alan.hayward@arm.com>
16929 David Sherwood <david.sherwood@arm.com>
16931 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16932 numbers of elements.
16934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16935 Alan Hayward <alan.hayward@arm.com>
16936 David Sherwood <david.sherwood@arm.com>
16938 * match.pd: Cope with polynomial numbers of vector elements.
16940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16941 Alan Hayward <alan.hayward@arm.com>
16942 David Sherwood <david.sherwood@arm.com>
16944 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16945 in a POINTER_PLUS_EXPR.
16947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16948 Alan Hayward <alan.hayward@arm.com>
16949 David Sherwood <david.sherwood@arm.com>
16951 * omp-simd-clone.c (simd_clone_subparts): New function.
16952 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16953 (ipa_simd_modify_function_body): Likewise.
16955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16956 Alan Hayward <alan.hayward@arm.com>
16957 David Sherwood <david.sherwood@arm.com>
16959 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16960 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16961 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16962 (expand_vector_condition, vector_element): Likewise.
16963 (subparts_gt): New function.
16964 (get_compute_type): Use subparts_gt.
16965 (count_type_subparts): Delete.
16966 (expand_vector_operations_1): Use subparts_gt instead of
16967 count_type_subparts.
16969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16970 Alan Hayward <alan.hayward@arm.com>
16971 David Sherwood <david.sherwood@arm.com>
16973 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16974 (vect_compile_time_alias): ...this new function. Do the calculation
16975 on poly_ints rather than trees.
16976 (vect_prune_runtime_alias_test_list): Update call accordingly.
16978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16979 Alan Hayward <alan.hayward@arm.com>
16980 David Sherwood <david.sherwood@arm.com>
16982 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16984 (vect_schedule_slp_instance): Likewise.
16986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16987 Alan Hayward <alan.hayward@arm.com>
16988 David Sherwood <david.sherwood@arm.com>
16990 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16991 constant and extern definitions for variable-length vectors.
16992 (vect_get_constant_vectors): Note that the number of units
16993 is known to be constant.
16995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16996 Alan Hayward <alan.hayward@arm.com>
16997 David Sherwood <david.sherwood@arm.com>
16999 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
17000 of units as polynomial. Choose between WIDE and NARROW based
17003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17004 Alan Hayward <alan.hayward@arm.com>
17005 David Sherwood <david.sherwood@arm.com>
17007 * tree-vect-stmts.c (simd_clone_subparts): New function.
17008 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
17010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17011 Alan Hayward <alan.hayward@arm.com>
17012 David Sherwood <david.sherwood@arm.com>
17014 * tree-vect-stmts.c (vectorizable_call): Treat the number of
17015 vectors as polynomial. Use build_index_vector for
17016 IFN_GOMP_SIMD_LANE.
17018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17019 Alan Hayward <alan.hayward@arm.com>
17020 David Sherwood <david.sherwood@arm.com>
17022 * tree-vect-stmts.c (get_load_store_type): Treat the number of
17023 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
17024 for variable-length vectors.
17025 (vectorizable_mask_load_store): Treat the number of units as
17026 polynomial, asserting that it is constant if the condition has
17027 already been enforced.
17028 (vectorizable_store, vectorizable_load): Likewise.
17030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17031 Alan Hayward <alan.hayward@arm.com>
17032 David Sherwood <david.sherwood@arm.com>
17034 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
17035 of units as polynomial. Punt if we can't tell at compile time
17036 which vector contains the final result.
17038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17039 Alan Hayward <alan.hayward@arm.com>
17040 David Sherwood <david.sherwood@arm.com>
17042 * tree-vect-loop.c (vectorizable_induction): Treat the number
17043 of units as polynomial. Punt on SLP inductions. Use an integer
17044 VEC_SERIES_EXPR for variable-length integer reductions. Use a
17045 cast of such a series for variable-length floating-point
17048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17049 Alan Hayward <alan.hayward@arm.com>
17050 David Sherwood <david.sherwood@arm.com>
17052 * tree.h (build_index_vector): Declare.
17053 * tree.c (build_index_vector): New function.
17054 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
17055 of units as polynomial, forcibly converting it to a constant if
17056 vectorizable_reduction has already enforced the condition.
17057 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
17058 to create a {1,2,3,...} vector.
17059 (vectorizable_reduction): Treat the number of units as polynomial.
17060 Choose vectype_in based on the largest scalar element size rather
17061 than the smallest number of units. Enforce the restrictions
17064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17065 Alan Hayward <alan.hayward@arm.com>
17066 David Sherwood <david.sherwood@arm.com>
17068 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
17069 number of units as polynomial.
17071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17072 Alan Hayward <alan.hayward@arm.com>
17073 David Sherwood <david.sherwood@arm.com>
17075 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
17076 * target.def (autovectorize_vector_sizes): Return the vector sizes
17077 by pointer, using vector_sizes rather than a bitmask.
17078 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
17079 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
17080 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
17082 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
17083 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
17084 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
17085 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
17086 * omp-general.c (omp_max_vf): Likewise.
17087 * omp-low.c (omp_clause_aligned_alignment): Likewise.
17088 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
17089 * tree-vect-loop.c (vect_analyze_loop): Likewise.
17090 * tree-vect-slp.c (vect_slp_bb): Likewise.
17091 * doc/tm.texi: Regenerate.
17092 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
17094 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
17095 the vector size as a poly_uint64 rather than an unsigned int.
17096 (current_vector_size): Change from an unsigned int to a poly_uint64.
17097 (get_vectype_for_scalar_type): Update accordingly.
17098 * tree.h (build_truth_vector_type): Take the size and number of
17099 units as a poly_uint64 rather than an unsigned int.
17100 (build_vector_type): Add a temporary overload that takes
17101 the number of units as a poly_uint64 rather than an unsigned int.
17102 * tree.c (make_vector_type): Likewise.
17103 (build_truth_vector_type): Take the number of units as a poly_uint64
17104 rather than an unsigned int.
17106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17107 Alan Hayward <alan.hayward@arm.com>
17108 David Sherwood <david.sherwood@arm.com>
17110 * target.def (get_mask_mode): Take the number of units and length
17111 as poly_uint64s rather than unsigned ints.
17112 * targhooks.h (default_get_mask_mode): Update accordingly.
17113 * targhooks.c (default_get_mask_mode): Likewise.
17114 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17115 * doc/tm.texi: Regenerate.
17117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17118 Alan Hayward <alan.hayward@arm.com>
17119 David Sherwood <david.sherwood@arm.com>
17121 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17122 * omp-general.c (omp_max_vf): Likewise.
17123 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17124 (expand_omp_simd): Handle polynomial safelen.
17125 * omp-low.c (omplow_simd_context): Add a default constructor.
17126 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17127 (lower_rec_simd_input_clauses): Update accordingly.
17128 (lower_rec_input_clauses): Likewise.
17130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17131 Alan Hayward <alan.hayward@arm.com>
17132 David Sherwood <david.sherwood@arm.com>
17134 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17135 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17136 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17137 (vect_analyze_slp_cost): Likewise.
17138 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17139 (vect_model_load_cost): Likewise.
17141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17142 Alan Hayward <alan.hayward@arm.com>
17143 David Sherwood <david.sherwood@arm.com>
17145 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17146 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17147 from an unsigned int * to a poly_uint64_pod *.
17148 (calculate_unrolling_factor): New function.
17149 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17152 Alan Hayward <alan.hayward@arm.com>
17153 David Sherwood <david.sherwood@arm.com>
17155 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17156 from an unsigned int to a poly_uint64.
17157 (_loop_vec_info::slp_unrolling_factor): Likewise.
17158 (_loop_vec_info::vectorization_factor): Change from an int
17160 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17161 (vect_get_num_vectors): New function.
17162 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17163 (vect_get_num_copies): Use vect_get_num_vectors.
17164 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17165 to an unsigned int *.
17166 (vect_analyze_data_refs): Change min_vf from an int * to a
17168 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17169 than an unsigned HOST_WIDE_INT.
17170 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17171 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17172 to an unsigned int *.
17173 (vect_analyze_data_ref_dependences): Likewise.
17174 (vect_compute_data_ref_alignment): Handle polynomial vf.
17175 (vect_enhance_data_refs_alignment): Likewise.
17176 (vect_prune_runtime_alias_test_list): Likewise.
17177 (vect_shift_permute_load_chain): Likewise.
17178 (vect_supportable_dr_alignment): Likewise.
17179 (dependence_distance_ge_vf): Take the vectorization factor as a
17180 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17181 (vect_analyze_data_refs): Change min_vf from an int * to a
17183 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17184 vfm1 as a poly_uint64 rather than an int. Make the same change
17185 for the returned bound_scalar.
17186 (vect_gen_vector_loop_niters): Handle polynomial vf.
17187 (vect_do_peeling): Likewise. Update call to
17188 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17189 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17191 * tree-vect-loop.c (vect_determine_vectorization_factor)
17192 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17193 (vect_get_known_peeling_cost): Likewise.
17194 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17195 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17196 (vect_transform_loop): Likewise. Use the lowest possible VF when
17197 updating the upper bounds of the loop.
17198 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17199 rather than an int.
17200 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17201 polynomial unroll factors.
17202 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17203 (vect_make_slp_decision): Likewise.
17204 (vect_supported_load_permutation_p): Likewise, and polynomial
17206 (vect_analyze_slp_cost): Handle polynomial vf.
17207 (vect_slp_analyze_node_operations): Likewise.
17208 (vect_slp_analyze_bb_1): Likewise.
17209 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17210 than an unsigned HOST_WIDE_INT.
17211 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17212 (vectorizable_load): Handle polynomial vf.
17213 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17215 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17218 Alan Hayward <alan.hayward@arm.com>
17219 David Sherwood <david.sherwood@arm.com>
17221 * match.pd: Handle bit operations involving three constants
17222 and try to fold one pair.
17224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17226 * tree-vect-loop-manip.c: Include gimple-fold.h.
17227 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17228 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17229 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17230 Add a path that uses a step of VF instead of 1, but disable it
17232 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17233 and niters_no_overflow parameters. Update calls to
17234 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17235 Create a new SSA name if the latter choses to use a ste other
17236 than zero, and return it via niters_vector_mult_vf_var.
17237 * tree-vect-loop.c (vect_transform_loop): Update calls to
17238 vect_do_peeling, vect_gen_vector_loop_niters and
17239 slpeel_make_loop_iterate_ntimes.
17240 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17241 (vect_gen_vector_loop_niters): Update declarations after above changes.
17243 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17245 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17246 128-bit round to integer instructions.
17247 (ceil<mode>2): Likewise.
17248 (btrunc<mode>2): Likewise.
17249 (round<mode>2): Likewise.
17251 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17253 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17254 unaligned VSX load/store on P8/P9.
17255 (expand_block_clear): Allow the use of unaligned VSX
17256 load/store on P8/P9.
17258 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17260 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17262 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17263 swap associated with both a load and a store.
17265 2018-01-02 Andrew Waterman <andrew@sifive.com>
17267 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17268 * config/riscv/riscv.md (clear_cache): Use it.
17270 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17272 * web.c: Remove out-of-date comment.
17274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17276 * expr.c (fixup_args_size_notes): Check that any existing
17277 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17278 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17279 (emit_single_push_insn): ...here.
17281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17283 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17284 (const_vector_encoded_nelts): New function.
17285 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17286 (const_vector_int_elt, const_vector_elt): Declare.
17287 * emit-rtl.c (const_vector_int_elt_1): New function.
17288 (const_vector_elt): Likewise.
17289 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17290 of CONST_VECTOR_ELT.
17292 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17294 * expr.c: Include rtx-vector-builder.h.
17295 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17296 directly on the tree encoding.
17297 (const_vector_from_tree): Likewise.
17298 * optabs.c: Include rtx-vector-builder.h.
17299 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17300 sequence of "u" values.
17301 * vec-perm-indices.c: Include rtx-vector-builder.h.
17302 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17303 directly on the vec_perm_indices encoding.
17305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17307 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17308 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17309 * rtx-vector-builder.h: New file.
17310 * rtx-vector-builder.c: Likewise.
17311 * rtl.h (rtx_def::u2): Add a const_vector field.
17312 (CONST_VECTOR_NPATTERNS): New macro.
17313 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17314 (CONST_VECTOR_DUPLICATE_P): Likewise.
17315 (CONST_VECTOR_STEPPED_P): Likewise.
17316 (CONST_VECTOR_ENCODED_ELT): Likewise.
17317 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17318 (unwrap_const_vec_duplicate): Likewise.
17319 (const_vec_series_p): Check for a non-duplicated vector encoding.
17320 Say that the function only returns true for integer vectors.
17321 * emit-rtl.c: Include rtx-vector-builder.h.
17322 (gen_const_vec_duplicate_1): Delete.
17323 (gen_const_vector): Call gen_const_vec_duplicate instead of
17324 gen_const_vec_duplicate_1.
17325 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17326 (gen_const_vec_duplicate): Use rtx_vector_builder.
17327 (gen_const_vec_series): Likewise.
17328 (gen_rtx_CONST_VECTOR): Likewise.
17329 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17330 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17331 Build a new vector rather than modifying a CONST_VECTOR in-place.
17332 (handle_special_swappables): Update call accordingly.
17333 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17334 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17335 Build a new vector rather than modifying a CONST_VECTOR in-place.
17336 (handle_special_swappables): Update call accordingly.
17338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17340 * simplify-rtx.c (simplify_const_binary_operation): Use
17341 CONST_VECTOR_ELT instead of XVECEXP.
17343 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17345 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17346 the selector elements to be different from the data elements
17347 if the selector is a VECTOR_CST.
17348 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17349 ssizetype for the selector.
17351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17353 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17354 before testing each element individually.
17355 * tree-vect-generic.c (lower_vec_perm): Likewise.
17357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17359 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17360 * selftest-run-tests.c (selftest::run_tests): Call it.
17361 * vector-builder.h (vector_builder::operator ==): New function.
17362 (vector_builder::operator !=): Likewise.
17363 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17364 (vec_perm_indices::all_from_input_p): New function.
17365 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17366 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17367 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17368 instead of reading the VECTOR_CST directly. Detect whether both
17369 vector inputs are the same before constructing the vec_perm_indices,
17370 and update the number of inputs argument accordingly. Use the
17371 utility functions added above. Only construct sel2 if we need to.
17373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17375 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17376 the broadcast of the low byte.
17377 (expand_mult_highpart): Use an explicit encoding for the permutes.
17378 * optabs-query.c (can_mult_highpart_p): Likewise.
17379 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17380 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17381 (vectorizable_bswap): Likewise.
17382 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17383 explicit encoding for the power-of-2 permutes.
17384 (vect_permute_store_chain): Likewise.
17385 (vect_grouped_load_supported): Likewise.
17386 (vect_permute_load_chain): Likewise.
17388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17390 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17391 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17392 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17393 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17394 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17395 (vect_gen_perm_mask_any): Likewise.
17397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17399 * int-vector-builder.h: New file.
17400 * vec-perm-indices.h: Include int-vector-builder.h.
17401 (vec_perm_indices): Redefine as an int_vector_builder.
17402 (auto_vec_perm_indices): Delete.
17403 (vec_perm_builder): Redefine as a stand-alone class.
17404 (vec_perm_indices::vec_perm_indices): New function.
17405 (vec_perm_indices::clamp): Likewise.
17406 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17407 (vec_perm_indices::new_vector): New function.
17408 (vec_perm_indices::new_expanded_vector): Update for new
17409 vec_perm_indices class.
17410 (vec_perm_indices::rotate_inputs): New function.
17411 (vec_perm_indices::all_in_range_p): Operate directly on the
17412 encoded form, without computing elided elements.
17413 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17414 encoding. Update for new vec_perm_indices class.
17415 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17416 the given vec_perm_builder.
17417 (expand_vec_perm_var): Update vec_perm_builder constructor.
17418 (expand_mult_highpart): Use vec_perm_builder instead of
17419 auto_vec_perm_indices.
17420 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17421 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17422 or double series encoding as appropriate.
17423 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17424 vec_perm_indices instead of auto_vec_perm_indices.
17425 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17426 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17427 (vect_permute_store_chain): Likewise.
17428 (vect_grouped_load_supported): Likewise.
17429 (vect_permute_load_chain): Likewise.
17430 (vect_shift_permute_load_chain): Likewise.
17431 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17432 (vect_transform_slp_perm_load): Likewise.
17433 (vect_schedule_slp_instance): Likewise.
17434 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17435 (vectorizable_mask_load_store): Likewise.
17436 (vectorizable_bswap): Likewise.
17437 (vectorizable_store): Likewise.
17438 (vectorizable_load): Likewise.
17439 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17440 vec_perm_indices instead of auto_vec_perm_indices. Use
17441 tree_to_vec_perm_builder to read the vector from a tree.
17442 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17443 vec_perm_builder instead of a vec_perm_indices.
17444 (have_whole_vector_shift): Use vec_perm_builder and
17445 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17446 truncation to calc_vec_perm_mask_for_shift.
17447 (vect_create_epilog_for_reduction): Likewise.
17448 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17449 from auto_vec_perm_indices to vec_perm_indices.
17450 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17451 instead of changing individual elements.
17452 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17453 the vector in d.perm.
17454 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17455 from auto_vec_perm_indices to vec_perm_indices.
17456 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17457 instead of changing individual elements.
17458 (arm_vectorize_vec_perm_const): Use new_vector to install
17459 the vector in d.perm.
17460 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17461 Update vec_perm_builder constructor.
17462 (rs6000_expand_interleave): Likewise.
17463 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17464 (rs6000_expand_interleave): Likewise.
17466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17468 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17469 to qimode could truncate the indices.
17470 * optabs.c (expand_vec_perm_var): Likewise.
17472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17474 * Makefile.in (OBJS): Add vec-perm-indices.o.
17475 * vec-perm-indices.h: New file.
17476 * vec-perm-indices.c: Likewise.
17477 * target.h (vec_perm_indices): Replace with a forward class
17479 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17480 * optabs.h: Include vec-perm-indices.h.
17481 (expand_vec_perm): Delete.
17482 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17483 (expand_vec_perm_const): Declare.
17484 * target.def (vec_perm_const_ok): Replace with...
17485 (vec_perm_const): ...this new hook.
17486 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17487 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17488 * doc/tm.texi: Regenerate.
17489 * optabs.def (vec_perm_const): Delete.
17490 * doc/md.texi (vec_perm_const): Likewise.
17491 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17492 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17493 expand_vec_perm for constant permutation vectors. Assert that
17494 the mode of variable permutation vectors is the integer equivalent
17495 of the mode that is being permuted.
17496 * optabs-query.h (selector_fits_mode_p): Declare.
17497 * optabs-query.c: Include vec-perm-indices.h.
17498 (selector_fits_mode_p): New function.
17499 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17500 is defined, instead of checking whether the vec_perm_const_optab
17501 exists. Use targetm.vectorize.vec_perm_const instead of
17502 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17503 fit in the vector mode before using a variable permute.
17504 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17505 vec_perm_indices instead of an rtx.
17506 (expand_vec_perm): Replace with...
17507 (expand_vec_perm_const): ...this new function. Take the selector
17508 as a vec_perm_indices rather than an rtx. Also take the mode of
17509 the selector. Update call to shift_amt_for_vec_perm_mask.
17510 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17511 Use vec_perm_indices::new_expanded_vector to expand the original
17512 selector into bytes. Check whether the indices fit in the vector
17513 mode before using a variable permute.
17514 (expand_vec_perm_var): Make global.
17515 (expand_mult_highpart): Use expand_vec_perm_const.
17516 * fold-const.c: Includes vec-perm-indices.h.
17517 * tree-ssa-forwprop.c: Likewise.
17518 * tree-vect-data-refs.c: Likewise.
17519 * tree-vect-generic.c: Likewise.
17520 * tree-vect-loop.c: Likewise.
17521 * tree-vect-slp.c: Likewise.
17522 * tree-vect-stmts.c: Likewise.
17523 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17525 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17526 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17527 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17528 (aarch64_vectorize_vec_perm_const): ...this new function.
17529 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17530 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17531 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17532 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17533 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17534 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17535 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17537 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17538 check for NEON modes.
17539 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17540 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17541 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17542 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17544 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17545 the old VEC_PERM_CONST conditions.
17546 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17547 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17548 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17549 (ia64_vectorize_vec_perm_const_ok): Merge into...
17550 (ia64_vectorize_vec_perm_const): ...this new function.
17551 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17552 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17553 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17554 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17555 * config/mips/mips.c (mips_expand_vec_perm_const)
17556 (mips_vectorize_vec_perm_const_ok): Merge into...
17557 (mips_vectorize_vec_perm_const): ...this new function.
17558 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17559 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17560 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17561 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17562 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17563 (rs6000_expand_vec_perm_const): Delete.
17564 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17566 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17567 (altivec_expand_vec_perm_const_le): Take each operand individually.
17568 Operate on constant selectors rather than rtxes.
17569 (altivec_expand_vec_perm_const): Likewise. Update call to
17570 altivec_expand_vec_perm_const_le.
17571 (rs6000_expand_vec_perm_const): Delete.
17572 (rs6000_vectorize_vec_perm_const_ok): Delete.
17573 (rs6000_vectorize_vec_perm_const): New function.
17574 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17575 an element count and rtx array.
17576 (rs6000_expand_extract_even): Update call accordingly.
17577 (rs6000_expand_interleave): Likewise.
17578 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17579 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17580 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17581 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17582 (rs6000_expand_vec_perm_const): Delete.
17583 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17584 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17585 (altivec_expand_vec_perm_const_le): Take each operand individually.
17586 Operate on constant selectors rather than rtxes.
17587 (altivec_expand_vec_perm_const): Likewise. Update call to
17588 altivec_expand_vec_perm_const_le.
17589 (rs6000_expand_vec_perm_const): Delete.
17590 (rs6000_vectorize_vec_perm_const_ok): Delete.
17591 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17592 reference to the SPE evmerge intructions.
17593 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17594 an element count and rtx array.
17595 (rs6000_expand_extract_even): Update call accordingly.
17596 (rs6000_expand_interleave): Likewise.
17597 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17598 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17604 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17605 vector mode and that that mode matches the mode of the data
17607 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17608 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17609 directly using expand_vec_perm_1 when forcing selectors into
17611 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17615 * optabs-query.h (can_vec_perm_p): Delete.
17616 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17617 * optabs-query.c (can_vec_perm_p): Split into...
17618 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17619 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17620 particular selector is valid.
17621 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17622 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17623 (vect_grouped_load_supported): Likewise.
17624 (vect_shift_permute_load_chain): Likewise.
17625 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17626 (vect_transform_slp_perm_load): Likewise.
17627 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17628 (vectorizable_bswap): Likewise.
17629 (vect_gen_perm_mask_checked): Likewise.
17630 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17631 implementations of variable permutation vectors into account
17632 when deciding which selector to use.
17633 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17634 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17635 with a false third argument.
17636 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17637 to test whether the constant selector is valid and can_vec_perm_var_p
17638 to test whether a variable selector is valid.
17640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17642 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17643 * optabs-query.c (can_vec_perm_p): Likewise.
17644 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17645 instead of vec_perm_indices.
17646 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17647 (vect_gen_perm_mask_checked): Likewise,
17648 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17649 (vect_gen_perm_mask_checked): Likewise,
17651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17653 * optabs-query.h (qimode_for_vec_perm): Declare.
17654 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17655 (qimode_for_vec_perm): ...this new function.
17656 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17658 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17660 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17661 does not have a conditional at the top.
17663 2018-01-02 Richard Biener <rguenther@suse.de>
17665 * ipa-inline.c (big_speedup_p): Fix expression.
17667 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17670 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17673 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17677 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17678 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17679 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17680 cond_taken_branch_cost 3->4.
17682 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17684 PR tree-optimization/83581
17685 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17686 TODO_cleanup_cfg if any changes have been made.
17688 PR middle-end/83608
17689 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17690 convert_modes if target mode has the right side, but different mode
17693 PR middle-end/83609
17694 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17695 last argument when extracting from CONCAT. If either from_real or
17696 from_imag is NULL, use expansion through memory. If result is not
17697 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17698 the parts directly to inner mode, if even that fails, use expansion
17701 PR middle-end/83623
17702 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17703 check for bswap in mode rather than HImode and use that in expand_unop
17706 Copyright (C) 2018 Free Software Foundation, Inc.
17708 Copying and distribution of this file, with or without modification,
17709 are permitted in any medium without royalty provided the copyright
17710 notice and this notice are preserved.