1 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4 behavior of vec_pack (vector double, vector double) to match
5 behavior of vec_float2 (vector double, vector double).
7 2018-06-22 Olivier Hainque <hainque@adacore.com>
9 * gimplify.c (gimplify_function_tree): Prevent creation
10 of a trampoline for the address of the current function
11 passed to entry/exit instrumentation hooks.
13 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
16 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
19 2018-06-22 Martin Liska <mliska@suse.cz>
21 PR tree-optimization/86263
22 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
23 Bail out if is_enabled is false.
24 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
26 (jump_table_cluster::is_enabled): New function.
28 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
30 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
31 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
32 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
33 (lto_input_ts_binfo_tree_pointers): Likewise.
34 * tree-streamer-out.c (streamer_write_tree_bitfields,
35 write_ts_binfo_tree_pointers): Likewise.
36 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
38 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
40 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
42 2018-06-22 Martin Liska <mliska@suse.cz>
44 * symbol-summary.h (get): Make it pure and inline move
45 functionality from ::get function.
46 (get): Remove and inline into ::get and ::get_create.
47 (get_create): Move code from ::get function.
49 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
52 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
53 -x assembler-with-cpp.
55 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
57 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
58 _FILE_OFFSET_BITS=64 for C++.
60 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
62 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
63 conversion insn that shows up when pr85657-3.c is compiled using
64 IEEE 128-bit long double.
65 (neg<mode>2_internal): Use the correct mode to check whether the
67 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
68 multiply and divide external functions from being created more
71 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
73 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
75 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
76 the edge can be forwarded.
77 (cfg_layout_merge_blocks): Likewise.
79 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
81 * except.c (finish_eh_generation): Commit edge insertions only after
82 the EH edges have been redirected from post-landing to landing pads.
84 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
86 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
87 create_tmp_var_for to create the FRAME decl.
88 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
90 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
92 * tree-inline.c (copy_edges_for_bb): Minor tweak.
93 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
94 debug statement when resetting its value.
95 (expand_call_inline): Copy the locus of the call onto the assignment
96 of the return value, if any. Use local variable in more cases.
98 2018-06-21 Martin Liska <mliska@suse.cz>
100 * ipa-pure-const.c (propagate_nothrow): Use
101 funct_state_summaries->get.
102 (dump_malloc_lattice): Likewise.
103 (propagate_malloc): Likewise.
105 2018-06-21 Richard Biener <rguenther@suse.de>
107 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
108 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
109 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
111 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
112 BLOCK_ABSTRACT_ORIGIN unconditionally.
114 2018-06-21 David Malcolm <dmalcolm@redhat.com>
116 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
118 * ipa-reference.c (ipa_reference_c_finalize): Delete
119 ipa_ref_opt_sum_summaries and set it to NULL.
121 2018-06-21 Tom de Vries <tdevries@suse.de>
123 PR tree-optimization/85859
124 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
125 test with comment from bb_no_side_effects_p.
127 2018-06-21 Richard Biener <rguenther@suse.de>
129 PR tree-optimization/86232
130 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
131 max for constant niter.
133 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
135 * config/aarch64/aarch64-simd.md
136 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
138 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
140 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
141 Make opernads of the unspec commutative.
143 2018-06-21 Richard Biener <rguenther@suse.de>
145 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
146 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
147 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
148 (vect_analyze_data_ref_dependence): Re-order checks to deal with
150 (vect_record_base_alignments): Do not record base alignment
151 for gathers or scatters.
152 (vect_compute_data_ref_alignment): Drop return value that is always
153 true. Bail out early for gathers or scatters.
154 (vect_enhance_data_refs_alignment): Bail out early for gathers
156 (vect_find_same_alignment_drs): Likewise.
157 (vect_analyze_data_refs_alignment): Remove dead code.
158 (vect_slp_analyze_and_verify_node_alignment): Likewise.
159 (vect_analyze_data_refs): For possible gathers or scatters do
160 not create an alternate DR, just check their possible validity
161 and mark them. Adjust DECL_NONALIASED handling to not rely
163 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
164 update inits of gathers or scatters.
165 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
166 Also copy gather/scatter flag to pattern vinfo.
168 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
170 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
171 behavior of vec_packsu (vector unsigned long long, vector unsigned
172 long long) to match behavior of vec_packs with same signature.
174 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
175 Thomas Schwinge <thomas@codesourcery.com>
176 Cesar Philippidis <cesar@codesourcery.com>
178 * gimplify.c (gimplify_scan_omp_clauses): Add support for
179 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
180 (gimplify_adjust_omp_clauses): Likewise.
181 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
182 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
183 (gimplify_omp_target_update): Update handling of acc update and
185 * omp-low.c (install_var_field): Remove unused parameter
186 base_pointers_restrict.
187 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
188 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
190 (omp_target_base_pointers_restrict_p): Delete.
191 (scan_omp_target): Update call to scan_sharing_clauses.
192 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
194 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
195 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
196 (convert_local_omp_clauses): Likewise.
197 * tree-pretty-print.c (dump_omp_clause): Likewise.
198 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
200 (omp_clause_code_name): Likewise.
202 2018-06-20 Jakub Jelinek <jakub@redhat.com>
205 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
208 PR tree-optimization/86231
209 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
210 anti-range don't overwrite *vr0min before using it to compute *vr0max.
212 2018-06-20 Tom de Vries <tdevries@suse.de>
214 PR tree-optimization/86097
215 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
216 iv type if signedness of iv type is not the same as that of *nit.
218 2018-06-20 Jakub Jelinek <jakub@redhat.com>
220 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
221 EDGE_EH edges, verify they are all EDGE_EH.
223 2018-06-20 Maya Rashish <coypu@sdf.org>
225 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
227 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
229 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
230 * config/aarch64/aarch64.c (xgene1_tunings): Add
231 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
232 (aarch64_mode_valid_for_sched_fusion_p):
234 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
235 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
237 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
239 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
240 * config/aarch64/iterators.md (VQ2): New mode iterator.
242 2018-06-20 Martin Liska <mliska@suse.cz>
244 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
245 Change default ratio from 10 to 8.
247 2018-06-20 Martin Liska <mliska@suse.cz>
249 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
251 (bit_test_cluster::find_bit_tests): Likewise.
252 (switch_decision_tree::analyze_switch_statement): Find clusters.
253 * tree-switch-conversion.h (struct jump_table_cluster): Document
256 2018-06-20 Martin Liska <mliska@suse.cz>
258 * tree-switch-conversion.c (switch_conversion::collect):
259 Record m_uniq property.
260 (switch_conversion::expand): Bail out for special conditions.
261 (group_cluster::~group_cluster): New.
262 (group_cluster::group_cluster): Likewise.
263 (group_cluster::dump): Likewise.
264 (jump_table_cluster::emit): New.
265 (switch_decision_tree::fix_phi_operands_for_edges): New.
266 (struct case_node): Remove struct.
267 (jump_table_cluster::can_be_handled): New.
268 (case_values_threshold): Moved to header.
269 (reset_out_edges_aux): Likewise.
270 (jump_table_cluster::is_beneficial): New.
271 (bit_test_cluster::can_be_handled): Likewise.
272 (add_case_node): Remove.
273 (bit_test_cluster::is_beneficial): New.
274 (case_bit_test::cmp): New.
275 (bit_test_cluster::emit): New.
276 (expand_switch_as_decision_tree_p): Remove.
277 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
278 (fix_phi_operands_for_edge): Likewise.
279 (switch_decision_tree::analyze_switch_statement): New.
280 (compute_cases_per_edge): Move ...
281 (switch_decision_tree::compute_cases_per_edge): ... here.
282 (try_switch_expansion): Likewise.
283 (switch_decision_tree::try_switch_expansion): Likewise.
284 (record_phi_operand_mapping): Likewise.
285 (switch_decision_tree::record_phi_operand_mapping): Likewise.
286 (emit_case_decision_tree): Likewise.
287 (switch_decision_tree::emit): Likewise.
288 (balance_case_nodes): Likewise.
289 (switch_decision_tree::balance_case_nodes): Likewise.
290 (dump_case_nodes): Likewise.
291 (switch_decision_tree::dump_case_nodes): Likewise.
292 (emit_jump): Likewise.
293 (switch_decision_tree::emit_jump): Likewise.
294 (emit_cmp_and_jump_insns): Likewise.
295 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
296 (emit_case_nodes): Likewise.
297 (switch_decision_tree::emit_case_nodes): Likewise.
298 (conditional_probability): Remove.
299 * tree-switch-conversion.h (enum cluster_type): New.
301 (struct cluster): Likewise.
302 (cluster::cluster): Likewise.
303 (struct simple_cluster): Likewise.
304 (simple_cluster::simple_cluster): Likewise.
305 (struct group_cluster): Likewise.
306 (struct jump_table_cluster): Likewise.
307 (struct bit_test_cluster): Likewise.
308 (struct min_cluster_item): Likewise.
309 (struct case_tree_node): Likewise.
310 (case_tree_node::case_tree_node): Likewise.
311 (jump_table_cluster::case_values_threshold): Likewise.
312 (struct case_bit_test): Likewise.
313 (struct switch_decision_tree): Likewise.
314 (struct switch_conversion): Likewise.
315 (switch_decision_tree::reset_out_edges_aux): Likewise.
317 2018-06-20 Martin Liska <mliska@suse.cz>
319 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
320 (hoist_edge_and_branch_if_true): Likewise.
321 (expand_switch_using_bit_tests_p): Likewise.
322 (struct case_bit_test): Likewise.
323 (case_bit_test_cmp): Likewise.
324 (emit_case_bit_tests): Likewise.
325 (switch_conversion::switch_conversion): New class.
326 (struct switch_conv_info): Remove old struct.
327 (collect_switch_conv_info): More to ...
328 (switch_conversion::collect): ... this.
329 (check_range): Likewise.
330 (switch_conversion::check_range): Likewise.
331 (check_all_empty_except_final): Likewise.
332 (switch_conversion::check_all_empty_except_final): Likewise.
333 (check_final_bb): Likewise.
334 (switch_conversion::check_final_bb): Likewise.
335 (create_temp_arrays): Likewise.
336 (switch_conversion::create_temp_arrays): Likewise.
337 (free_temp_arrays): Likewise.
338 (gather_default_values): Likewise.
339 (switch_conversion::gather_default_values): Likewise.
340 (build_constructors): Likewise.
341 (switch_conversion::build_constructors): Likewise.
342 (constructor_contains_same_values_p): Likewise.
343 (switch_conversion::contains_same_values_p): Likewise.
344 (array_value_type): Likewise.
345 (switch_conversion::array_value_type): Likewise.
346 (build_one_array): Likewise.
347 (switch_conversion::build_one_array): Likewise.
348 (build_arrays): Likewise.
349 (switch_conversion::build_arrays): Likewise.
350 (gen_def_assigns): Likewise.
351 (switch_conversion::gen_def_assigns): Likewise.
352 (prune_bbs): Likewise.
353 (switch_conversion::prune_bbs): Likewise.
354 (fix_phi_nodes): Likewise.
355 (switch_conversion::fix_phi_nodes): Likewise.
356 (gen_inbound_check): Likewise.
357 (switch_conversion::gen_inbound_check): Likewise.
358 (process_switch): Use the newly created class.
359 (switch_conversion::expand): New.
360 (switch_conversion::~switch_conversion): New.
361 * tree-switch-conversion.h: New file.
363 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
365 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
366 tree-vect-patterns.c.
367 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
368 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
369 (vect_recog_sad_pattern): Likewise.
370 (vect_recog_widen_sum_pattern): Likewise.
371 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
372 (vect_recog_widen_shift_pattern): Remove the type_in argument.
373 (vect_recog_rotate_pattern): Likewise.
374 (vect_recog_mult_pattern): Likewise.
375 (vect_recog_vector_vector_shift_pattern): Likewise.
376 (vect_recog_divmod_pattern): Likewise.
377 (vect_recog_mixed_size_cond_pattern): Likewise.
378 (vect_recog_bool_pattern): Likewise.
379 (vect_recog_mask_conversion_pattern): Likewise.
380 (vect_try_gather_scatter_pattern): Likewise.
381 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
382 (vect_recog_over_widening_pattern): Likewise.
383 (vect_recog_gather_scatter_pattern): Likewise.
384 (vect_recog_func_ptr): Move from tree-vectorizer.h
385 (vect_vect_recog_func_ptrs): Move further down the file.
386 (vect_recog_func): Likewise. Remove the third argument.
387 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
388 (vect_pattern_recog_1): Expect the pattern function to do any
389 necessary target tests. Also expect it to provide a vector type.
390 Remove the type_in handling.
392 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
394 * tree-vect-patterns.c (vect_pattern_detected): New function.
395 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
396 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
397 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
398 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
399 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
400 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
401 (vect_recog_mask_conversion_pattern)
402 (vect_try_gather_scatter_pattern): Likewise.
404 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
406 * tree-vect-patterns.c (vect_get_internal_def): New function.
407 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
408 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
409 (search_type_for_mask_1): Use it.
411 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
413 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
414 redundant WIDEN_SUM_EXPR handling.
415 (vect_recog_sad_pattern): Likewise.
417 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
419 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
420 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
421 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
422 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
423 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
425 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
427 * tree-vect-stmts.c (vectorizable_call): Make sure that we
428 use the stmt_vec_info of the original bb statement for the
429 new zero assignment, even if the call is part of a pattern.
431 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
433 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
434 that the sequence is attached to the original statement rather
435 than the pattern statement.
436 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
437 PATTERN_DEF_SEQ from the original statement rather than
438 the main pattern statement.
439 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
440 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
441 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
443 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
445 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
446 definition statements before the early exit for statements that aren't
448 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
450 (vect_transform_loop): ...here. Process pattern definition
451 statements without first checking whether the main pattern
452 statement is live or relevant.
454 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
456 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
457 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
459 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
461 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
462 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
463 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
464 (expand_block_compare): Change select_block_compare_mode call.
465 (expand_strncmp_align_check): Use new functions, fix comment.
466 (emit_final_str_compare_gpr): New function.
467 (expand_strn_compare): Refactor and clean up code.
468 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
470 2018-06-19 Tony Reix <tony.reix@atos.com>
471 Damien Bergamini <damien.bergamini@atos.com>
472 David Edelsohn <dje.gcc@gmail.com>
474 * collect2.c (static_obj): New variable.
475 (static_libs): New variable.
476 (is_in_list): Uncomment declaration.
477 (main): Track AIX libraries linked statically.
478 (is_in_list): Uncomment definition.
479 (scan_prog_file): Don't add AIX shared libraries initializer
480 to constructor list if linking statically.
482 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
484 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
486 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
488 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
490 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
493 2018-06-19 Martin Liska <mliska@suse.cz>
495 * config/i386/i386.c (ix86_can_inline_p): Do not use
496 ipa_fn_summaries::get_create.
497 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
499 (devirtualization_time_bonus): Likewise.
500 (ipcp_propagate_stage): Likewise.
501 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
502 (edge_set_predicate): Likewise.
503 (evaluate_conditions_for_known_args): Likewise.
504 (evaluate_properties_for_edge): Likewise.
505 (ipa_call_summary::reset): Tranform to ...
506 (ipa_call_summary::~ipa_call_summary): ... this.
507 (ipa_fn_summary::reset): Transform to ...
508 (ipa_fn_summary::~ipa_fn_summary): ... this.
509 (ipa_fn_summary_t::remove): Rename to ...
510 (ipa_fn_summary_t::remove_callees): ... this.
511 (ipa_fn_summary_t::duplicate): Use placement new
512 instead of memory copy.
513 (ipa_call_summary_t::duplicate): Likewise.
514 (ipa_call_summary_t::remove): Remove.
515 (dump_ipa_call_summary): Change get_create to get.
516 (ipa_dump_fn_summary): Dump only when summary exists.
517 (analyze_function_body): Use symbol_summary::get instead
519 (compute_fn_summary): Likewise.
520 (estimate_edge_devirt_benefit): Likewise.
521 (estimate_edge_size_and_time): Likewise.
522 (inline_update_callee_summaries): Likewise.
523 (remap_edge_change_prob): Likewise.
524 (remap_edge_summaries): Likewise.
525 (ipa_merge_fn_summary_after_inlining): Likewise.
526 (write_ipa_call_summary): Likewise.
527 (ipa_fn_summary_write): Likewise.
528 (ipa_free_fn_summary): Likewise.
529 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
530 (struct ipa_call_summary): Likewise.
531 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
533 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
534 (estimate_size_after_inlining): Likewise.
535 (estimate_growth): Likewise.
536 (growth_likely_positive): Likewise.
537 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
538 (inline_call): Likewise.
539 * ipa-inline.c (caller_growth_limits): Likewise.
540 (can_inline_edge_p): Likewise.
541 (can_inline_edge_by_limits_p): Likewise.
542 (compute_uninlined_call_time): Likewise.
543 (compute_inlined_call_time): Likewise.
544 (want_inline_small_function_p): Likewise.
545 (edge_badness): Likewise.
546 (update_caller_keys): Likewise.
547 (update_callee_keys): Likewise.
548 (inline_small_functions): Likewise.
549 (inline_to_all_callers_1): Likewise.
550 (dump_overall_stats): Likewise.
551 (early_inline_small_functions): Likewise.
552 (early_inliner): Likewise.
553 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
554 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
555 * ipa-pure-const.c (malloc_candidate_p): Likewise.
556 * ipa-split.c (execute_split_functions): Likewise.
557 * symbol-summary.h: Likewise.
558 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
560 2018-06-19 Richard Biener <rguenther@suse.de>
562 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
563 (vectorize_loops): ... here. Fix dbgcnt handling.
564 (try_vectorize_loop): Wrap try_vectorize_loop_1.
566 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
569 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
570 ieee128 argument takes up only one (vector) register, not two (floating
573 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
575 * gimplify.c (gimplify_init_constructor): Really never clear for an
576 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
578 2018-06-19 Richard Biener <rguenther@suse.de>
580 PR tree-optimization/86179
581 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
582 after failed recognition.
584 2018-06-18 Martin Sebor <msebor@redhat.com>
587 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
588 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
589 Handle integer subtraction.
590 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
591 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
593 2018-06-18 David Malcolm <dmalcolm@redhat.com>
595 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
596 param from rtx to rtx_insn *.
597 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
599 (frv_ifcvt_modify_insn): Likwise.
600 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
601 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
602 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
603 as_a <rtx_insn *> cast to local "unprotected_region" once
604 it's been established that it's not NULL or pc_rtx.
605 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
606 param "sethi" from rtx to rtx_insn *.
607 (nds32_group_float_insns): Likewise for param "insn".
608 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
610 (vax_output_int_subtract): Likewise.
611 * config/vax/vax.c (vax_output_int_add): Likewise for param
613 (vax_output_int_subtract): Likewise.
614 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
615 (emit_pattern_after): Likewise for param "after".
616 (emit_insn_after): Likewise.
617 (emit_jump_insn_after): Likewise.
618 (emit_call_insn_after): Likewise.
619 (emit_debug_insn_after): Likewise.
620 (emit_pattern_before): Likewise for param "before".
621 (emit_insn_before): Likewise.
622 (emit_jump_insn_before): Likewise.
623 * final.c (get_insn_template): Likewise for param "insn", removing
625 * output.h (get_insn_template): Likewise for 2nd param.
626 * rtl.h (emit_insn_before): Likewise.
627 (emit_jump_insn_before): Likewise.
628 (emit_debug_insn_before_noloc): Likewise.
629 (emit_insn_after): Likewise.
630 (emit_jump_insn_after): Likewise.
631 (emit_call_insn_after): Likewise.
632 (emit_debug_insn_after): Likewise.
633 (set_insn_deleted): Likewise for param.
635 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
638 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
639 floating point modes, so that IFmode is numerically greater than
640 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
641 to declare the ordering. This prevents IFmode from being
642 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
643 machine. Include rs6000-modes.h to share the fractional values
644 between genmodes* and the rest of the compiler.
648 * config/rs6000/rs6000-modes.h: New file.
649 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
650 meaning of rs6000_long_double_size so that 126..128 selects an
651 appropriate 128-bit floating point type.
652 (rs6000_option_override_internal): Likewise.
653 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
654 (TARGET_LONG_DOUBLE_128): Change the meaning of
655 rs6000_long_double_size so that 126..128 selects an appropriate
656 128-bit floating point type.
657 (LONG_DOUBLE_TYPE_SIZE): Update comment.
658 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
659 source and destination to match the standard usage.
660 (truncifkf2): Likewise.
661 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
662 ISA 2.07 to use an explicit clobber, instead of passing in a
664 (copysign<mode>3_soft): Likewise.
666 2018-06-18 David Malcolm <dmalcolm@redhat.com>
668 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
669 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
670 (vect_slp_analyze_instance_dependence): Likewise.
671 (vect_enhance_data_refs_alignment): Likewise.
672 (vect_analyze_data_refs_alignment): Likewise.
673 (vect_slp_analyze_and_verify_instance_alignment
674 (vect_analyze_data_ref_accesses): Likewise.
675 (vect_prune_runtime_alias_test_list): Likewise.
676 (vect_analyze_data_refs): Likewise.
677 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
678 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
679 (vect_analyze_scalar_cycles_1): Likewise.
680 (vect_get_loop_niters): Likewise.
681 (vect_analyze_loop_form_1): Likewise.
682 (vect_update_vf_for_slp): Likewise.
683 (vect_analyze_loop_operations): Likewise.
684 (vect_analyze_loop): Likewise.
685 (vectorizable_induction): Likewise.
686 (vect_transform_loop): Likewise.
687 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
688 * tree-vect-slp.c (vect_analyze_slp): Likewise.
689 (vect_make_slp_decision): Likewise.
690 (vect_detect_hybrid_slp): Likewise.
691 (vect_slp_analyze_operations): Likewise.
692 (vect_slp_bb): Likewise.
693 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
694 (vectorizable_bswap): Likewise.
695 (vectorizable_call): Likewise.
696 (vectorizable_simd_clone_call): Likewise.
697 (vectorizable_conversion): Likewise.
698 (vectorizable_assignment): Likewise.
699 (vectorizable_shift): Likewise.
700 (vectorizable_operation): Likewise.
701 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
703 2018-06-18 Martin Sebor <msebor@redhat.com>
705 PR tree-optimization/81384
706 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
707 * builtins.c (expand_builtin_strnlen): New function.
708 (expand_builtin): Call it.
709 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
710 * builtins.def (BUILT_IN_STRNLEN): New.
711 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
712 Warn for bounds in excess of maximum object size.
713 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
714 single-value ranges. Handle strnlen.
715 (handle_builtin_strlen): Handle strnlen.
716 (strlen_check_and_optimize_stmt): Same.
717 * doc/extend.texi (Other Builtins): Document strnlen.
719 2018-06-18 Maya Rashish <coypu@sdf.org>
721 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
722 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
723 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
725 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
727 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
729 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
731 * tree.c (escaped_string::escape): Replace cast to char * by
732 const_cast<char *> (unescaped).
734 2018-06-18 Nick Clifton <nickc@redhat.com>
737 * tree.c (escaped_string): New class. Converts an unescaped
738 string into its escaped equivalent.
739 (warn_deprecated_use): Use the new class to convert the
740 deprecation message, if present.
741 (test_escaped_strings): New self test.
742 (test_c_tests): Add test_escaped_strings.
743 * doc/extend.texi (deprecated): Add a note that the
744 deprecation message is affected by the -fmessage-length
745 option, and that control characters will be escaped.
746 (#pragma GCC error): Document this pragma.
747 (#pragma GCC warning): Likewise.
748 * doc/invoke.texi (-fmessage-length): Document this option's
749 effect on the #warning and #error preprocessor directives and
750 the deprecated attribute.
752 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
754 * tree.c (decl_value_expr_lookup): Revert latest change.
755 (decl_value_expr_insert): Likewise.
757 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
759 * gimplify.c (nonlocal_vlas): Delete.
760 (nonlocal_vla_vars): Likewise.
761 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
763 (gimplify_body): Do not create and destroy nonlocal_vlas.
764 * tree-nested.c: Include diagnostic.h.
765 (use_pointer_in_frame): Tweak.
766 (lookup_field_for_decl): Add assertion and declare the transformation.
767 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
768 internal error when the reference is in a wrong context. Do not
769 create a debug decl by default.
770 (note_nonlocal_block_vlas): Delete.
771 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
772 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
773 create a debug decl by default.
774 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
775 call is in a wrong context.
776 (fixup_vla_decls): New function.
777 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
778 debug variables were created.
779 * tree.c (decl_value_expr_lookup): Add checking assertion.
780 (decl_value_expr_insert): Likewise.
782 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
785 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
786 * tree-scalar-evolution.c (interpret_expr): Likewise.
787 (expression_expensive_p): Likewise.
788 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
789 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
790 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
791 (ssa_defined_by_minus_one_stmt_p): New.
793 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
796 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
797 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
798 * dojump.c (do_jump): Likewise.
799 * expr.c (expand_expr_real_2): Check operand type's sign.
800 * fold-const.c (const_unop): Handle ABSU_EXPR.
801 (fold_abs_const): Likewise.
802 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
803 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
804 (strip_sign_op_1): Likesise.
805 * match.pd: Add new pattern to generate ABSU_EXPR.
806 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
807 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
808 * tree-eh.c (operation_could_trap_helper_p): Likewise.
809 * tree-inline.c (estimate_operator_cost): Likewise.
810 * tree-pretty-print.c (dump_generic_node): Likewise.
811 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
812 * tree.def (ABSU_EXPR): New.
814 2018-06-16 Jakub Jelinek <jakub@redhat.com>
817 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
818 documented as preserved for backward compatibility only.
819 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
821 PR rtl-optimization/86108
822 * bb-reorder.c (create_forwarder_block): Renamed to ...
823 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
824 jump from new landing pad to the second part.
825 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
828 2018-06-15 Jakub Jelinek <jakub@redhat.com>
831 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
832 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
833 Only call store_expr for halves if the mode is the same.
836 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
837 Fix up comment formatting.
839 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
841 * typed-splay-tree.h (typed_splay_tree::remove): New function.
842 (typed_splay_tree::closure,
843 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
844 (typed_splay_tree::typed_splay_tree,
845 typed_splay_tree::operator =): Declared private.
846 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
847 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
848 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
849 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
850 typed_splay_tree::splay_tree_splay,
851 typed_splay_tree::splay_tree_foreach_helper,
852 typed_splay_tree::splay_tree_insert,
853 typed_splay_tree::splay_tree_remove,
854 typed_splay_tree::splay_tree_lookup,
855 typed_splay_tree::splay_tree_predecessor,
856 typed_splay_tree::splay_tree_successor,
857 typed_splay_tree::splay_tree_min,
858 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
859 (typed_splay_tree::root, typed_splay_tree::comp,
860 typed_splay_tree::delete_key,
861 typed_splay_tree::delete_value): New data members.
862 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
863 typed_splay_tree::remove.
865 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
867 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
868 -mginv and -mno-ginv to the assembler.
869 * config/mips/mips.opt (-mcrc): New option.
871 * doc/invoke.text (-mcrc): Document.
874 2018-06-15 Nick Clifton <nickc@redhat.com>
877 * tree.c (escaped_string): New class. Converts an unescaped
878 string into its escaped equivalent.
879 (warn_deprecated_use): Use the new class to convert the
880 deprecation message, if present.
881 (test_escaped_strings): New self test.
882 (test_c_tests): Add test_escaped_strings.
883 * doc/extend.texi (deprecated): Add a note that the
884 deprecation message is affected by the -fmessage-length
885 option, and that control characters will be escaped.
886 (#pragma GCC error): Document this pragma.
887 (#pragma GCC warning): Likewise.
888 * doc/invoke.texi (-fmessage-length): Document this option's
889 effect on the #warning and #error preprocessor directives and
890 the deprecated attribute.
892 2018-06-15 Richard Biener <rguenther@suse.de>
894 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
895 here, also noting vector size used.
896 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
897 size used in MSG_OPTIMIZED_LOCATIONS dump.
898 (pass_slp_vectorize::execute): Adjust.
900 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
903 * config/arc/arc.c (arc_return_address_register): Fix
906 2018-06-15 Richard Biener <rguenther@suse.de>
909 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
910 leave useless conversion stripping to force_gimple_operand_gsi.
911 (gimplify_build2): Likewise.
912 (gimplify_build1): Likewise.
914 2018-06-15 Richard Biener <rguenther@suse.de>
917 * tree-cfg.c (move_stmt_op): unshare invariant addresses
918 before adjusting their block.
920 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
922 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
923 multilibs for *-*-rtems*.
924 * config/riscv/t-rtems: New file.
926 2018-06-14 Jakub Jelinek <jakub@redhat.com>
929 * match.pd ((A +- CST1) +- CST2): Punt if last resort
930 unsigned_type_for returns NULL.
933 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
934 subregs of multi-word pseudos unless the float mode has word size.
936 2018-06-14 Richard Biener <rguenther@suse.de>
939 * tree-vect-generic.c (build_word_mode_vector_type): Remove
940 duplicate and harmful type_hash_canon.
941 * tree.c (type_hash_canon): Assert we didn't find ourselves.
943 2018-06-14 Richard Biener <rguenther@suse.de>
946 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
949 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
951 * config/rtems.h (STDINT_LONG32): Define.
953 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
954 Prachi Godbole <prachi.godbole@imgtec.com>
956 * config/mips/mips-cpus.def: Define P6600.
957 * config/mips/mips-tables.opt: Regenerate.
958 * config/mips/mips.c (mips_ucbranch_type): New enum.
959 (mips_rtx_cost_data): Add support for P6600.
960 (mips_issue_rate): Likewise.
961 (mips_multipass_dfa_lookahead): Likewise.
962 (mips_avoid_hazard): Likewise.
963 (mips_reorg_process_insns): Likewise.
964 (mips_classify_branch_p6600): New function.
965 * config/mips/mips.h (TUNE_P6600): New define.
966 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
967 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
968 * config/mips/mips.md: Include p6600.md.
969 (processor): Add p6600.
970 * config/mips/p6600.md: New file.
971 * doc/invoke.texi: Add p6600 to supported architectures.
973 2018-06-13 Martin Sebor <msebor@redhat.com>
975 PR tree-optimization/86114
976 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
978 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
980 2018-06-13 Richard Biener <rguenther@suse.de>
982 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
983 Properly set vector type of the intermediate stmt.
984 * tree-vect-stmts.c (vectorizable_operation): The destination
985 var always has vectype_out type.
987 2018-06-13 Jeff Law <law@redhat.com>
989 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
990 integer 0 for argument to print_rtl_with_bb.
991 (rl78_reorg): Likewise.
993 2018-06-13 David Malcolm <dmalcolm@redhat.com>
995 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
996 from rtx to rtx_insn *.
997 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
999 (add_sched_insns_for_speculation): Likewise for local "target",
1000 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1001 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
1002 from rtx_insn ** to rtx_code_label **.
1003 (reorg_emit_nops): Likewise.
1004 (c6x_reorg): Likewise for local "call_labels".
1005 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
1007 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
1008 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
1009 the loops over LABEL_REFS.
1010 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
1012 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
1013 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
1014 (split_branches): Strengthen local "olabel" from rtx to
1015 rtx_insn *, adding a safe_as_a cast.
1016 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
1018 (add_insn_after): Likewise for first two params.
1019 (add_insn_before): Likewise.
1020 (remove_insn): Likewise for param.
1021 (emit_pattern_before_noloc): Likewise for second and third params.
1022 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
1023 (emit_call_insn_before_noloc): Likewise.
1024 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
1026 (emit_barrier_before): Likewise.
1027 (emit_label_before): Strengthen "label" param from "rtx" to
1028 "rtx_code_label *". Strengthen "before" param from "rtx" to
1030 (emit_insn_after_1): Strengthen "after" param from "rtx" to
1032 (emit_pattern_after_noloc): Likewise.
1033 (emit_insn_after_noloc): Likewise.
1034 (emit_jump_insn_after_noloc): Likewise.
1035 (emit_call_insn_after_noloc): Likewise.
1036 (emit_debug_insn_after_noloc): Likewise.
1037 (emit_barrier_after): Likewise.
1038 (emit_label_after): Likewise for both params.
1039 (emit_pattern_after_setloc): Likewise for "after" param. Convert
1040 "loc" param from "int" to "location_t".
1041 (emit_insn_after_setloc): Likewise.
1042 (emit_jump_insn_after_setloc): Likewise.
1043 (emit_call_insn_after_setloc): Likewise.
1044 (emit_debug_insn_after_setloc): Likewise.
1045 (emit_pattern_before_setloc): Likewise for "before" param. Convert
1046 "loc" param from "int" to "location_t".
1047 (emit_pattern_before): Convert NULL_RTX to NULL.
1048 (emit_insn_before_setloc): Convert "loc" param from "int" to
1050 (emit_jump_insn_before_setloc): Likewise.
1051 (emit_call_insn_before_setloc): Likewise.
1052 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
1053 rtx_insn *. Convert "loc" param from "int" to "location_t".
1054 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
1055 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
1056 Convert 3rd param from "int" to "location_t".
1057 (emit_barrier_before, emit_barrier_after, next_real_insn):
1058 Strengthen param from rtx to rtx_insn *.
1059 (emit_label_before): Strengthen 1st param from "rtx" to
1060 "rtx_code_label *". Strengthen 2nd param from "rtx" to
1062 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
1063 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
1064 Strengthen 2nd param from "rtx" to "rtx_insn *".
1065 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
1066 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
1067 Likewise. Convert 3rd param from "int" to "location_t".
1068 (emit_label_after): Strengthen 1st param from "rtx" to
1070 (next_real_insn, remove_insn): Strengthen param from "rtx" to
1072 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
1073 from "rtx" to "rtx_insn *".
1075 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
1077 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
1078 bodies streamed in with -Q.
1079 * dumpfile.c (dump_files): Add lto-stream-out dump file.
1080 * dumpfile.h (tree_dump_index): Add lto_stream_out.
1081 * gimple-streamer-out.c: Include gimple-pretty-print.h
1082 (output_bb): Dump stmts streamed.
1083 * lto-section-out.c: Include print-tree.h
1084 (lto_begin_section): Dump sections created.
1085 (lto_output_decl_index): Dump decl encoded.
1086 * lto-streamer-out.c: Include print-tree.h
1087 (create_output_block): Dump output block created.
1088 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
1089 (output_function): Dump function output.
1090 (output_constructor): Dump constructor streamed.
1091 (write_global_stream): Output indexes encoded.
1092 (produce_asm_for_decls): Dump streams encoded.
1093 * lto-streamer.c (streamer_dump_file): New global var.
1094 * lto-streamer.h (streamer_dump_file): Declare.
1095 * passes.c (ipa_write_summaries): Initialize streamer dump.
1096 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
1099 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
1102 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
1103 offsets for register save directives. Emit a second batch of save
1104 directives, if need be, when the function accesses prior frames.
1106 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1108 * config/arc/fpu.md (fmasf4): Force operand to register.
1109 (fnmasf4): Likewise.
1111 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1113 * config/arc/arc-protos.h (arc_pad_return): Remove.
1114 * config/arc/arc.c (machine_function): Remove force_short_suffix
1116 (arc_print_operand): Adjust printing of '&'.
1117 (arc_verify_short): Remove conditional printing of short suffix.
1118 (arc_final_prescan_insn): Remove reference to size_reason.
1119 (pad_return): New function.
1120 (arc_reorg): Call pad_return.
1121 (arc_pad_return): Remove.
1122 (arc_init_machine_status): Remove reference to force_short_suffix.
1123 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
1124 (attr length): When attribute iscompact is true force to 2
1125 regardless; in the case of maybe check if we want to force the
1126 instruction to have 4 bytes length.
1127 (nopv): Change it to generate 4 byte long nop as well.
1128 (blockage): New pattern.
1129 (simple_return): Remove call to arc_pad_return.
1130 (p_return_i): Likewise.
1132 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1134 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1136 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1138 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1141 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1143 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1144 for ARC700 and ARCv2.
1146 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1149 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1150 operands[2] instead of operands[1].
1153 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1155 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1156 case, check whether the outer register overlaps an unallocatable
1157 register, not just whether it fits the required class.
1159 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1161 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1163 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1164 (memrefs_conflict_p): Likewise.
1165 (init_alias_analysis): Likewise.
1166 * cfgexpand.c (expand_debug_expr): Likewise.
1167 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1168 * cse.c (fold_rtx): Likewise.
1169 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1170 * expr.c (emit_block_move_hints): Likewise.
1171 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1172 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1173 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1174 (emit_group_store): Likewise.
1175 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1176 to read the PRE/POST_MODIFY increment.
1177 * calls.c (store_one_arg): Use strip_offset.
1178 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1180 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1182 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1183 (simplify_binary_operation_1): Extend CONST_INT handling to
1185 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1186 than a HOST_WIDE_INT.
1187 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1189 (adjust_mems, add_stores): Update accodingly.
1190 (vt_canonicalize_addr): Track polynomial offsets.
1191 (emit_note_insn_var_location): Likewise.
1192 (vt_add_function_parameter): Likewise.
1193 (vt_initialize): Likewise.
1195 2018-06-12 Jeff Law <law@redhat.com>
1197 * config.gcc (alpha*-*-freebsd*): Remove.
1198 * config/alpha/freebsd.h: Remove.
1200 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1203 * spellcheck-tree.c (levenshtein_distance): Rename to...
1204 (get_edit_distance): ...this, and update for underlying renaming.
1205 * spellcheck-tree.h (levenshtein_distance): Rename to...
1206 (get_edit_distance): ...this.
1207 * spellcheck.c (levenshtein_distance): Rename to...
1208 (get_edit_distance): ...this. Convert from Levenshtein distance
1209 to Damerau-Levenshtein distance by supporting transpositions of
1210 adjacent characters. Rename "v1" to "v_next" and "v0" to
1212 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1213 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1214 update for underlying renaming.
1215 (selftest::levenshtein_distance_unit_test): Rename to...
1216 (selftest::test_get_edit_distance_unit): ...this, and update for
1217 underlying renaming.
1218 (selftest::test_find_closest_string): Add example from PR 69968
1219 where transposition helps
1220 (selftest::test_metric_conditions): Update for renaming.
1221 (selftest::test_metric_conditions): Likewise.
1222 (selftest::spellcheck_c_tests): Likewise.
1223 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1224 (get_edit_distance): ...this.
1225 (best_match::consider): Update for renaming.
1227 2018-06-12 Martin Sebor <msebor@redhat.com>
1229 PR tree-optimization/85259
1230 * builtins.c (compute_objsize): Handle constant offsets.
1231 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1232 true iff a warning has been issued.
1233 * gimple.h (gimple_nonartificial_location): New function.
1234 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1235 gimple_nonartificial_location and handle -Wno-system-headers.
1236 (handle_builtin_stxncpy): Same.
1238 2018-06-12 Martin Sebor <msebor@redhat.com>
1241 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1243 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1245 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1246 BUILTIN_VEC_XST entries for pointer to double and long long.
1248 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1251 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1253 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1256 2018-06-12 Martin Liska <mliska@suse.cz>
1258 * doc/options.texi: Document IntegerRange.
1260 2018-06-12 Martin Liska <mliska@suse.cz>
1262 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1263 * opt-functions.awk: Handle Deprecated flag.
1264 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1266 (read_cmdline_option): Report warning for a deprecated option.
1267 * opts.h (struct cl_option): Add new field cl_deprecated.
1268 (CL_ERR_DEPRECATED): New.
1270 2018-06-12 Martin Liska <mliska@suse.cz>
1272 * doc/options.texi: Document Deprecated option flag.
1274 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1276 * config/arc/arc-arch.h (arc_extras): New enum.
1277 (arc_cpu_t):Add field extra.
1278 (arc_cpu_types): Consider the extras.
1279 * config/arc/arc-cpus.def: Add extras info.
1280 * config/arc/arc-opts.h (processor_type): Consider extra field.
1281 * config/arc/arc.c (arc_override_options): Handle extra field.
1283 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1285 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1286 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1288 * config/arc/arc.c (arc_init): Update pic warning.
1289 (irq_range): Update irq range parsing warnings.
1290 (arc_override_options): Update various warning messages.
1291 (arc_handle_aux_attribute): Likewise.
1293 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1295 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1297 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1299 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1300 compiler messages directives.
1302 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1304 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1305 * config/mips/mips-tables.opt: Regenerate.
1306 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1308 * doc/invoke.texi: Document -march=i6500.
1310 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1312 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1313 (i6400_gpmul): Add cpu_unit.
1314 (i6400_gpdiv): Likewise.
1315 (i6400_msa_add_d): Update reservations.
1316 (i6400_msa_int_add) Likewise.
1317 (i6400_msa_short_logic3) Likewise.
1318 (i6400_msa_short_logic2) Likewise.
1319 (i6400_msa_short_logic) Likewise.
1320 (i6400_msa_move) Likewise.
1321 (i6400_msa_cmp) Likewise.
1322 (i6400_msa_short_float2) Likewise.
1323 (i6400_msa_div_d) Likewise.
1324 (i6400_msa_long_logic1) Likewise.
1325 (i6400_msa_long_logic2) Likewise.
1326 (i6400_msa_mult) Likewise.
1327 (i6400_msa_long_float2) Likewise.
1328 (i6400_msa_long_float4) Likewise.
1329 (i6400_msa_long_float5) Likewise.
1330 (i6400_msa_long_float8) Likewise.
1331 (i6400_fpu_fadd): Include frint type.
1332 (i6400_fpu_store): New define_insn_reservation.
1333 (i6400_fpu_load): Likewise.
1334 (i6400_fpu_move): Likewise.
1335 (i6400_fpu_fcmp): Likewise.
1336 (i6400_fpu_fmadd): Likewise.
1337 (i6400_int_mult): Include imul3nc type and update reservation.
1338 (i6400_int_div): Include idiv3 type and update reservation.
1339 (i6400_int_load): Update to check type not move_type.
1340 (i6400_int_store): Likewise.
1341 (i6400_int_prefetch): Set zero latency.
1343 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1345 * gcc.c: Document new %@{...} sequence.
1346 (LINK_COMMAND_SPEC): Use it for the -L switches.
1347 (cpp_unique_options): Use it for the -I switches.
1348 (at_file_argbuf): New global variable.
1349 (in_at_file): Likewise.
1350 (alloc_args): Create at_file_argbuf.
1351 (clear_args): Truncate at_file_argbuf.
1352 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1353 (open_at_file): New function.
1354 (close_at_file): Likewise.
1355 (create_at_file): Delete.
1356 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1359 (validate_switches_from_spec): Deal with %@{...} sequence.
1360 (validate_switches): Likewise.
1361 (driver::finalize): Call clear_args.
1363 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1365 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1367 2018-06-11 Martin Sebor <msebor@redhat.com>
1369 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1370 (Wno-class-memaccess): Add @opindex.
1371 (Wno-templates, Wno-multiple-inheritance): Same.
1372 (Wno-virtual-inheritance, Wno-namespaces): Same.
1373 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1374 (Wno-format-overflow, Wno-format-truncation): Same.
1375 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1376 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1377 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1378 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1379 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1380 (Wno-misspelled-isr): Same.
1382 2018-06-11 Martin Sebor <msebor@redhat.com>
1384 * PR tree-optimization/86083
1385 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1387 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1389 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1391 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1394 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1395 on the correct operand.
1396 (*movdi_internal64): Ditto.
1398 2018-06-11 Martin Liska <mliska@suse.cz>
1400 PR tree-optimization/86089
1401 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1403 2018-06-11 Julia Koval <julia.koval@intel.com>
1405 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1406 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1407 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1409 2018-06-11 Olivier Hainque <hainque@adacore.com>
1411 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1412 for Ada with strict dwarf2.
1414 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1417 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1420 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1422 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1424 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1426 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1429 2018-06-08 Martin Liska <mliska@suse.cz>
1431 * tree-cfg.h (debug_function): Fix argument type to match
1434 2018-06-08 Martin Liska <mliska@suse.cz>
1436 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1437 Remove usage of MPX-related (and removed) fields.
1438 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1440 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1442 * cfg.c (debug): Use TDF_NONE rather than 0.
1443 * cfghooks.c (debug): Likewise.
1444 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1445 (struct dump_option_value_info): Convert to...
1446 (struct kv_pair): ...this template type.
1447 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1449 (optinfo_verbosity_options): Likewise.
1450 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1452 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1453 than int for "optgroup_flags" param.
1454 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1456 (dump_dec): Likewise.
1457 (dump_finish): Use TDF_NONE rather than 0.
1458 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1459 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1460 than 0. Update for change to option_ptr.
1461 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1462 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1463 0. Update for changes to optinfo_verbosity_options and
1465 (opt_info_switch_p): Convert optgroup_flags from int to
1467 (dump_basic_block): Use dump_flags_t rather than int
1468 for "dump_kind" param.
1469 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1470 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1471 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1472 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1473 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1474 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1475 TDF_NONE): Convert from macros to...
1476 (enum dump_flag): ...this new enum.
1477 (dump_flags_t): Update to use enum.
1478 (operator|, operator&, operator~, operator|=, operator&=):
1479 Implement for dump_flags_t.
1480 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1481 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1482 Convert from macros to...
1483 (enum optgroup_flag): ...this new enum.
1484 (optgroup_flags_t): New typedef.
1485 (operator|, operator|=): Implement for optgroup_flags_t.
1486 (struct dump_file_info): Convert field "alt_flags" to
1487 dump_flags_t. Convert field "optgroup_flags" to
1489 (dump_basic_block): Use dump_flags_t rather than int for param.
1490 (dump_generic_expr_loc): Likewise.
1491 (dump_dec): Likewise.
1492 (dump_register): Convert param "optgroup_flags" to
1494 (opt_info_enable_passes): Likewise.
1495 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1497 * gimple-pretty-print.c (debug): Likewise.
1498 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1499 (merged_store_group::apply_stores): Likewise.
1500 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1501 * gimple.c (verify_gimple_pp): Likewise.
1502 * graphite-poly.c (print_pbb_body): Likewise.
1503 * passes.c (pass_manager::register_one_dump_file): Convert
1504 local "optgroup_flags" to optgroup_flags_t.
1505 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1507 (debug_body): Likewise.
1508 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1509 to optgroup_flags_t.
1510 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1512 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1513 (convert_mult_to_fma): Likewise.
1514 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1515 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1516 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1517 "dump_kind" to dump_flags_t.
1519 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1521 * config/rs6000/rs6000.c (min, max): Delete.
1523 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1525 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1526 -mabi=spe and -mabi=no-spe.
1528 2018-06-08 Martin Liska <mliska@suse.cz>
1530 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1531 where we expect an existing summary.
1533 2018-06-08 Martin Liska <mliska@suse.cz>
1535 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1536 * ipa-inline.h (estimate_edge_growth): Likewise.
1538 2018-06-08 Martin Liska <mliska@suse.cz>
1540 * cgraph.c (function_version_hasher::hash): Use
1541 cgraph_node::get_uid ().
1542 (function_version_hasher::equal):
1543 * cgraph.h (cgraph_node::get_uid): New method.
1544 * ipa-inline.c (update_caller_keys): Use
1545 cgraph_node::get_uid ().
1546 (update_callee_keys): Likewise.
1547 * ipa-utils.c (searchc): Likewise.
1548 (ipa_reduced_postorder): Likewise.
1549 * lto-cgraph.c (input_node): Likewise.
1550 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1551 * symbol-summary.h (symtab_insertion): Likewise.
1552 (symtab_removal): Likewise.
1553 (symtab_duplication): Likewise.
1554 * tree-pretty-print.c (dump_function_header): Likewise.
1555 * tree-sra.c (convert_callers_for_node): Likewise.
1557 2018-06-08 Martin Liska <mliska@suse.cz>
1559 * cgraph.c (symbol_table::create_edge): Always assign a new
1561 (symbol_table::free_edge): Do not recycle numbers.
1562 * cgraph.h (cgraph_edge::get): New method.
1563 * symbol-summary.h (symtab_removal): Use it.
1564 (symtab_duplication): Likewise.
1565 (call_summary::hashable_uid): Remove.
1567 2018-06-08 Martin Liska <mliska@suse.cz>
1569 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1570 (initialize_growth_caches): Remove.
1571 (free_growth_caches): Likewise.
1572 (do_estimate_edge_time): Use edge_growth_cache.
1573 (do_estimate_edge_size): Likewise.
1574 (do_estimate_edge_hints): Likewise.
1575 * ipa-inline.c (reset_edge_caches): Likewise.
1576 (recursive_inlining): Likewise.
1577 (inline_small_functions): Likewise.
1578 * ipa-inline.h (initialize_growth_caches): Remove.
1579 (estimate_edge_size): Likewise.
1580 (estimate_edge_time): Likewise.
1581 (estimate_edge_hints): Likewise.
1582 (reset_edge_growth_cache): Likewise.
1583 * symbol-summary.h (call_summary::remove): New method.
1585 2018-06-08 Martin Liska <mliska@suse.cz>
1587 * ipa-cp.c (class edge_clone_summary): New summary.
1588 (grow_edge_clone_vectors): Remove.
1589 (ipcp_edge_duplication_hook): Remove.
1590 (class edge_clone_summary_t): New call_summary class.
1591 (ipcp_edge_removal_hook): Remove.
1592 (edge_clone_summary_t::duplicate): New function.
1593 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1594 (create_specialized_node): Likewise.
1595 (ipcp_driver): Initialize edge_clone_summaries and do not
1598 2018-06-08 Martin Liska <mliska@suse.cz>
1600 * symbol-summary.h (get): New function.
1601 (call_summary::m_initialize_when_cloning): New class member.
1603 2018-06-08 Martin Liska <mliska@suse.cz>
1605 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1606 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1607 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1608 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1609 (remove_cgraph_node_from_order): Use the removed_nodes set.
1610 (do_per_function_toporder): Likwise.
1611 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1612 instead of summary_uid.
1613 (symtab_removal): Likewise.
1614 (symtab_duplication): Likewise.
1616 2018-06-08 Martin Liska <mliska@suse.cz>
1618 * ipa-cp.c (ipcp_store_bits_results): Use
1619 ipcp_transformation_sum.
1620 (ipcp_store_vr_results): Likewise.
1621 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1623 (ipcp_transformation_initialize): ... this.
1624 (ipa_set_node_agg_value_chain):
1625 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1626 (write_ipcp_transformation_info): Likewise.
1627 (read_ipcp_transformation_info): Likewise.
1628 (ipcp_update_bits): Likewise.
1629 (ipcp_update_vr): Likewise.
1630 (ipcp_transform_function): Likewise.
1631 * ipa-prop.h: Rename ipcp_transformation_summary to
1632 ipcp_transformation.
1633 (class ipcp_transformation_t): New function summary.
1634 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1635 (ipa_get_agg_replacements_for_node): Likewise.
1637 2018-06-08 Martin Liska <mliska@suse.cz>
1639 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1641 (class funct_state_summary_t): New function_summary class.
1642 (has_function_state): Remove.
1643 (get_function_state): Likewise.
1644 (set_function_state): Likewise.
1645 (add_new_function): Likewise.
1646 (funct_state_summary_t::insert): New function.
1647 (duplicate_node_data): Remove.
1648 (remove_node_data): Remove.
1649 (funct_state_summary_t::duplicate): New function.
1650 (register_hooks): Create new funct_state_summaries.
1651 (pure_const_generate_summary): Use it.
1652 (pure_const_write_summary): Likewise.
1653 (pure_const_read_summary): Likewise.
1654 (propagate_pure_const): Likewise.
1655 (propagate_nothrow): Likewise.
1656 (dump_malloc_lattice): Likewise.
1657 (propagate_malloc): Likewise.
1658 (execute): Do not register hooks, just remove summary
1660 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1663 2018-06-08 Martin Liska <mliska@suse.cz>
1665 * ipa-reference.c (remove_node_data): Remove.
1666 (duplicate_node_data): Likewise.
1667 (class ipa_ref_var_info_summary_t): New class.
1668 (class ipa_ref_opt_summary_t): Likewise.
1669 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1670 (get_reference_optimization_summary): Use
1671 ipa_ref_opt_sum_summaries.
1672 (set_reference_vars_info): Remove.
1673 (set_reference_optimization_summary): Likewise.
1674 (ipa_init): Create summaries.
1675 (init_function_info): Use function summary.
1676 (ipa_ref_opt_summary_t::duplicate): New function.
1677 (ipa_ref_opt_summary_t::remove): New function.
1678 (get_read_write_all_from_node): Fix GNU coding style.
1679 (propagate): Use function summary.
1680 (write_node_summary_p): Fix GNU coding style.
1681 (stream_out_bitmap): Likewise.
1682 (ipa_reference_read_optimization_summary): Use function summary.
1683 (ipa_reference_c_finalize): Do not release hooks.
1685 2018-06-08 Martin Liska <mliska@suse.cz>
1687 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1688 (analyze_function_body): Extract multiple calls of get_create.
1689 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1690 * ipa-inline.c (recursive_inlining): Use ::get method.
1691 * ipa-inline.h (estimate_edge_growth): Likewise.
1693 2018-06-08 Martin Liska <mliska@suse.cz>
1695 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1697 (hsa_function_summary::hsa_function_summary): Use the new enum
1699 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1700 * hsa-gen.c (hsa_get_host_function): Likewise.
1701 (get_brig_function_name): Likewise.
1702 * ipa-hsa.c (process_hsa_functions): Likewise.
1703 (ipa_hsa_write_summary): Likewise.
1704 * symbol-summary.h (symtab_duplication): Use ::get function/
1705 (get): New function.
1707 2018-06-08 Martin Liska <mliska@suse.cz>
1709 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1711 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1712 (hsa_register_kernel): Likewise.
1713 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1714 * hsa-gen.c (hsa_get_host_function): Likewise.
1715 (get_brig_function_name): Likewise.
1716 (generate_hsa): Likewise.
1717 (pass_gen_hsail::execute): Likewise.
1718 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1719 (devirtualization_time_bonus): Likewise.
1720 (ipcp_propagate_stage): Likewise.
1721 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1722 (edge_set_predicate): Likewise.
1723 (evaluate_conditions_for_known_args): Likewise.
1724 (evaluate_properties_for_edge): Likewise.
1725 (ipa_fn_summary::reset): Likewise.
1726 (ipa_fn_summary_t::duplicate): Likewise.
1727 (dump_ipa_call_summary): Likewise.
1728 (ipa_dump_fn_summary): Likewise.
1729 (analyze_function_body): Likewise.
1730 (compute_fn_summary): Likewise.
1731 (estimate_edge_devirt_benefit): Likewise.
1732 (estimate_edge_size_and_time): Likewise.
1733 (estimate_calls_size_and_time): Likewise.
1734 (estimate_node_size_and_time): Likewise.
1735 (inline_update_callee_summaries): Likewise.
1736 (remap_edge_change_prob): Likewise.
1737 (remap_edge_summaries): Likewise.
1738 (ipa_merge_fn_summary_after_inlining): Likewise.
1739 (ipa_update_overall_fn_summary): Likewise.
1740 (read_ipa_call_summary): Likewise.
1741 (inline_read_section): Likewise.
1742 (write_ipa_call_summary): Likewise.
1743 (ipa_fn_summary_write): Likewise.
1744 (ipa_free_fn_summary): Likewise.
1745 * ipa-hsa.c (process_hsa_functions): Likewise.
1746 (ipa_hsa_write_summary): Likewise.
1747 (ipa_hsa_read_section): Likewise.
1748 * ipa-icf.c (sem_function::merge): Likewise.
1749 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1750 (do_estimate_edge_time): Likewise.
1751 (estimate_size_after_inlining): Likewise.
1752 (estimate_growth): Likewise.
1753 (growth_likely_positive): Likewise.
1754 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1755 (inline_call): Likewise.
1756 * ipa-inline.c (caller_growth_limits): Likewise.
1757 (can_inline_edge_p): Likewise.
1758 (can_inline_edge_by_limits_p): Likewise.
1759 (compute_uninlined_call_time): Likewise.
1760 (compute_inlined_call_time): Likewise.
1761 (want_inline_small_function_p): Likewise.
1762 (edge_badness): Likewise.
1763 (update_caller_keys): Likewise.
1764 (update_callee_keys): Likewise.
1765 (recursive_inlining): Likewise.
1766 (inline_small_functions): Likewise.
1767 (inline_to_all_callers_1): Likewise.
1768 (dump_overall_stats): Likewise.
1769 (early_inline_small_functions): Likewise.
1770 (early_inliner): Likewise.
1771 * ipa-inline.h (estimate_edge_growth): Likewise.
1772 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1773 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1774 * ipa-prop.h (IPA_NODE_REF): Likewise.
1775 (IPA_EDGE_REF): Likewise.
1776 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1777 (propagate_malloc): Likewise.
1778 * ipa-split.c (execute_split_functions): Likewise.
1779 * symbol-summary.h: Rename get to get_create.
1781 (get_create): Likewise.
1782 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1784 2018-06-08 Martin Liska <mliska@suse.cz>
1786 * symbol-summary.h (release): Move definition out of class
1788 (symtab_removal): Likewise.
1789 (symtab_duplication): Likewise.
1791 2018-06-08 Martin Liska <mliska@suse.cz>
1793 * symbol-summary.h (function_summary): Move constructor
1794 implementation out of class declaration.
1795 (release): Likewise.
1796 (symtab_insertion): Likewise.
1797 (symtab_removal): Likewise.
1798 (symtab_duplication): Likewise.
1801 2018-06-08 Martin Liska <mliska@suse.cz>
1803 * Makefile.in: Remove support for MPX (macros, related functions,
1804 fields in cgraph_node, ...).
1805 * builtin-types.def (BT_BND): Likewise.
1806 (BT_FN_BND_CONST_PTR): Likewise.
1807 (BT_FN_CONST_PTR_BND): Likewise.
1808 (BT_FN_VOID_PTR_BND): Likewise.
1809 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1810 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1811 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1812 (expand_builtin_mempcpy_with_bounds): Likewise.
1813 (expand_builtin_memset_with_bounds): Likewise.
1814 (expand_builtin_memset_args): Likewise.
1815 (std_expand_builtin_va_start): Likewise.
1816 (expand_builtin): Likewise.
1817 (expand_builtin_with_bounds): Likewise.
1818 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1819 (DEF_LIB_BUILTIN_CHKP): Likewise.
1820 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1821 (DEF_CHKP_BUILTIN): Likewise.
1822 (BUILT_IN_MEMCPY): Likewise.
1823 (BUILT_IN_MEMMOVE): Likewise.
1824 (BUILT_IN_MEMPCPY): Likewise.
1825 (BUILT_IN_MEMSET): Likewise.
1826 (BUILT_IN_STPCPY): Likewise.
1827 (BUILT_IN_STRCAT): Likewise.
1828 (BUILT_IN_STRCHR): Likewise.
1829 (BUILT_IN_STRCPY): Likewise.
1830 (BUILT_IN_STRLEN): Likewise.
1831 (BUILT_IN_MEMCPY_CHK): Likewise.
1832 (BUILT_IN_MEMMOVE_CHK): Likewise.
1833 (BUILT_IN_MEMPCPY_CHK): Likewise.
1834 (BUILT_IN_MEMSET_CHK): Likewise.
1835 (BUILT_IN_STPCPY_CHK): Likewise.
1836 (BUILT_IN_STRCAT_CHK): Likewise.
1837 (BUILT_IN_STRCPY_CHK): Likewise.
1838 * calls.c (store_bounds): Likewise.
1839 (emit_call_1): Likewise.
1840 (special_function_p): Likewise.
1841 (maybe_warn_nonstring_arg): Likewise.
1842 (initialize_argument_information): Likewise.
1843 (finalize_must_preallocate): Likewise.
1844 (compute_argument_addresses): Likewise.
1845 (expand_call): Likewise.
1846 * cfgexpand.c (expand_call_stmt): Likewise.
1847 (expand_return): Likewise.
1848 (expand_gimple_stmt_1): Likewise.
1849 (pass_expand::execute): Likewise.
1850 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1851 (cgraph_node::remove): Likewise.
1852 (cgraph_node::dump): Likewise.
1853 (cgraph_node::verify_node): Likewise.
1854 * cgraph.h (chkp_function_instrumented_p): Likewise.
1855 (symtab_node::get_alias_target): Likewise.
1856 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1857 (cgraph_local_p): Likewise.
1858 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1859 (cgraph_edge::rebuild_references): Likewise.
1860 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1861 (walk_polymorphic_call_targets): Likewise.
1862 (cgraph_node::expand_thunk): Likewise.
1863 (symbol_table::output_weakrefs): Likewise.
1864 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1865 (ix86_handle_option): Likewise.
1866 * config/i386/constraints.md: Likewise.
1867 * config/i386/i386-builtin-types.def (BND): Likewise.
1871 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1872 (BDESC_FIRST): Likewise.
1874 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1875 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1876 * config/i386/i386.c (enum reg_class): Likewise.
1877 (ix86_target_string): Likewise.
1878 (ix86_option_override_internal): Likewise.
1879 (ix86_conditional_register_usage): Likewise.
1880 (ix86_valid_target_attribute_inner_p): Likewise.
1881 (ix86_set_indirect_branch_type): Likewise.
1882 (ix86_set_current_function): Likewise.
1883 (ix86_function_arg_regno_p): Likewise.
1884 (init_cumulative_args): Likewise.
1885 (ix86_function_arg_advance): Likewise.
1886 (ix86_function_arg): Likewise.
1887 (ix86_pass_by_reference): Likewise.
1888 (ix86_function_value_regno_p): Likewise.
1889 (ix86_function_value_1): Likewise.
1890 (ix86_function_value_bounds): Likewise.
1891 (ix86_return_in_memory): Likewise.
1892 (ix86_setup_incoming_vararg_bounds): Likewise.
1893 (ix86_va_start): Likewise.
1894 (indirect_thunk_need_prefix): Likewise.
1895 (print_reg): Likewise.
1896 (ix86_print_operand): Likewise.
1897 (ix86_expand_call): Likewise.
1898 (ix86_output_function_return): Likewise.
1899 (reg_encoded_number): Likewise.
1900 (BDESC_VERIFYS): Likewise.
1901 (ix86_init_mpx_builtins): Likewise.
1902 (ix86_init_builtins): Likewise.
1903 (ix86_emit_cmove): Likewise.
1904 (ix86_emit_move_max): Likewise.
1905 (ix86_expand_builtin): Likewise.
1906 (ix86_builtin_mpx_function): Likewise.
1907 (ix86_get_arg_address_for_bt): Likewise.
1908 (ix86_load_bounds): Likewise.
1909 (ix86_store_bounds): Likewise.
1910 (ix86_load_returned_bounds): Likewise.
1911 (ix86_store_returned_bounds): Likewise.
1912 (ix86_class_likely_spilled_p): Likewise.
1913 (ix86_hard_regno_mode_ok): Likewise.
1914 (x86_order_regs_for_local_alloc): Likewise.
1915 (ix86_mitigate_rop): Likewise.
1916 (ix86_bnd_prefixed_insn_p): Likewise.
1917 (ix86_mpx_bound_mode): Likewise.
1918 (ix86_make_bounds_constant): Likewise.
1919 (ix86_initialize_bounds): Likewise.
1920 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1921 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1922 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1923 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1924 (TARGET_CHKP_BOUND_MODE): Likewise.
1925 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1926 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1927 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1928 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1929 * config/i386/i386.h (TARGET_MPX): Likewise.
1930 (TARGET_MPX_P): Likewise.
1931 (VALID_BND_REG_MODE): Likewise.
1932 (FIRST_BND_REG): Likewise.
1933 (LAST_BND_REG): Likewise.
1934 (enum reg_class): Likewise.
1935 (BND_REG_P): Likewise.
1936 (BND_REGNO_P): Likewise.
1937 (BNDmode): Likewise.
1938 (ADJUST_INSN_LENGTH): Likewise.
1939 * config/i386/i386.md: Likewise.
1940 * config/i386/i386.opt: Likewise.
1941 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1942 (defined): Likewise.
1943 (LINK_MPX): Likewise.
1944 (MPX_SPEC): Likewise.
1945 (LIBMPX_SPEC): Likewise.
1946 (LIBMPXWRAPPERS_SPEC): Likewise.
1947 (CHKP_SPEC): Likewise.
1948 * config/i386/predicates.md: Likewise.
1949 * dbxout.c (dbxout_type): Likewise.
1950 * doc/extend.texi: Likewise.
1951 * doc/invoke.texi: Likewise.
1952 * doc/md.texi: Likewise.
1953 * doc/tm.texi: Likewise.
1954 * doc/tm.texi.in: Likewise.
1955 * dwarf2out.c (is_base_type): Likewise.
1956 (gen_formal_types_die): Likewise.
1957 (gen_subprogram_die): Likewise.
1958 (gen_type_die_with_usage): Likewise.
1959 (gen_decl_die): Likewise.
1960 (dwarf2out_late_global_decl): Likewise.
1961 * expr.c (expand_assignment): Likewise.
1962 (emit_storent_insn): Likewise.
1963 (store_expr_with_bounds): Likewise.
1964 (store_expr): Likewise.
1965 (expand_expr_real_1): Likewise.
1966 * expr.h (store_expr_with_bounds): Likewise.
1967 * function.c (use_register_for_decl): Likewise.
1968 (struct bounds_parm_data): Likewise.
1969 (assign_parms_augmented_arg_list): Likewise.
1970 (assign_parm_find_entry_rtl): Likewise.
1971 (assign_parm_is_stack_parm): Likewise.
1972 (assign_parm_load_bounds): Likewise.
1973 (assign_bounds): Likewise.
1974 (assign_parms): Likewise.
1975 (expand_function_start): Likewise.
1976 * gcc.c (CHKP_SPEC): Likewise.
1977 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1978 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1979 (wrestrict_dom_walker::check_call): Likewise.
1980 * gimple.c (gimple_build_call_from_tree): Likewise.
1981 * gimple.h (enum gf_mask): Likewise.
1982 (gimple_call_with_bounds_p): Likewise.
1983 (gimple_call_set_with_bounds): Likewise.
1984 * gimplify.c (gimplify_init_constructor): Likewise.
1985 * ipa-cp.c (initialize_node_lattices): Likewise.
1986 (propagate_constants_across_call): Likewise.
1987 (find_more_scalar_values_for_callers_subset): Likewise.
1988 * ipa-hsa.c (process_hsa_functions): Likewise.
1989 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1990 * ipa-icf.c (sem_function::merge): Likewise.
1991 * ipa-inline.c (early_inliner): Likewise.
1992 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1993 (warn_function_cold): Likewise.
1994 (propagate_pure_const): Likewise.
1995 * ipa-ref.h (enum GTY): Likewise.
1996 * ipa-split.c (find_retbnd): Likewise.
1997 (consider_split): Likewise.
1998 (split_function): Likewise.
1999 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2000 * ipa.c (walk_polymorphic_call_targets): Likewise.
2001 (symbol_table::remove_unreachable_nodes): Likewise.
2002 (process_references): Likewise.
2003 (cgraph_build_static_cdtor_1): Likewise.
2004 * lto-cgraph.c (lto_output_node): Likewise.
2005 (output_refs): Likewise.
2006 (compute_ltrans_boundary): Likewise.
2007 (input_overwrite_node): Likewise.
2008 (input_node): Likewise.
2009 (input_cgraph_1): Likewise.
2010 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
2011 * passes.c (pass_manager::execute_early_local_passes): Likewise.
2012 (class pass_chkp_instrumentation_passes): Likewise.
2013 (make_pass_chkp_instrumentation_passes): Likewise.
2014 * passes.def: Likewise.
2015 * rtl.h (struct GTY): Likewise.
2016 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
2017 * stor-layout.c (layout_type): Likewise.
2018 * symtab.c: Likewise.
2019 * target.def: Likewise.
2020 * targhooks.c (default_chkp_bound_type): Likewise.
2021 (default_chkp_bound_mode): Likewise.
2022 (default_builtin_chkp_function): Likewise.
2023 (default_chkp_function_value_bounds): Likewise.
2024 (default_chkp_make_bounds_constant): Likewise.
2025 (default_chkp_initialize_bounds): Likewise.
2026 * targhooks.h (default_chkp_bound_type): Likewise.
2027 (default_chkp_bound_mode): Likewise.
2028 (default_builtin_chkp_function): Likewise.
2029 (default_chkp_function_value_bounds): Likewise.
2030 (default_chkp_make_bounds_constant): Likewise.
2031 (default_chkp_initialize_bounds): Likewise.
2032 * toplev.c (compile_file): Likewise.
2033 (process_options): Likewise.
2034 * tree-core.h (DEF_BUILTIN): Likewise.
2035 (DEF_BUILTIN_CHKP): Likewise.
2036 * tree-inline.c (declare_return_variable): Likewise.
2037 (remap_gimple_stmt): Likewise.
2038 (copy_bb): Likewise.
2039 (initialize_inlined_parameters): Likewise.
2040 (expand_call_inline): Likewise.
2041 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
2042 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
2043 (make_pass_ipa_chkp_produce_thunks): Likewise.
2044 (make_pass_chkp): Likewise.
2045 (make_pass_chkp_opt): Likewise.
2046 (make_pass_chkp_instrumentation_passes): Likewise.
2047 * tree-pretty-print.c (dump_generic_node): Likewise.
2048 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
2049 * tree-ssa-dce.c (propagate_necessity): Likewise.
2050 (eliminate_unnecessary_stmts): Likewise.
2051 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
2052 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
2053 * tree-ssa-sccvn.h: Likewise.
2054 * tree-ssa-strlen.c (get_string_length): Likewise.
2055 (valid_builtin_call): Likewise.
2056 (adjust_last_stmt): Likewise.
2057 (handle_builtin_strchr): Likewise.
2058 (handle_builtin_strcpy): Likewise.
2059 (handle_builtin_stxncpy): Likewise.
2060 (handle_builtin_memcpy): Likewise.
2061 (handle_builtin_strcat): Likewise.
2062 (strlen_check_and_optimize_stmt): Likewise.
2063 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
2064 * tree-streamer-in.c: Likewise.
2065 * tree-streamer.c (record_common_node): Likewise.
2066 * tree.c (tree_code_size): Likewise.
2067 (wide_int_to_tree_1): Likewise.
2068 (type_contains_placeholder_1): Likewise.
2069 (build_common_tree_nodes): Likewise.
2070 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
2071 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
2072 (POINTER_BOUNDS_P): Likewise.
2073 (BOUNDED_TYPE_P): Likewise.
2074 (BOUNDED_P): Likewise.
2075 (CALL_WITH_BOUNDS_P): Likewise.
2076 (pointer_bounds_type_node): Likewise.
2077 * value-prof.c (gimple_ic): Likewise.
2078 * var-tracking.c (vt_add_function_parameters): Likewise.
2079 * varasm.c (make_decl_rtl): Likewise.
2080 (assemble_start_function): Likewise.
2081 (output_constant): Likewise.
2082 (maybe_assemble_visibility): Likewise.
2083 * varpool.c (ctor_for_folding): Likewise.
2084 * chkp-builtins.def: Remove.
2085 * ipa-chkp.c: Remove.
2086 * ipa-chkp.h: Remove.
2087 * rtl-chkp.c: Remove.
2088 * rtl-chkp.h: Remove.
2089 * tree-chkp-opt.c: Remove.
2090 * tree-chkp.c: Remove.
2091 * tree-chkp.h: Remove.
2093 2018-06-07 Carl Love <cel@us.ibm.com>
2095 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
2096 vextract_fp_from_shortl): Add BE support.
2098 2018-06-07 Paul Koning <ni1d@arrl.net>
2100 * compare-elim.c (try_merge_compare): Don't merge compare if
2101 address contains a side effect.
2102 (try_eliminate_compare): Likewise.
2104 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
2106 * config.gcc: Support "tremont".
2107 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
2108 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2110 * config/i386/i386.c (m_TREMONT): Define.
2111 (processor_target_table): Add "tremont".
2112 (PTA_TREMONT): Define.
2113 (ix86_lea_outperforms): Add TARGET_TREMONT.
2114 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
2115 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
2116 and M_INTEL_GOLDMONT_PLUS.
2117 (fold_builtin_cpu): Add "tremont".
2118 (ix86_add_stmt_cost): Add TARGET_TREMONT.
2119 (ix86_option_override_internal): Add "tremont".
2120 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
2121 (processor_type): Add PROCESSOR_TREMONT.
2122 * config/i386/x86-tune.def: Add m_TREMONT.
2123 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
2125 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2127 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2128 symbol defined for msp430i* devices to be lower case.
2130 gcc/testsuite/gcc.target/msp430/
2131 * msp430i-device-symbol.c: New test.
2132 * msp430f-device-symbol.c: New test.
2133 * msp430.h: New test header file.
2135 2018-06-07 Richard Biener <rguenther@suse.de>
2137 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2138 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2139 Properly wrap signed arithmetic if overflow wraps.
2141 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2143 PR tree-optimization/69615
2144 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2145 of a cast from a same precision integral SSA_NAME in a bb dominated
2146 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2147 cast to utype if rhs2 has already a compatible type.
2149 2018-06-07 Richard Biener <rguenther@suse.de>
2151 PR tree-optimization/85935
2152 * graphite-scop-detection.c (find_params_in_bb): Analyze
2153 condition operands with respect to the correct loop. Assert
2154 the analysis doesn't fail.
2156 2018-06-04 Carl Love <cel@us.ibm.com>
2158 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2159 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2160 as it is slightly cheaper.
2161 (first_match_or_eos_index_<mode>):
2162 Calculate index using natural element order.
2163 (first_match_index_<mode>):
2164 Calculate index using natural element order.
2165 (first_match_or_eos_index_<mode>):
2166 Calculate index using natural order.
2167 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2168 for BE and LE modes.
2169 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2170 P9V_BUILTIN_VCLZLSBB_V16QI.
2171 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2174 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2176 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2177 indentation and line wrap for many prototypes. Add missing
2178 @smallexample directives around block of prototypes for vec_xl and
2181 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2183 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2184 track if we pass or return IEEE 128-bit floating point.
2185 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2186 C++ mangling that is compatible with GCC 8.1.
2187 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2188 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2189 floating point types.
2190 (rs6000_function_arg_advance_1): Likewise.
2191 (rs6000_mangle_type): Optionally generate mangled names that match
2192 what GCC 8.1 generated for IEEE 128-bit floating point types.
2193 (rs6000_globalize_decl_name): If we have an external function that
2194 passes or returns IEEE 128-bit types, generate a weak reference
2195 from the mangled name used in GCC 8.1 to the current mangled
2197 (rs6000_init_builtins): Make __ibm128 use the long double type if
2198 long double is IBM extended double. Make __float128 use the long
2199 double type if long double is IEEE 128-bit.
2202 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2203 macro for __ibm128 built-in functions.
2204 (PACK_IF): Add __ibm128 pack/unpack functions.
2205 (UNPACK_IF): Likewise.
2206 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2207 enable long double built-in functions if long double is IEEE
2208 128-bit floating point.
2209 (rs6000_invalid_builtin): Update long double built-in function
2211 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2212 functions, adjust the built-in function to use the long double
2213 built-in function if __ibm128 and long double are the same type.
2214 * doc/extend.texi (PowerPC builtins): Update documention for
2215 __builtin_{,un}pack_longdouble. Add documentation for
2216 __builtin_{,un}pack_ibm128.
2218 2018-06-06 Jim Wilson <jimw@sifive.com>
2220 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2221 (struct machine_function): New field interrupt_mode.
2222 (riscv_handle_type_attribute): New function. Add forward declaration.
2223 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2224 (riscv_expand_epilogue): Check interrupt_mode field.
2225 (riscv_set_current_function): Check interrupt attribute args and
2226 set interrupt_mode field.
2227 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2228 (riscv_sret, riscv_uret): New.
2229 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2230 new arguments to interrupt attribute.
2232 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2235 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2236 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2238 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2240 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2241 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2242 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2243 several redundant entries.
2245 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2247 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2248 type from "rtx" to "rtx_insn *".
2249 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2250 for local "call_insn", removing cast.
2251 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2253 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2255 PR tree-optimization/86066
2256 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2257 for BIT_INSERT_EXPR stores.
2259 2018-06-06 Richard Biener <rguenther@suse.de>
2261 PR tree-optimization/86062
2262 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2263 component refs ontop
2264 of to be offsetted base.
2266 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2268 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2269 to be static and remove check on interrupt attribute name.
2271 gcc/testsuite/gcc.target/msp430/
2272 * function-attributes-4.c: New test.
2273 * static-interrupts.c: New test.
2275 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2277 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2278 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2280 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2283 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2285 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2286 Remove second argument, change how error is called.
2287 (aarch64_layout_arg): Remove second argument from
2288 aarch64_err_no_fpadvsimd call.
2289 (aarch64_init_cumulative_args): Ditto.
2290 (aarch64_gimplify_va_arg_expr): Ditto.
2291 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2293 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2295 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2296 (*simple_return_indirect_internal<mode>): Rename from
2297 simple_return_indirect_internal. Use W mode iterator.
2298 (rstorssp): New expander.
2299 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2300 (clrssbsy): New expander.
2301 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2303 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2305 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2307 (cmse_check_pointed_object): Likewise.
2309 2018-06-05 Martin Liska <mliska@suse.cz>
2311 PR gcov-profile/47618
2312 * doc/invoke.texi: Document how -fprofile-dir format
2315 2018-06-05 Richard Biener <rguenther@suse.de>
2317 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2318 removal pretend DOM info isn't available so we do not update
2319 it and only remove edges, not dominated blocks. Actually free
2320 DOM info in case we removed something. Remove unreachable blocks.
2321 (mfb_keep_latches): Work with either DOM info or marked backedges.
2322 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2323 first. Mark backedges if DOM info isn't available.
2324 (Re-)compute DOM info after cleanup_control_flow_pre.
2326 2018-06-05 Richard Biener <rguenther@suse.de>
2328 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2329 (locus_discrim_hasher::hash): Adjust.
2330 (locus_discrim_hasher::equal): Likewise.
2331 (next_discriminator_for_locus): Work on line directly.
2332 (same_line_p): Pass in expanded locus1 as well.
2333 (assign_discriminators): Avoid redundant location expansions.
2335 2018-06-05 Richard Biener <rguenther@suse.de>
2337 PR tree-optimization/86046
2338 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2339 if required after clearing TREE_ADDRESSABLE.
2341 2018-06-05 Richard Biener <rguenther@suse.de>
2343 PR tree-optimization/86047
2344 * tree-ssa-loop.c (for_each_index): Glob handling of all
2345 decls and constants and really handle all of them.
2347 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2350 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2351 qualifier_void_pointer and qualifier_const_void_pointer.
2352 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2353 (arm_init_builtins): Handle the above.
2354 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2355 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2356 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2359 2018-06-05 Martin Liska <mliska@suse.cz>
2361 * auto-profile.c (read_autofdo_file): Do not use
2362 gcov_ctr_summary struct.
2363 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2364 * coverage.c (struct counts_entry): Likewise.
2365 (read_counts_file): Read just single summary entry.
2366 (get_coverage_counts): Use gcov_summary struct.
2367 * coverage.h (get_coverage_counts): Likewise.
2368 * gcov-dump.c (dump_working_sets): Likewise.
2369 (tag_summary): Dump just single summary.
2370 * gcov-io.c (gcov_write_summary): Write just histogram
2372 (gcov_read_summary): Read just single summary.
2373 (compute_working_sets): Use gcov_summary struct.
2374 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2375 of GCOV_COUNTERS_SUMMABLE.
2376 (GCOV_COUNTERS_SUMMABLE): Remove.
2377 (GCOV_FIRST_VALUE_COUNTER): Replace with
2378 GCOV_COUNTER_V_INTERVAL.
2379 (struct gcov_ctr_summary): Remove.
2380 (struct gcov_summary): Directly use fields of former
2382 (compute_working_sets): Use gcov_summary struct.
2383 * gcov.c (read_count_file): Do not use ctrs fields.
2384 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2386 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2388 * profile.c: Likewise.
2389 * profile.h: Likewise.
2391 2018-06-05 Martin Liska <mliska@suse.cz>
2393 PR gcov-profile/84846
2394 * gcov.c (output_lines): Print working directory only
2395 in intermediate format.
2397 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2399 * config/s390/s390-builtin-types.def: Add void function type.
2400 * config/s390/s390-builtins.def: Use the function type for the
2403 2018-06-04 Jim Wilson <jimw@sifive.com>
2405 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2407 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2408 and maybe_eh_return. Change regno to unsigned int. Use new args to
2409 handle EH_RETURN_DATA_REGNO registers properly.
2410 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2411 (riscv_expand_epilogue): Update comment. Change argument name and
2412 type. Update code to use new name and type. Pass new args to
2413 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2415 * config/riscv/riscv.md (NORMAL_RETURN): New.
2416 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2417 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2418 (eh_return): Call gen_eh_return_internal and emit barrier.
2419 (eh_return_internal): Call riscv_expand_epilogue.
2421 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2423 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2424 bit_insertion field and declare can_be_merged_into method.
2425 (merged_store_group::can_be_merged_into): New method.
2426 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2427 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2428 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2430 2018-06-04 Richard Biener <rguenther@suse.de>
2432 PR tree-optimization/85955
2433 * builtins.c (fold_builtin_sincos): Convert pointers to
2434 destination to appropriate type before dereferencing.
2436 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2438 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2440 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2442 * expr.c (expand_expr_real_1): Force the operand into memory if
2443 its TYPE_MODE is BLKmode and if there is no integer mode for
2444 the number of bits being extracted.
2446 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2450 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2451 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2453 2018-06-04 Richard Biener <rguenther@suse.de>
2455 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2456 (cleanup_tree_cfg_noloop): ... single caller. Do
2457 start_recording_case_labels later.
2459 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2461 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2462 to _IMMINTRIN_H_INCLUDED.
2463 * config/i386/pconfigintrin.h: Ditto.
2464 * config/i386/waitpkgintrin.h: Ditto.
2465 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2466 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2467 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2468 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2469 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2470 waitpkgintrin.h and cldemoteintrin.h.
2472 2018-06-04 Richard Biener <rguenther@suse.de>
2474 PR tree-optimization/86038
2475 * tracer.c (find_best_successor): Check probability for
2476 being initialized, bail out if not.
2478 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2481 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2482 of bits to ignore when comparing architectures.
2484 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2486 PR tree-optimization/69615
2487 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2488 maximum or minimum of the type, try to merge it also as if
2489 range1 is + [-, x - 1] or + [x + 1, -].
2492 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2494 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2496 PR tree-optimization/86034
2497 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2498 the unsigned bitfield type in a bit insertion sequence if it does not
2499 have a larger precision than the bitfield size.
2500 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2502 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2504 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2506 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2508 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2509 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2510 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2511 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2513 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2515 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2516 Disable -fdelete-null-pointer-checks for ELF toolchain.
2518 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2519 Kito Cheng <kito.cheng@gmail.com>
2521 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2522 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2523 (nds32*-*-*): Add glibc and uclibc conditions.
2524 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2525 (TARGET_EXCEPT_UNWIND_INFO): Define.
2526 * config/nds32/elf.h: New file.
2527 * config/nds32/linux.h: New file.
2528 * config/nds32/nds32-elf.opt: New file.
2529 * config/nds32/nds32-linux.opt: New file.
2530 * config/nds32/nds32-fp-as-gp.c
2531 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2532 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2534 (nds32_asm_file_end): Ditto.
2535 (nds32_print_operand): Ditto.
2536 (nds32_insert_attributes): Ditto.
2537 (nds32_init_libfuncs): New function.
2538 (TARGET_HAVE_TLS): Define.
2539 (TARGET_INIT_LIBFUNCS): Define.
2540 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2542 (TARGET_ELF): Apply different mcmodel setting.
2543 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2544 been migrated into elf.h and linux.h files.
2545 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2546 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2547 (mcmodel): The content has been migrated into nds32-elf.opt and
2548 nds32-linux.opt files.
2549 * config/nds32/t-elf: New file.
2550 * config/nds32/t-linux: New file.
2552 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2553 Shiva Chen <shiva0217@gmail.com>
2555 * config/nds32/constants.md (unspec_volatile_element): Add
2556 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2557 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2559 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2560 (make_pass_nds32_fp_as_gp): Declare.
2561 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2563 (nds32_asm_function_end_prologue): Remove unused asm output.
2564 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2565 (nds32_asm_file_start): Output necessary fp_as_gp information.
2566 (nds32_option_override): Adjust register usage.
2567 (nds32_expand_prologue): Consider fp_as_gp situation.
2568 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2569 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2572 (simple_return): Ditto.
2573 (omit_fp_begin): Output special directive for fp_as_gp.
2574 (omit_fp_end): Output special directive for fp_as_gp.
2575 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2576 mforbid-fp-as-gp): New options.
2578 2018-06-01 Mark Wielaard <mark@klomp.org>
2580 * dwarf2out.c (dwarf2out_finish): Remove generation of
2581 DW_AT_loclists_base.
2583 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2585 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2586 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2587 (struct merged_store_group): Add bit_insertion field.
2588 (dump_char_array): Use standard hexadecimal format.
2589 (merged_store_group::merged_store_group): Set bit_insertion to false.
2590 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2591 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2592 also print the mask in the dump file.
2593 (pass_store_merging::gate): Minor tweak.
2594 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2595 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2596 stores with INTEGER_CST stores.
2597 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2598 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2599 and use it throughout. Generate bit insertion sequences if need be.
2600 (pass_store_merging::process_store): Remove redundant condition.
2601 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2603 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2605 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2606 the 128-bit floating point types. Fix function comment.
2608 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2610 * config/aarch64/aarch64-simd.md
2611 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2613 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2616 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2618 PR tree-optimization/85989
2619 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2621 (backprop::intersect_uses): Check it when deciding whether this
2622 is a backedge reference.
2623 (backprop::process_block): Add each phi to m_visited_phis
2624 after visiting it, then clear it at the end.
2626 2018-06-01 Richard Biener <rguenther@suse.de>
2628 * tree-vectorizer.h (vect_dr_stmt): New function.
2629 (vect_get_load_cost): Adjust.
2630 (vect_get_store_cost): Likewise.
2631 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2632 Use vect_dr_stmt instead of DR_SMTT.
2633 (vect_record_base_alignments): Likewise.
2634 (vect_calculate_target_alignment): Likewise.
2635 (vect_compute_data_ref_alignment): Likewise and make static.
2636 (vect_update_misalignment_for_peel): Likewise.
2637 (vect_verify_datarefs_alignment): Likewise.
2638 (vector_alignment_reachable_p): Likewise.
2639 (vect_get_data_access_cost): Likewise. Pass down
2640 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2641 (vect_get_peeling_costs_all_drs): Likewise.
2642 (vect_peeling_hash_get_lowest_cost): Likewise.
2643 (vect_enhance_data_refs_alignment): Likewise.
2644 (vect_find_same_alignment_drs): Likewise.
2645 (vect_analyze_data_refs_alignment): Likewise.
2646 (vect_analyze_group_access_1): Likewise.
2647 (vect_analyze_group_access): Likewise.
2648 (vect_analyze_data_ref_access): Likewise.
2649 (vect_analyze_data_ref_accesses): Likewise.
2650 (vect_vfa_segment_size): Likewise.
2651 (vect_small_gap_p): Likewise.
2652 (vectorizable_with_step_bound_p): Likewise.
2653 (vect_prune_runtime_alias_test_list): Likewise.
2654 (vect_analyze_data_refs): Likewise.
2655 (vect_supportable_dr_alignment): Likewise.
2656 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2657 (vect_gen_prolog_loop_niters): Likewise.
2658 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2659 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2661 (vect_recog_mask_conversion_pattern): Likewise.
2662 (vect_try_gather_scatter_pattern): Likewise.
2663 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2664 to vect_get_store_cost.
2665 (vect_get_store_cost): Get stmt_info instead of DR.
2666 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2667 (vect_get_load_cost): Get stmt_info instead of DR.
2669 2018-06-01 Richard Biener <rguenther@suse.de>
2672 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2673 vars wrapped in MEM_REFs.
2675 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2677 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2678 Fix subreg tests so that we only return a choice between
2679 GENERAL_REGS and FP_REGS if the original classes included both.
2681 2018-06-01 Richard Biener <rguenther@suse.de>
2684 * tree-ssa-structalias.c (get_function_part_constraint):
2685 Handle NULL fi->decl.
2686 (find_func_aliases_for_call): Properly handle indirect
2687 fi from direct call.
2688 (find_func_clobbers): Likewise.
2689 (ipa_pta_execute): Likewise.
2690 (create_variable_info_for): For functions that are ifunc_resolver
2691 resolve to a varinfo that contains the result of the resolver call.
2692 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2695 2018-05-31 Michael Collison <michael.collison@arm.com>
2697 * config/aarch64/aarch64.md:
2698 (*fix_to_zero_extenddfdi2): New pattern.
2699 * gcc.target/aarch64/fix_extend1.c: New testcase.
2701 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2705 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2706 and BUILT_IN_STRNCMP_EQ.
2707 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2708 BUILT_IN_STRNCMP_EQ.
2709 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2710 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2711 (gimple_fold_builtin): Likewise.
2712 * tree-ssa-strlen.c (compute_string_length): New function.
2713 (determine_min_obsize): New function.
2714 (handle_builtin_string_cmp): New function to handle calls to
2715 string compare functions.
2716 (strlen_optimize_stmt): Add handling to builtin string compare
2718 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2719 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2720 * tree.c (build_common_builtin_nodes): Add new defines of
2721 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2723 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2726 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2727 functions with naked attribute.
2729 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2731 * config/i386/sse.md (avx_vec_concat<mode>):
2732 Substitute concat_tg_mode mode attribute with xtg_mode.
2733 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2734 (concat_tg_mode): Remove mode attribute.
2736 2018-05-31 Martin Sebor <msebor@redhat.com>
2739 * calls.c (alloc_max_size): Correct a logic error/typo.
2740 Treat excessive arguments as infinite. Warn for invalid arguments.
2741 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2743 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2746 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2747 and movx for Haswell.
2749 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2750 Cesar Philippidis <cesar@codesourcery.com>
2753 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2754 when emitting error on private/firstprivate reductions.
2755 * omp-low.c (lower_omp_target): Avoid reference-type processing
2756 on pointers for firstprivate clause.
2758 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2760 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2763 * config/aarch64/aarch64-simd.md
2764 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2765 (aarch64_ld1_x3_<mode>): Likewise
2766 (aarch64_st1x2<VALLDIF:mode>): Likewise
2767 (aarch64_st1_x2_<mode>): Likewise
2768 (aarch64_st1x3<VALLDIF:mode>): Likewise
2769 (aarch64_st1_x3_<mode>): Likewise
2770 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2771 (vld1_s8_x3): Likewise.
2772 (vld1_u16_x3): Likewise.
2773 (vld1_s16_x3): Likewise.
2774 (vld1_u32_x3): Likewise.
2775 (vld1_s32_x3): Likewise.
2776 (vld1_u64_x3): Likewise.
2777 (vld1_s64_x3): Likewise.
2778 (vld1_f16_x3): Likewise.
2779 (vld1_f32_x3): Likewise.
2780 (vld1_f64_x3): Likewise.
2781 (vld1_p8_x3): Likewise.
2782 (vld1_p16_x3): Likewise.
2783 (vld1_p64_x3): Likewise.
2784 (vld1q_u8_x3): Likewise.
2785 (vld1q_s8_x3): Likewise.
2786 (vld1q_u16_x3): Likewise.
2787 (vld1q_s16_x3): Likewise.
2788 (vld1q_u32_x3): Likewise.
2789 (vld1q_s32_x3): Likewise.
2790 (vld1q_u64_x3): Likewise.
2791 (vld1q_s64_x3): Likewise.
2792 (vld1q_f16_x3): Likewise.
2793 (vld1q_f32_x3): Likewise.
2794 (vld1q_f64_x3): Likewise.
2795 (vld1q_p8_x3): Likewise.
2796 (vld1q_p16_x3): Likewise.
2797 (vld1q_p64_x3): Likewise.
2798 (vst1_s64_x2): Likewise.
2799 (vst1_u64_x2): Likewise.
2800 (vst1_f64_x2): Likewise.
2801 (vst1_s8_x2): Likewise.
2802 (vst1_p8_x2): Likewise.
2803 (vst1_s16_x2): Likewise.
2804 (vst1_p16_x2): Likewise.
2805 (vst1_s32_x2): Likewise.
2806 (vst1_u8_x2): Likewise.
2807 (vst1_u16_x2): Likewise.
2808 (vst1_u32_x2): Likewise.
2809 (vst1_f16_x2): Likewise.
2810 (vst1_f32_x2): Likewise.
2811 (vst1_p64_x2): Likewise.
2812 (vst1q_s8_x2): Likewise.
2813 (vst1q_p8_x2): Likewise.
2814 (vst1q_s16_x2): Likewise.
2815 (vst1q_p16_x2): Likewise.
2816 (vst1q_s32_x2): Likewise.
2817 (vst1q_s64_x2): Likewise.
2818 (vst1q_u8_x2): Likewise.
2819 (vst1q_u16_x2): Likewise.
2820 (vst1q_u32_x2): Likewise.
2821 (vst1q_u64_x2): Likewise.
2822 (vst1q_f16_x2): Likewise.
2823 (vst1q_f32_x2): Likewise.
2824 (vst1q_f64_x2): Likewise.
2825 (vst1q_p64_x2): Likewise.
2826 (vst1_s64_x3): Likewise.
2827 (vst1_u64_x3): Likewise.
2828 (vst1_f64_x3): Likewise.
2829 (vst1_s8_x3): Likewise.
2830 (vst1_p8_x3): Likewise.
2831 (vst1_s16_x3): Likewise.
2832 (vst1_p16_x3): Likewise.
2833 (vst1_s32_x3): Likewise.
2834 (vst1_u8_x3): Likewise.
2835 (vst1_u16_x3): Likewise.
2836 (vst1_u32_x3): Likewise.
2837 (vst1_f16_x3): Likewise.
2838 (vst1_f32_x3): Likewise.
2839 (vst1_p64_x3): Likewise.
2840 (vst1q_s8_x3): Likewise.
2841 (vst1q_p8_x3): Likewise.
2842 (vst1q_s16_x3): Likewise.
2843 (vst1q_p16_x3): Likewise.
2844 (vst1q_s32_x3): Likewise.
2845 (vst1q_s64_x3): Likewise.
2846 (vst1q_u8_x3): Likewise.
2847 (vst1q_u16_x3): Likewise.
2848 (vst1q_u32_x3): Likewise.
2849 (vst1q_u64_x3): Likewise.
2850 (vst1q_f16_x3): Likewise.
2851 (vst1q_f32_x3): Likewise.
2852 (vst1q_f64_x3): Likewise.
2853 (vst1q_p64_x3): Likewise.
2855 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2857 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2858 user_label_prefix to name.
2860 * tree-core.h: Update comment about the format of NAME string
2861 passed to handler in attribute_spec.
2863 * config/msp430/msp430.md: Remove erroneous subreg expression from
2864 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2865 zero_extend{q,h}isi2.
2867 2018-05-30 Borislav Petkov <bp@suse.de>
2869 * doc/extend.texi: Document some architecture specific
2870 constraints and sort entries.
2872 2018-05-30 Martin Sebor <msebor@redhat.com>
2875 * builtins.c (expand_builtin_stpcpy_1): New function.
2876 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2877 only if the former succeeds.
2879 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2881 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2884 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2886 * doc/invoke.texi (-flinker-output): Document
2888 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2890 * passes.c (ipa_write_summaries): Only modify statements if body
2892 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2893 incrementally linking.
2894 (ipa_passes): Likewise.
2895 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2896 pass down resolution info.
2897 * common.opt (flag_incremental_link): Update info.
2898 * gcc.c (plugin specs): Turn flinker-output=* to
2899 -plugin-opt=-linker-output-known
2900 * toplev.c (compile_file): Also cut compilation when doing incremental
2902 * flag-types. (enum lto_partition_model): Add
2903 LTO_LINKER_OUTPUT_NOLTOREL.
2904 (invoke.texi): Add -flinker-output docs.
2905 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2906 link same way as WPA; do not stream in dead initializers.
2908 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2909 darf2out_register_external_decl): Support incremental link.
2911 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2913 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2915 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2917 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2918 it down to simple_object_copy_lto_debug_sections.
2919 (run_gcc): Determine incremental LTO link time and configure
2920 lto1 into non-wpa mode, disable renaming of debug sections.
2922 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2924 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2925 descriptions of various incorrectly documented functions.
2927 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2930 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2931 address check not strict.
2933 2018-05-30 Richard Biener <rguenther@suse.de>
2935 PR tree-optimization/85964
2936 * tracer.c (better_p): Drop initialized count check, we only
2937 call the function with initialized counts now.
2938 (find_best_successor): Do find a best edge if one
2939 has uninitialized count.
2940 (find_best_predecessor): Likewise. Do BB frequency check only
2941 if count is initialized.
2943 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2945 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2946 (aarch64_ldrstr_offset_compare): New.
2947 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2948 load/store orderings.
2949 (aarch64_gen_adjusted_ldpstp): Likewise.
2951 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2953 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2954 Check for subset of GENERAL_REGS and FP_REGS.
2955 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2958 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2960 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2961 and wi::to_poly_offset. Add the current offset and then check
2962 whether the sum fits, rather than using an unchecked addition of
2963 a checked term. Check for a shwi rather than a uhwi.
2964 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2965 (store_constructor): Use poly_int_tree_p.
2966 (expand_expr_real_1): Likewise.
2967 * function.c (assign_temp): Likewise.
2968 * fold-const.c (const_binop): Use poly_int_tree_p and
2970 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2972 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2973 to_poly_offset for MEM offsets.
2974 * ipa-icf.c (sem_variable::equals): Likewise.
2975 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2976 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2977 wi::to_poly_offset for BIT_FIELD_REF offsets.
2978 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2980 * var-tracking.c (emit_note_insn_var_location): Use
2981 tree_to_poly_uint64.
2983 2018-05-29 Jim Wilson <jimw@sifive.com>
2985 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2987 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2990 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2991 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2993 (sse4_1_round<mode>2): Use nonimmediate_operand
2994 for operand 1 predicate.
2996 2018-05-29 Martin Sebor <msebor@redhat.com>
2997 Richard Biener <rguenther@suse.de>
3000 * calls.c (get_size_range): Call determine_value_range instead
3001 of get_value_range..
3002 * tree-vrp.h (determine_value_range): Declared new function.
3003 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
3005 2018-05-29 Richard Biener <rguenther@suse.de>
3007 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
3008 sure to use non-pattern stmts for get_earlier_stmt arguments.
3009 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
3010 called on pattern stmts.
3011 (get_later_stmt): Likewise.
3013 2018-05-29 Martin Liska <mliska@suse.cz>
3015 PR gcov-profile/85759
3016 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
3019 2018-05-29 Jakub Jelinek <jakub@redhat.com>
3021 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
3023 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
3027 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3028 VEC_PACK_FLOAT_EXPR): New tree codes.
3029 * tree-pretty-print.c (op_code_prio): Handle
3030 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3031 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3032 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3033 * tree-inline.c (estimate_operator_cost): Likewise.
3034 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
3035 * fold-const.c (const_binop): Likewise.
3036 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
3037 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
3038 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3039 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
3040 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
3041 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
3042 * expr.c (expand_expr_real_2): Likewise.
3043 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
3044 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
3045 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
3047 * optabs.c (expand_widen_pattern_expr): For
3048 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
3049 sign from result type rather than operand's type.
3050 (expand_binop_directly): For vec_packu_float_optab and
3051 vec_packs_float_optab allow result type to be different from operand's
3053 * optabs-tree.c (optab_for_tree_code): Handle
3054 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3055 VEC_PACK_FLOAT_EXPR. Formatting fixes.
3056 * tree-vect-generic.c (expand_vector_operations_1): Handle
3057 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
3058 VEC_PACK_FLOAT_EXPR.
3059 * tree-vect-stmts.c (supportable_widening_operation): Handle
3061 (supportable_narrowing_operation): Handle FLOAT_EXPR.
3062 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
3063 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
3064 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
3065 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
3067 (vec_pack<floatprefix>_float_<mode>): New expander.
3068 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
3070 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
3071 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
3072 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
3073 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
3074 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
3076 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
3077 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
3078 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
3079 VEC_PACK_FLOAT_EXPR): Document.
3081 2018-05-29 Richard Biener <rguenther@suse.de>
3083 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
3085 (stmt_vec_info_vec): Make pointer.
3086 (init_stmt_vec_info_vec): Remove.
3087 (free_stmt_vec_info_vec): Likewise.
3088 (set_stmt_vec_info_vec): New function.
3089 (free_stmt_vec_infos): Likewise.
3090 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
3091 (set_vinfo_for_stmt): Likewise.
3092 (get_earlier_stmt): Likewise.
3093 (get_later_stmt): Likewise.
3094 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
3095 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
3096 (vec_info::~vec_info): Free stmt_vec_infos.
3097 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
3098 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
3099 (pass_slp_vectorize::execute): Likewise.
3100 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
3101 (free_stmt_vec_info_vec): Likewise.
3102 (set_stmt_vec_info_vec): New function.
3103 (free_stmt_vec_infos): Likewise.
3104 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
3105 the global stmt_vec_info_vec.
3106 * tree-parloops.c (gather_scalar_reductions): Use
3107 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
3110 2018-05-29 Richard Biener <rguenther@suse.de>
3112 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
3114 2018-05-29 Martin Liska <mliska@suse.cz>
3115 David Malcolm <dmalcolm@redhat.com>
3117 * vec.c (test_reverse): New.
3118 (vec_c_tests): Add new test.
3119 * vec.h (vl_ptr>::reverse): New function.
3121 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
3123 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
3125 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3128 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3130 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3132 2018-05-28 Richard Biener <rguenther@suse.de>
3134 PR tree-optimization/85933
3135 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3136 look at stmts marked as vectorizable.
3138 2018-05-28 Richard Biener <rguenther@suse.de>
3140 PR tree-optimization/85934
3141 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3142 vector boolean check before scalar optimization.
3144 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3146 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3149 2018-05-28 Mark Wielaard <mark@klomp.org>
3151 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3152 if it is an expression containing a minus sign.
3154 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3156 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3158 2018-05-27 Paul Koning <ni1d@arrl.net>
3160 * config/pdp11/pdp11.md (truncsihi2): Remove.
3162 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3163 Chung-Ju Wu <jasonwucj@gmail.com>
3165 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3167 (unaligned_store_dw): Ditto.
3168 * config/nds32/nds32-memory-manipulation.c
3169 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3170 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3171 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3172 (emit_setmem_word_loop): Rename to ...
3173 (emit_setmem_doubleword_loop): ... this.
3174 (nds32_gen_dup_4_byte_to_word_value): New function.
3175 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3176 (nds32_expand_setmem_loop): Refine implementation.
3177 (nds32_expand_setmem_loop_v3m): Ditto.
3178 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3181 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3183 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3185 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3187 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3188 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3189 machine->attr_no_prologue_p.
3190 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3191 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3192 (nds32_expand_epilogue): Consider attr_naked_p.
3193 (nds32_expand_epilogue_v3pop): Likewise.
3194 (nds32_can_use_return_insn): Likewise.
3195 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3196 attr_no_prologue_p fields.
3197 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3199 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3202 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3204 * config/i386/sse.md
3205 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3207 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3209 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3211 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3213 (*<floatsuffix>floatv2div2sf2): Rename to ...
3214 (*float<floatunssuffix>v2div2sf2): ... this.
3215 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3216 (float<floatunssuffix>v2div2sf2_mask): ... this.
3217 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3218 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3219 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3221 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3223 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3225 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3227 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3229 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3231 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3232 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3233 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3234 gen_ufix_truncv8dfv8si2.
3235 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3236 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3237 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3238 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3239 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3240 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3241 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3242 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3244 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3248 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3250 2018-05-25 Jim Wilson <jimw@sifive.com>
3252 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3253 * config/riscv/riscv.c (struct machine_function): Add
3254 interrupt_handler_p and attribute_checked_p fields.
3255 (riscv_attribute_table): Add interrupt.
3256 (riscv_interrupt_type_p): New.
3257 (riscv_save_reg_p): Save extra regs for interrupt handler.
3258 (riscv_use_save_libcall): Return false for interrupt handler.
3259 (riscv_first_stack_step): Add forward declaration.
3260 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3261 for interrupt handler with large frame. Use it for saved reg list.
3262 (riscv_expand_prologue): Move flag_stack_usage_info support to
3263 eliminate duplication.
3264 (riscv_expand_epilogue): Generate mret for interrupt handler.
3265 (riscv_epilogue_uses): New.
3266 (riscv_can_use_return_insn): Return false for interrupt handler.
3267 (riscv_function_ok_for_sibcall): Likewise.
3268 (riscv_set_current_function): Add interrupt handler support.
3269 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3270 * config/riscv/riscv.md (UNSPECV_MRET): New.
3272 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3274 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3276 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3278 PR tree-optimization/85712
3279 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3280 this candidate has already been replaced in-situ by a copy.
3282 2018-05-25 Jason Merrill <jason@redhat.com>
3284 PR c++/80485 - inline function non-zero address.
3285 * symtab.c (nonzero_address): Check DECL_COMDAT.
3287 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3290 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3291 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3292 extension of SImode operation. Use const123_operand predicate.
3293 (*saddsi_1): Remove.
3294 (*saddl_se_1): Ditto.
3296 (*ssubl_se_1): Ditto.
3297 * config/alpha/predicates.md (const123_operand): New predicate.
3298 * config/alpha/constraints.md (P): Use IN_RANGE.
3300 2018-05-25 Richard Biener <rguenther@suse.de>
3302 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3304 (ref_maybe_used_by_stmt_p): Likewise.
3305 (stmt_may_clobber_ref_p): Likewise.
3306 (stmt_may_clobber_ref_p_1): Likewise.
3307 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3309 (ref_maybe_used_by_stmt_p): Likewise.
3310 (stmt_may_clobber_ref_p): Likewise.
3311 (stmt_may_clobber_ref_p_1): Likewise.
3312 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3313 the alias oracle to disambiguate DRs with stmts DR analysis
3315 (vect_analyze_data_refs): Do not give up on not analyzable
3316 DRs for BB vectorization. Remove code truncating the dataref
3319 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3322 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3323 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3324 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3326 2018-05-25 Richard Biener <rguenther@suse.de>
3328 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3329 function, combining stmt data ref gathering and fatal analysis
3331 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3332 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3333 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3334 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3335 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3337 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3339 PR tree-optimization/85720
3340 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3341 SCC if all partitions are builtins.
3342 (version_loop_by_alias_check): New parameter. Generate cancelable
3343 runtime alias check if all partitions are builtins.
3344 (distribute_loop): Update call to above function.
3346 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3348 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3349 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3350 (parm_default_def_partition_arg): Ditto.
3351 (set_parm_default_def_partition): Ditto.
3352 (get_parm_default_def_partitions): Ditto and make it static.
3353 (get_undefined_value_partitions): Ditto and make it static.
3354 (remove_ssa_form): Refactor call to init_var_map here.
3355 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3356 computation for loop region.
3357 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3358 (register_default_def): Delete.
3359 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3360 (parm_default_def_partition_arg): Ditto.
3361 (set_parm_default_def_partition): Ditto.
3362 (get_parm_default_def_partitions): Ditto and make it static.
3363 (get_undefined_value_partitions): Ditto and make it static.
3364 (coalesce_with_default, coalesce_with_default): Update comment.
3365 (create_coalesce_list_for_region): New func factored out from
3366 create_outofssa_var_map.
3367 (populate_coalesce_list_for_outofssa): New func factored out from
3368 create_outofssa_var_map and coalesce_ssa_name.
3369 (create_outofssa_var_map): Delete.
3370 (coalesce_ssa_name): Refactor to support live range computation.
3371 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3372 (get_parm_default_def_partitions): Delete.
3373 (get_undefined_value_partitions): Ditto.
3374 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3375 computation for loop region.
3376 (new_tree_live_info, loe_visit_block): Ditto.
3377 (live_worklist, set_var_live_on_entry): Ditto.
3378 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3379 * tree-ssa-live.h (struct _var_map): New fields.
3380 (init_var_map): Change decl.
3381 (region_contains_p): New.
3383 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3385 * tree-ssa-live.h (live_merge_and_clear): Delete.
3387 2018-05-25 Richard Biener <rguenther@suse.de>
3390 * tree-dump.c (dequeue_and_dump): Remove access to removed
3391 operand 2 of a SWITCH_EXPR.
3393 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3395 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3396 multiplication and division.
3397 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3398 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3399 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3400 (cond_udiv_optab, cond_umod_optab): New optabs.
3401 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3402 (IFN_COND_RDIV): New internal functions.
3403 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3404 TRUNC_MOD_EXPR and RDIV_EXPR.
3405 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3406 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3408 (SVE_INT_BINARY): Include mult.
3409 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3410 (optab, sve_int_op): Handle mult.
3411 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3413 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3414 for SVE_INT_BINARY_SD.
3416 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3418 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3419 (optab, sve_int_op): Handle div and udiv.
3420 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3421 for SVE_INT_BINARY_SD.
3422 (*<optab><mode>3): New insn for the same.
3424 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3426 * tree-vect-patterns.c: Include predict.h.
3427 (vect_recog_divmod_pattern): Restrict check for division support
3428 to when optimizing for size.
3430 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3432 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3433 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3434 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3435 (gimple_match_op::set_op): Likewise.
3436 (gimple_resimplify4): Declare.
3437 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3438 (expr::gen_transform): Likewise.
3439 (decision_tree::gen): Generate a simplification routine for 4 operands.
3440 * gimple-match-head.c (gimple_simplify): Add an overload for
3441 4 operands. In the top-level function, handle up to 4 call
3442 arguments and call gimple_resimplify4.
3443 (gimple_resimplify4): New function.
3444 (build_call_internal): Pass a fourth operand.
3445 (maybe_push_to_seq): Likewise.
3446 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3447 Fold VEC_COND_EXPRs of an operation and a default value into
3448 an IFN_COND_* function if possible.
3449 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3451 (SVE_COND_FP_BINARY): Include them.
3452 (optab, sve_fp_op): Handle them.
3453 (SVE_INT_BINARY_REV): New code iterator.
3454 (SVE_COND_FP_BINARY_REV): New int iterator.
3455 (commutative): New int attribute.
3456 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3458 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3460 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3461 (*cond_<optab><mode>): New patterns for reversed operands.
3463 2018-05-25 Richard Biener <rguenther@suse.de>
3465 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3466 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3467 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3468 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3469 * tree-vect-data-refs.c (everywhere): Adjust users.
3470 * tree-vect-loop.c (everywhere): Likewise.
3471 * tree-vect-slp.c (everywhere): Likewise.
3472 * tree-vect-stmts.c (everywhere): Likewise.
3473 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3475 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3477 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3479 (gcc_cv_as_section_exclude): ... this.
3480 Try Solaris as #exclude syntax.
3481 * configure: Regenerate.
3482 * config.in: Regenerate.
3483 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3485 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3486 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3488 * varasm.c (default_elf_asm_named_section): Don't check if
3489 HAVE_GAS_SECTION_EXCLUDE is defined.
3491 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3493 * doc/md.texi: Update the documentation of the cond_* optabs
3494 to mention the new final operand. Fix GET_MODE_NUNITS call.
3495 Describe the scalar case too.
3496 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3497 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3499 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3500 (get_conditional_internal_fn): Update comment.
3501 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3502 accumulator value as a final argument to conditional functions.
3503 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3504 a define_expand and add an "else" operand. Assert for now that
3505 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3506 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3507 (*cond_<optab><mode>): New patterns.
3508 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3509 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3510 (UNSPEC_COND_EOR): Delete.
3511 (optab): Remove associated mappings.
3512 (SVE_INT_BINARY): New code iterator.
3513 (sve_int_op): Remove int attribute and add "minus" to the code
3515 (SVE_COND_INT_OP): Delete.
3516 (SVE_COND_FP_OP): Rename to...
3517 (SVE_COND_FP_BINARY): ...this.
3519 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3521 * optabs.c (can_reuse_operands_p): New function.
3522 (maybe_legitimize_operands): Try to reuse the results for
3525 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3527 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3528 Add {q} suffix to insn mnemonic.
3530 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3532 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3533 (msp430_warn_func_return): New.
3535 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3537 * fold-const.c (tree_nonzero_bits): New function.
3538 * fold-const.h (tree_nonzero_bits): Likewise.
3539 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3540 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3542 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3546 * varasm.c (assemble_alias): Check ifunc_resolver only on
3549 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3552 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3553 when memory input operand is handled.
3555 2018-05-24 Luis Machado <luis.machado@linaro.org>
3557 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3559 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3561 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3563 * match.pd: Delay FMA folds until after vectorization.
3565 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3568 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3569 address check not strict.
3571 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3573 * gimple-match.h (gimple_match_op): New class.
3574 (mprts_hook): Replace parameters with a gimple_match_op *.
3575 (maybe_build_generic_op): Likewise.
3576 (gimple_simplified_result_is_gimple_val): Replace parameters with
3577 a const gimple_match_op *.
3578 (gimple_simplify): Replace code_helper * and tree * parameters with
3579 a gimple_match_op * parameter.
3580 (gimple_resimplify1): Replace code_helper *, tree and tree *
3581 parameters with a gimple_match_op * parameter.
3582 (gimple_resimplify2): Likewise.
3583 (gimple_resimplify3): Likewise.
3584 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3585 parameters with a gimple_match_op * parameter.
3586 * gimple-match-head.c (gimple_simplify): Change prototypes of
3587 auto-generated functions to take a gimple_match_op * instead of
3588 separate code_helper * and tree * parameters. Make the same
3589 change in the top-level overload and update calls to the
3590 gimple_resimplify routines. Update calls to the auto-generated
3591 functions and to maybe_push_res_to_seq in the publicly-facing
3592 operation-specific gimple_simplify overloads.
3593 (gimple_match_op::MAX_NUM_OPS): Define.
3594 (gimple_resimplify1): Replace rcode and ops with a single res_op
3595 parameter. Update call to gimple_simplify.
3596 (gimple_resimplify2): Likewise.
3597 (gimple_resimplify3): Likewise.
3598 (mprts_hook): Replace parameters with a gimple_match_op *.
3599 (maybe_build_generic_op): Likewise.
3600 (build_call_internal): Replace type, nargs and ops with
3601 a gimple_match_op *.
3602 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3603 with a single gimple_match_op *. Update calls to mprts_hook,
3604 build_call_internal and gimple_simplified_result_is_gimple_val.
3605 Factor out code that is common to the tree_code and combined_fn cases.
3606 * genmatch.c (expr::gen_transform): Replace tem_code and
3607 tem_ops with a gimple_match_op called tem_op. Update calls
3608 to the gimple_resimplify functions and maybe_push_res_to_seq.
3609 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3610 res_ops. Update call to the gimple_resimplify functions.
3611 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3612 (decision_tree::gen): Make the functions take a gimple_match_op *
3613 called res_op instead of separate res_code and res_ops parameters.
3614 Update call accordingly.
3615 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3616 and ops with a single res_op parameter. Update calls to
3617 maybe_build_generic_op and maybe_push_res_to_seq.
3618 (fold_stmt_1): Update calls to gimple_simplify and
3619 replace_stmt_with_simplification.
3620 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3621 and gimple_simplified_result_is_gimple_val.
3622 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3624 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3625 with a gimple_match_op *.
3626 (vn_nary_build_or_lookup): Likewise. Update call to
3627 vn_nary_build_or_lookup_1.
3628 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3629 gimple_match_op *. Update calls to the gimple_resimplify routines
3630 and to gimple_simplified_result_is_gimple_val.
3631 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3632 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3633 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3634 (visit_nary_op): Likewise.
3635 (visit_reference_op_load): Likewise.
3637 2018-05-23 Luis Machado <luis.machado@linaro.org>
3639 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3640 modifier for printing the step amount.
3642 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3645 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3648 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3650 * doc/sourcebuild.texi (Endianness): New subsubsection.
3652 2018-05-23 Luis Machado <luis.machado@linaro.org>
3654 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3655 <prefetch_dynamic_strides>: New const bool field.
3656 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3657 prefetch_dynamic_strides.
3658 (exynosm1_prefetch_tune): Likewise.
3659 (thunderxt88_prefetch_tune): Likewise.
3660 (thunderx_prefetch_tune): Likewise.
3661 (thunderx2t99_prefetch_tune): Likewise.
3662 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3664 (aarch64_override_options_internal): Update to set
3665 PARAM_PREFETCH_DYNAMIC_STRIDES.
3666 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3667 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3668 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3669 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3670 prefetch-dynamic-strides setting.
3672 2018-05-23 Luis Machado <luis.machado@linaro.org>
3674 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3675 <minimum_stride>: New const int field.
3676 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3677 minimum_stride field defaulting to -1.
3678 (exynosm1_prefetch_tune): Likewise.
3679 (thunderxt88_prefetch_tune): Likewise.
3680 (thunderx_prefetch_tune): Likewise.
3681 (thunderx2t99_prefetch_tune): Likewise.
3682 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3683 <default_opt_level>: Set to 3.
3684 (aarch64_override_options_internal): Update to set
3685 PARAM_PREFETCH_MINIMUM_STRIDE.
3686 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3687 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3688 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3689 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3690 stride is constant and is below the minimum stride threshold.
3692 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3694 * config/arm/arm-cpus.in (mode26): Delete.
3695 (armv4): Delete mode26 reference.
3696 * config/arm/arm.c (arm_configure_build_target): Delete use of
3699 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3701 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3703 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3704 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3706 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3707 (floatunsdidf2): Ditto.
3709 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3711 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3712 (fixuns_trunc<mode>si2_avx512f): Ditto.
3713 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3714 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3715 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3717 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3719 PR rtl-optimization/79985
3720 * df-scan.c (df_insn_refs_collect): Remove special case for
3721 global registers and asm statements.
3723 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3725 * extend.texi (Global Register Variables): Rewrite the bullet list.
3726 Note that the register is available for allocation. Note that access
3727 via inline asm must use constraints. Add note about async-signal
3728 handlers. Remove paragraph about automagic register selection.
3730 2018-05-23 Richard Biener <rguenther@suse.de>
3732 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3733 of fixed offset from memset VN.
3735 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3737 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3739 (alloc_cand_and_find_basis): Initialize first_interp field.
3740 (slsr_process_mul): Modify first_interp field.
3741 (slsr_process_add): Likewise.
3742 (slsr_process_cast): Modify first_interp field for each new
3744 (slsr_process_copy): Likewise.
3745 (dump_candidate): Dump first_interp field.
3746 (replace_mult_candidate): Process all interpretations, not just
3748 (replace_rhs_if_not_dup): Likewise.
3749 (replace_one_candidate): Likewise.
3751 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3753 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3755 (aarch64_needs_frame_chain): New function.
3756 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3758 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3761 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3762 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3763 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3764 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3765 as true for strict-align.
3766 (aarch64_can_inline_p): Perform checks even when callee has no
3767 attributes to check for strict alignment.
3768 * doc/extend.texi (AArch64 Function Attributes): Document
3770 * doc/invoke.texi: (AArch64 Options): Likewise.
3772 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3774 PR tree-optimization/85853
3775 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3776 the handling of the root of the node to...
3777 (vect_slp_analyze_node_operations_1): ...this new function,
3778 and run the whole thing with the child nodes' def types
3779 set according to their SLP node's def type.
3781 2018-05-23 Richard Biener <rguenther@suse.de>
3784 * tree-data-ref.c (create_runtime_alias_checks): Defer
3785 and ignore overflow warnings.
3787 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3789 PR tree-optimization/85822
3790 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3793 2018-05-23 Richard Biener <rguenther@suse.de>
3795 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3796 memset constants via native_interpret_expr.
3798 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3801 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3803 (cgraph_node::create_alias): Likewise.
3804 (cgraph_node::get_availability): Check ifunc_resolver instead
3805 of looking up ifunc attribute.
3806 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3807 * varasm.c (do_assemble_alias): Likewise.
3808 (assemble_alias): Likewise.
3809 (default_binds_local_p_3): Likewise.
3810 * cgraph.h (cgraph_node): Add ifunc_resolver.
3811 (cgraph_node::only_called_directly_or_aliased_p): Return false
3813 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3815 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3816 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3817 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3818 instead of looking up ifunc attribute.
3820 2018-05-22 Luis Machado <luis.machado@linaro.org>
3822 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3824 2018-05-22 Martin Sebor <msebor@redhat.com>
3827 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3828 only when expasion succeeds.
3829 (expand_builtin_strcmp): Same.
3830 (expand_builtin_strncmp): Same.
3832 2018-05-22 Martin Sebor <msebor@redhat.com>
3834 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3836 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3837 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3839 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3840 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3842 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3843 Add check for legitimate_address.
3844 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3845 (aarch64_swap_ldrstr_operands): New.
3846 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3849 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3850 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3852 * config/aarch64/aarch64.md: New patterns to generate stp
3854 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3855 single words and double words.
3856 (load_pair_sw, load_pair_dw): Likewise.
3857 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3859 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3861 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3862 for different mode ldpstp and add peephole for merged zero stores.
3864 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3866 (aarch64_gen_store_pair): Rename calls to match new patterns.
3867 (aarch64_gen_load_pair): Rename calls to match new patterns.
3868 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3869 (load_pair<DREG:mode><DREG2:mode>): ... This.
3870 (store_pair<mode>): Rename to...
3871 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3872 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3874 (V_INT_EQUIV): Handle SImode.
3875 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3878 2018-05-22 Martin Sebor <msebor@redhat.com>
3881 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3882 or ajust the presumed bound on an operation to avoid unnecessary
3885 2018-05-22 Martin Sebor <msebor@redhat.com>
3887 PR tree-optimization/85826
3888 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3889 assuming that a DECL necesarily has a constant size.
3891 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3894 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3896 2018-05-22 Richard Biener <rguenther@suse.de>
3898 PR tree-optimization/85834
3899 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3900 non-constant and non-zero memset arguments.
3902 2018-05-22 Martin Liska <mliska@suse.cz>
3905 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3907 2018-05-22 Richard Biener <rguenther@suse.de>
3909 PR tree-optimization/85863
3910 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3911 comparisons when vectype is specified.
3912 (vectorizable_condition): Do not specify vectype for
3913 vect_is_simple_cond when SLP vectorizing.
3915 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3918 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3919 define __ibm128 as long double.
3920 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3921 as a distinct type when IEEE 128-bit support is enabled.
3922 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3923 128-bit types to use the correct functions.
3924 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3925 convert between 128-bit floating point types that have different
3926 modes but the same representation, instead of using gen_lowpart to
3928 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3930 (IFKF_reg): New attributes to give the register constraints for
3932 (extend<mode>tf2_internal): New insns to mark an explicit
3933 conversion between 128-bit floating point types that have a
3934 different mode but share the same representation.
3936 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3938 PR tree-optimization/85814
3939 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3940 a null return from get_strinfo when unsharing the next
3941 strinfo in the chain.
3943 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3946 * varasm.c (weak_finish): Clean up weak_decls.
3948 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3950 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3951 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3952 UNSPEC_UADALP values.
3953 * config/aarch64/iterators.md (ABAL): New int iterator.
3956 (sur): Add mappings for the above.
3957 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3959 (aarch64_<sur>abal<mode>_4): Likewise.
3960 (aarch64_<sur>adalp<mode>_3): Likewise.
3961 (<sur>sadv16qi): New define_expand.
3963 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3965 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3966 (*movdf_internal): Ditto.
3967 (*rcpsf2_sse): Ditto.
3968 (*rsqrtsf2_sse): Ditto.
3969 (*sqrt<mode>2_sse): Ditto.
3971 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3973 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3975 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3976 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3977 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3978 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3980 * config/aarch64/arm_neon.h: Likewise.
3981 * config/aarch64/iterators.md (VQ_I): New.
3983 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3985 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3987 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3990 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3992 * config/nds32/constraints.md (S): New constraint.
3993 * config/nds32/nds32.md (call_internal): Use constraint S.
3994 (call_value_internal): Likewise.
3995 (sibcall_internal): Likewise.
3996 (sibcall_value_internal): Likewise.
3998 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3999 Chung-Ju Wu <jasonwucj@gmail.com>
4001 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
4004 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
4005 Chung-Ju Wu <jasonwucj@gmail.com>
4007 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
4008 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
4009 (nds32_rtx_costs_impl): Simplify.
4010 (nds32_address_cost_impl): Simplify.
4011 (nds32_init_rtx_costs): New function.
4012 (nds32_rtx_costs_speed_prefer): Likewise.
4013 (nds32_rtx_costs_size_prefer): Likewise.
4014 (nds32_address_cost_speed_prefer): Likewise.
4015 (nds32_address_cost_speed_fwprop): Likewise.
4016 (nds32_address_cost_size_prefer): Likewise.
4017 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
4018 * config/nds32/nds32.c (nds32_option_override): Use
4019 nds32_init_rtx_costs function.
4021 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
4023 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
4024 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
4025 (TARGET_PIPELINE_N8): Likewise.
4026 (TARGET_PIPELINE_N10): Likewise.
4027 (TARGET_PIPELINE_N13): Likewise.
4028 (TARGET_PIPELINE_GRAYWOLF): Likewise.
4030 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
4032 * config/nds32/nds32-fpu.md: Update copyright year.
4034 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4036 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
4038 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4040 * config/nds32/nds32.c
4041 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
4042 * config/nds32/nds32.opt (minline-asm-r15): New option.
4044 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4046 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
4048 * config/nds32/nds32.md (abssi2): New pattern.
4050 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
4052 * config/i386/i386.md (rex64namesuffix): New mode attribute.
4053 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
4054 Merge insn pattern from sse_cvtsi2ss<round_name> and
4055 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
4056 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
4057 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
4058 using SWI48 mode iterator.
4059 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
4060 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
4061 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
4062 pattern from sse_cvttss2si<round_saeonly_name>
4063 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
4064 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
4065 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
4066 using SWI48 mode iterator.
4067 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
4068 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
4069 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
4070 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
4071 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
4072 using SWI48 mode iterator.
4073 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
4074 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
4075 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
4076 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
4077 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
4078 SWI48 mode iterator.
4079 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
4080 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
4081 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
4082 pattern from sse_cvttsd2si<round_saeonly_name>
4083 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
4085 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4087 * config/nds32/nds32-md-auxiliary.c
4088 (nds32_valid_smw_lwm_base_p): Refine.
4089 (nds32_output_smw_single_word): Refine.
4090 (nds32_output_smw_double_word): New.
4091 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
4093 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4095 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
4096 (nds32_output_stack_pop): Refine.
4097 (nds32_expand_unaligned_load): Refine.
4098 (nds32_expand_unaligned_store): Refine.
4100 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4101 Chung-Ju Wu <jasonwucj@gmail.com>
4103 * config/nds32/constants.md: Add TP_REGNUM constant.
4104 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
4105 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
4107 * config/nds32/nds32-doubleword.md: Consider flag_pic.
4108 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
4109 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
4110 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
4111 and PIC code generation.
4112 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
4114 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
4116 * config/nds32/nds32.md: Support TLS and PIC.
4117 * config/nds32/nds32.c: Support TLS and PIC.
4118 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
4119 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
4122 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
4124 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
4125 mode with E_ prefix.
4127 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4128 Chung-Ju Wu <jasonwucj@gmail.com>
4130 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4131 * config/nds32/nds32-md-auxiliary.c
4132 (symbolic_reference_mentioned_p): New.
4133 (nds32_legitimize_ict_address): New.
4134 (nds32_expand_ict_move): New.
4135 (nds32_indirect_call_referenced_p): New.
4136 (nds32_symbol_binds_local_p): Delete.
4137 (nds32_long_call_p): Modify.
4138 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4139 * config/nds32/nds32-protos.h
4140 (symbolic_reference_mentioned_p): Declare.
4141 (nds32_legitimize_ict_address): Declare.
4142 (nds32_expand_ict_move): Declare.
4143 (nds32_indirect_call_referenced_p): Declare.
4144 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4145 (nds32_relax_group): Use nds32_ict_const_p as condition.
4146 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4147 (nds32_asm_file_start): Output ict_model directive in asm code.
4148 (nds32_legitimate_address_p): Consider indirect call.
4149 (nds32_print_operand): Consider indirect call.
4150 (nds32_print_operand_address): Consider indirect call.
4151 (nds32_insert_attributes): Handle "indirect_call" attribute.
4152 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4153 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4154 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4155 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4156 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4157 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4158 (TARGET_ICT_MODEL_SMALL): Define.
4159 (TARGET_ICT_MODEL_LARGE): Define.
4160 * config/nds32/nds32.md (movsi): Consider ict model.
4161 (call, call_value): Consider ict model.
4162 (sibcall, sibcall_value): Consider ict model.
4163 * config/nds32/nds32.opt (mict-model): New option.
4164 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4167 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4168 Monk Chiang <sh.chiang04@gmail.com>
4169 Jim Wilson <jimw@sifive.com>
4171 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4172 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4173 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4174 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4175 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4176 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4177 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4178 compute save_libcall_adjustment properly.
4179 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4180 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4181 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4182 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4183 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4184 (ABI_SPEC): Handle mabi=ilp32e.
4185 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4186 (RVE): Add RVE mask.
4187 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4188 <-march>: Add rv32e as an example.
4190 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4193 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4194 (intra_create_variable_infos): Handle C++ constructors.
4196 2018-05-18 Martin Liska <mliska@suse.cz>
4198 * passes.def: Remove a redundant pass.
4200 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4203 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4205 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4207 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4209 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4211 (armv2, armv2a, armv3, armv3m): Delete architectures.
4212 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4213 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4214 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4216 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4217 (*mulsidi3adddi): Likewise.
4218 (mulsidi3): Likewise.
4219 (*mulsidi3_nov6): Likewise.
4220 (umulsidi3): Likewise.
4221 (umulsidi3_nov6): Likewise.
4222 (umaddsidi4): Likewise.
4223 (*umulsidi3adddi): Likewise.
4224 (smulsi3_highpart): Likewise.
4225 (*smulsi3_highpart_nov6): Likewise.
4226 (umulsi3_highpart): Likewise.
4227 (*umulsi3_highpart_nov6): Likewise.
4228 * config/arm/arm.h (arm_arch3m): Delete.
4229 * config/arm/arm.c (arm_arch3m): Delete.
4230 (arm_option_override_internal): Update armv3-related comment.
4231 (arm_configure_build_target): Delete use of isa_bit_mode32.
4232 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4233 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4234 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4236 (mulusa3): Likewise.
4237 * config/arm/arm-protos.h (arm_arch3m): Delete.
4238 * config/arm/arm-tables.opt: Regenerate.
4239 * config/arm/arm-tune.md: Likewise.
4240 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4241 deleted architectures.
4243 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4245 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4246 (armv5t, armv5te): New features.
4247 (ARMv5, ARMv5e): Delete fgroups.
4248 (ARMv5t, ARMv5te): Adjust for above changes.
4250 (armv5, armv5e): Delete arches.
4251 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4253 (*call_reg_arm): Likewise.
4254 (*call_value_reg_armv5): Likewise.
4255 (*call_value_reg_arm): Likewise.
4256 (*call_symbol): Likewise.
4257 (*call_value_symbol): Likewise.
4258 (*sibcall_insn): Likewise.
4259 (*sibcall_value_insn): Likewise.
4261 (prefetch): Likewise.
4262 (define_split and define_peephole2 dependent on arm_arch5):
4264 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4266 (TARGET_ARM_QBIT): Likewise.
4267 (TARGET_DSP_MULTIPLY): Likewise.
4268 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4269 (arm_arch5, arm_arch5e): Delete.
4270 (arm_arch5t, arm_arch5te): Declare.
4271 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4272 (arm_arch5t): Declare.
4273 (arm_option_reconfigure_globals): Update for the above.
4274 (arm_options_perform_arch_sanity_checks): Update comment, replace
4275 use of arm_arch5 with arm_arch5t.
4276 (use_return_insn): Likewise.
4277 (arm_emit_call_insn): Likewise.
4278 (output_return_instruction): Likewise.
4279 (arm_final_prescan_insn): Likewise.
4280 (arm_coproc_builtin_available): Likewise.
4281 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4282 arm_arch5e with arm_arch5t and arm_arch5te.
4283 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4284 (arm_arch5t, arm_arch5te): Declare.
4285 * config/arm/arm-tables.opt: Regenerate.
4286 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4287 * config/arm/t-multilib: Likewise.
4288 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4289 instead of arm_arch5.
4290 (*call_reg_thumb1): Likewise.
4291 (*call_value_reg_thumb1_v5): Likewise.
4292 (*call_value_reg_thumb1): Likewise.
4293 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4295 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4297 2018-05-18 Martin Liska <mliska@suse.cz>
4299 PR gcov-profile/84846
4300 * doc/gcov.texi: Document -t option of gcov tool.
4302 2018-05-18 Martin Liska <mliska@suse.cz>
4304 PR gcov-profile/84846
4305 * gcov.c (print_usage): Add new -t option.
4306 (process_args): Handle the option.
4307 (generate_results): Use stdout as output when requested by
4310 2018-05-18 Martin Liska <mliska@suse.cz>
4312 PR gcov-profile/84846
4313 * coverage.c (coverage_init): Write PWD to .gcno file.
4314 * doc/gcov.texi: Document how working directory is printed.
4315 * gcov-dump.c (dump_gcov_file): Print PWD.
4316 * gcov.c (output_intermediate_file): Likewise.
4317 (read_graph_file): Read PWD string.
4318 (output_lines): Print PWD.
4320 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4323 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4324 for retval and return false if all args to phi are zero.
4326 2018-05-18 Richard Biener <rguenther@suse.de>
4328 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4330 (evrp_dom_walker::before_dom_children): Call it.
4332 2018-05-18 Richard Biener <rguenther@suse.de>
4334 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4335 results when processing array refs with variable index.
4337 2018-05-18 Toon Moene <toon@moene.org>
4339 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4340 directly after that of -floop-interchange. Indicate that both
4341 options are enabled by default when specifying -O3.
4343 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4345 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4346 iterator. Delete separate integer-mode vec_set<mode> expander.
4347 (aarch64_simd_vec_setv2di): Delete.
4348 (vec_setv2di): Delete.
4349 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4350 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4351 the "w, r" alternative.
4353 2018-05-18 Martin Liska <mliska@suse.cz>
4355 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4356 * tree-pass.h (make_pass_lower_switch_O0): New function.
4357 * tree-switch-conversion.c (node_has_low_bound): Remove.
4358 (node_has_high_bound): Likewise.
4359 (node_is_bounded): Likewise.
4360 (class pass_lower_switch): Make it a template type and create
4362 (pass_lower_switch::execute): Add template argument.
4363 (make_pass_lower_switch): New function.
4364 (make_pass_lower_switch_O0): New function.
4365 (do_jump_if_equal): Remove.
4366 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4367 all the hard work to tree optimization passes.
4369 2018-05-18 Martin Liska <mliska@suse.cz>
4371 * dbgcnt.c (limit_low): Renamed from limit.
4372 (limit_high): New variable.
4373 (dbg_cnt_is_enabled): Check for upper limit.
4374 (dbg_cnt): Adjust dumping.
4375 (dbg_cnt_set_limit_by_index): Add new argument for high
4377 (dbg_cnt_set_limit_by_name): Likewise.
4378 (dbg_cnt_process_single_pair): Parse new format.
4379 (dbg_cnt_process_opt): Use strtok.
4380 (dbg_cnt_list_all_counters): Remove 'value' and add
4382 * doc/invoke.texi: Document changes.
4384 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4386 * doc/sourcebuild.texi (scalar_all_fma): Document.
4387 * tree.def (FMA_EXPR): Delete.
4388 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4389 * internal-fn.c (ternary_direct): New macro.
4390 (expand_ternary_optab_fn): Likewise.
4391 (direct_ternary_optab_supported_p): Likewise.
4392 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4393 * builtins.c (fold_builtin_fma): Delete.
4394 (fold_builtin_3): Don't call it.
4395 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4396 * expr.c (expand_expr_real_2): Likewise.
4397 * fold-const.c (operand_equal_p): Likewise.
4398 (fold_ternary_loc): Likewise.
4399 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4400 * gimple.c (DEFTREECODE): Likewise.
4401 * gimplify.c (gimplify_expr): Likewise.
4402 * optabs-tree.c (optab_for_tree_code): Likewise.
4403 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4404 * tree-eh.c (operation_could_trap_p): Likewise.
4405 (stmt_could_throw_1_p): Likewise.
4406 * tree-inline.c (estimate_operator_cost): Likewise.
4407 * tree-pretty-print.c (dump_generic_node): Likewise.
4408 (op_code_prio): Likewise.
4409 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4410 * tree-ssa-operands.c (get_expr_operands): Likewise.
4411 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4412 * fold-const-call.h (fold_fma): Delete.
4413 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4414 CFN_FNMA and CFN_FNMS.
4416 * genmatch.c (combined_fn): New enum.
4417 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4418 (commutative_op): New function.
4419 (commutate): Use it. Handle more than 2 operands.
4420 (dt_operand::gen_gimple_expr): Use commutative_op.
4421 (parser::parse_expr): Allow :c to be used with non-binary
4422 operators if the commutative operand is known.
4423 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4424 CFN_FMS, CFN_FNMA and CFN_FNMS.
4425 (backprop::process_assign_use): Remove FMA_EXPR handling.
4426 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4427 (gen_hsa_fma): New function.
4428 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4429 IFN_FNMA and IFN_FNMS.
4430 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4431 * gimple-fold.h (follow_all_ssa_edges): Declare.
4432 * gimple-fold.c (follow_all_ssa_edges): New function.
4433 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4434 gimple_build interface and use follow_all_ssa_edges to fold the result.
4435 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4436 instead of checking for optabs directly.
4437 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4438 rather than FMA_EXPRs.
4439 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4440 call to IFN_FMA instead of an FMA_EXPR.
4442 2018-05-17 Jim Wilson <jimw@sifive.com>
4444 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4445 sign extended promoted subreg, and the range does not have the sign bit
4446 set, then do a sign extend.
4448 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4449 test, check for sign extended subreg and/or constant operands, and
4450 do a sign extend in that case.
4452 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4454 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4455 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4457 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4458 Change logics_shift_reg to logics_shift_imm.
4459 (thunderx2t99_fp_loadpair_basic): Delete.
4460 (thunderx2t99_fp_storepair_basic): Delete.
4461 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4462 (thunderx2t99_asimd_polynomial): Delete.
4463 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4464 and neon_fp_mul_d_scalar_q.
4465 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4466 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4467 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4468 (thunderx2t99_asimd_lut): Add missing tbl types.
4469 (thunderx2t99_asimd_ext): Delete.
4470 (thunderx2t99_asimd_load1_1_mult): Delete.
4471 (thunderx2t99_asimd_load1_2_mult): Delete.
4472 (thunderx2t99_asimd_load1_ldp): New.
4473 (thunderx2t99_asimd_load1): New.
4474 (thunderx2t99_asimd_load2): Add missing *load2* types.
4475 (thunderx2t99_asimd_load3): New.
4476 (thunderx2t99_asimd_load4): New.
4477 (thunderx2t99_asimd_store1_1_mult): Delete.
4478 (thunderx2t99_asimd_store1_2_mult): Delete.
4479 (thunderx2t99_asimd_store2_mult): Delete.
4480 (thunderx2t99_asimd_store2_onelane): Delete.
4481 (thunderx2t99_asimd_store_stp): New.
4482 (thunderx2t99_asimd_store1): New.
4483 (thunderx2t99_asimd_store2): New.
4484 (thunderx2t99_asimd_store3): New.
4485 (thunderx2t99_asimd_store4): New.
4487 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4489 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4490 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4492 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4493 Segher Boessenkool <segher@kernel.crashing.org>
4496 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4499 2018-05-17 Richard Biener <rguenther@suse.de>
4501 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4502 for pruning loop and prune defs feeding only already visited PHIs.
4504 2018-05-17 Richard Biener <rguenther@suse.de>
4506 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4508 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4509 Richard Biener <rguenther@suse.de>
4511 PR tree-optimization/85793
4512 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4513 for VMAT_ELEMENTWISE.
4515 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4517 * internal-fn.h (lookup_internal_fn): Declare
4518 * internal-fn.c (lookup_internal_fn): New function.
4519 * gimple.c (gimple_build_call_from_tree): Handle calls to
4521 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4522 internal function names.
4523 * tree-pretty-print.c (dump_generic_node): Likewise.
4524 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4526 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4528 * gimple-fold.h (gimple_build): Make the function forms take
4529 combined_fn rather than built_in_function.
4530 (gimple_simplify): Likewise.
4531 * gimple-match-head.c (gimple_simplify): Likewise.
4532 * gimple-fold.c (gimple_build): Likewise.
4533 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4534 rather than gimple_build_call_internal.
4535 (get_initial_defs_for_reduction): Likewise.
4536 (vect_create_epilog_for_reduction): Likewise.
4537 (vectorizable_live_operation): Likewise.
4539 2018-05-17 Martin Liska <mliska@suse.cz>
4541 * gimple-ssa-sprintf.c (format_directive): Do not use
4542 space in between 'G_' and '('.
4544 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4547 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4548 even if the mask is not all ones.
4551 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4553 (ix86_gimple_fold_builtin): Likewise.
4556 * config/i386/i386.c: Include tree-vector-builder.h.
4557 (ix86_vector_shift_count): New function.
4558 (ix86_fold_builtin): Fold shift builtins by scalar count.
4559 (ix86_gimple_fold_builtin): Likewise.
4561 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4562 _mm512_setzero): New intrinsics.
4564 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4565 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4567 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4568 code generation for cases where splatting a value is not useful.
4569 * simplify-rtx.c (simplify_ternary_operation): Simplify
4570 vec_merge across a vec_duplicate and a paradoxical subreg forming
4571 a vector mode to a vec_concat.
4573 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4575 * config.gcc: Support "goldmont-plus".
4576 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4578 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4579 PROCESSOR_GOLDMONT_PLUS.
4580 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4581 (processor_target_table): Add "goldmont-plus".
4582 (PTA_GOLDMONT_PLUS): Define.
4583 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4584 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4585 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4586 (fold_builtin_cpu): Add "goldmont-plus".
4587 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4588 (ix86_option_override_internal): Add "goldmont-plus".
4589 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4590 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4591 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4592 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4594 2018-05-17 Richard Biener <rguenther@suse.de>
4596 PR tree-optimization/85757
4597 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4598 remove defs that only feed that PHI from further processing.
4600 2018-05-16 Jim Wilson <jimw@sifive.com>
4602 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4604 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4606 2018-05-16 Mark Wielaard <mark@klomp.org>
4608 * dwarf2out.c (count_index_strings): New function.
4609 (output_indirect_strings): Call count_index_strings and generate
4610 header for dwarf_version >= 5.
4612 2018-05-16 Mark Wielaard <mark@klomp.org>
4614 * dwarf2out.c (dwarf_FORM): New function.
4615 (set_indirect_string): Use dwarf_FORM.
4616 (reset_indirect_string): Likewise.
4617 (size_of_die): Likewise.
4618 (value_format): Likewise.
4619 (output_die): Likewise.
4620 (add_skeleton_AT_string): Likewise.
4621 (output_macinfo_op): Likewise.
4622 (index_string): Likewise.
4623 (output_index_string_offset): Likewise.
4624 (output_index_string): Likewise.
4625 (count_index_strings): Likewise.
4627 2018-05-16 Carl Love <cel@us.ibm.com>
4629 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4630 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4632 2018-05-16 Martin Jambor <mjambor@suse.cz>
4634 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4635 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4637 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4639 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4640 (fnma<mode>4): Likewise.
4641 (fms<mode>4): Likewise.
4642 (fnms<mode>4): Likewise.
4643 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4644 (aarch64_fnma<mode>4): Likewise.
4645 (aarch64_fms<mode>4): Likewise.
4646 (aarch64_fnms<mode>4): Likewise.
4647 (aarch64_fnmadd<mode>4): Likewise.
4649 2018-05-16 Jason Merrill <jason@redhat.com>
4651 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4653 2018-05-16 Richard Biener <rguenther@suse.de>
4655 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4656 (dump_stmt_cost): Declare.
4657 (add_stmt_cost): Dump cost we add.
4658 (add_stmt_costs): New function.
4659 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4661 (vect_analyze_stmt): Adjust prototype.
4662 (vectorizable_condition): Likewise.
4663 (vectorizable_live_operation): Likewise.
4664 (vectorizable_reduction): Likewise.
4665 (vectorizable_induction): Likewise.
4666 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4667 cost vector to pass to vectorizable_ and record afterwards.
4668 (vect_model_reduction_cost): Take cost vector argument and adjust.
4669 (vect_model_induction_cost): Likewise.
4670 (vectorizable_reduction): Likewise.
4671 (vectorizable_induction): Likewise.
4672 (vectorizable_live_operation): Likewise.
4673 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4674 SLP_TREE_NUMBER_OF_VEC_STMTS.
4675 (vect_analyze_slp_cost_1): Remove.
4676 (vect_analyze_slp_cost): Likewise.
4677 (vect_slp_analyze_node_operations): Take visited args and
4678 a target cost vector. Avoid processing already visited stmt sets.
4679 (vect_slp_analyze_operations): Use a local cost vector to gather
4680 costs and register those of non-discarded instances.
4681 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4682 (vect_schedule_slp_instance): Remove copying of
4683 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4685 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4686 adding cost. Record cost entry location.
4687 (vect_prologue_cost_for_slp_op): Function to compute cost of
4688 a constant or invariant generated for SLP vect in the prologue,
4689 split out from vect_analyze_slp_cost_1.
4690 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4691 (vect_model_promotion_demotion_cost): Likewise.
4692 (vect_model_store_cost): Likewise, make static.
4693 (vect_model_load_cost): Likewise.
4694 (vectorizable_bswap): Add cost vector arg and adjust.
4695 (vectorizable_call): Likewise.
4696 (vectorizable_simd_clone_call): Likewise.
4697 (vectorizable_conversion): Likewise.
4698 (vectorizable_assignment): Likewise.
4699 (vectorizable_shift): Likewise.
4700 (vectorizable_operation): Likewise.
4701 (vectorizable_store): Likewise.
4702 (vectorizable_load): Likewise.
4703 (vectorizable_condition): Likewise.
4704 (vectorizable_comparison): Likewise.
4705 (can_vectorize_live_stmts): Likewise.
4706 (vect_analyze_stmt): Likewise.
4707 (vect_transform_stmt): Adjust calls to vectorizable_*.
4708 * tree-vectorizer.c: Include gimple-pretty-print.h.
4709 (dump_stmt_cost): New function.
4711 2018-05-16 Richard Biener <rguenther@suse.de>
4713 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4714 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4715 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4716 (check_name): New callback.
4717 (dse_classify_store): Track cycles via a visited bitmap of PHI
4718 defs and simplify handling of in-loop and across loop dead stores
4719 and properly fail for loop-variant refs. Handle byte-tracking with
4720 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4723 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4725 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4726 (vect_get_mask_type_for_stmt): Likewise.
4727 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4729 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4730 to determine the statement's vector type and the vector type that
4731 should be used for calculating nunits. Deal with cases in which
4732 the type has to be deferred.
4733 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4734 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4735 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4736 (vect_determine_vf_for_stmt): New functions, split out from...
4737 (vect_determine_vectorization_factor): ...here.
4738 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4739 (vect_get_mask_type_for_stmt): New functions, split out from
4740 vect_determine_vectorization_factor.
4742 2018-05-16 Richard Biener <rguenther@suse.de>
4744 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4745 verify the [VEC_]COND_EXPR embedded comparison.
4747 2018-05-15 Martin Sebor <msebor@redhat.com>
4749 PR tree-optimization/85753
4750 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4751 RECORD_TYPE in addition to ARRAY_TYPE.
4753 2018-05-15 Martin Sebor <msebor@redhat.com>
4756 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4758 2018-05-15 Richard Biener <rguenther@suse.de>
4760 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4761 add by_clobber_p one. Change algorithm to collect all defs
4762 representing uses we need to walk and try reducing them to
4763 a single one before failing.
4764 (dse_dom_walker::dse_optimize_stmt): Adjust.
4766 2018-05-13 Mark Wielaard <mark@klomp.org>
4768 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4769 (size_of_loc_descr): Likewise.
4770 (output_loc_operands): Likewise.
4771 (output_loc_operands_raw): Likewise.
4772 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4773 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4774 (hash_loc_operands): Likewise.
4775 (compare_loc_operands): Likewise.
4777 2018-05-14 Mark Wielaard <mark@klomp.org>
4779 * dwarf2out.c (count_index_addrs): New function.
4780 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4782 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4784 PR tree-optimization/83648
4785 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4786 return value as malloc candidate.
4788 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4791 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4792 param as true in call to suggest_attribute.
4794 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4796 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4797 -mreadonly-in-sdata.
4799 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4801 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4803 (aarch64_crypto_aesd_fused): Likewise.
4805 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4807 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4808 (movsi_aarch64): Likewise.
4809 (load_pairsi): Likewise.
4810 (load_pairdi): Likewise.
4811 (store_pairsi): Likewise.
4812 (store_pairdi): Likewise.
4813 (load_pairsf): Likewise.
4814 (load_pairdf): Likewise.
4815 (store_pairsf): Likewise.
4816 (store_pairdf): Likewise.
4817 (zero_extend): Likewise.
4818 (trunc): Swap alternatives.
4819 (fcvt_target): Add '?' to prefer w over r.
4821 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4824 * config/i386/i386.md: Disallow non-commutative arithmetics in
4825 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4826 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4827 in the peephole2 before it.
4829 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4831 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4832 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4833 (ix86_handle_option): Handle -mcldemote.
4834 * config.gcc: New header.
4835 * config/i386/cldemoteintrin.h: New file.
4836 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4837 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4839 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4840 OPTION_MASK_ISA_CLDEMOTE.
4841 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4842 (ix86_valid_target_attribute_inner_p): Ditto.
4843 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4844 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4845 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4846 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4847 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4849 * config/i386/i386.opt: Add -mcldemote.
4850 * config/i386/x86intrin.h: New header.
4851 * doc/invoke.texi: Add -mcldemote.
4853 2018-05-14 Richard Biener <rguenther@suse.de>
4855 * doc/match-and-simplify.texi: Adjust :s documentation.
4857 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4859 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4860 intended memcpy size.
4861 (REORDER_45): Likewise.
4863 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4865 * sort.cc: New file.
4866 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4867 * vec.c (qsort_chk): Use gcc_qsort.
4868 * Makefile.in (OBJS-libcommon): Add sort.o.
4869 (build/sort.o): New target. Use it...
4870 (BUILD_RTL): ... here, and...
4871 (build/gencfn-macros): ... here, and...
4872 (build/genmatch): ... here.
4874 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4875 Chung-Ju Wu <jasonwucj@gmail.com>
4877 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4878 * config/nds32/nds32-graywolf.md: New file.
4879 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4880 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4882 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4883 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4884 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4885 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4886 * config/nds32/pipelines.md: Include n15 settings.
4888 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4889 Chung-Ju Wu <jasonwucj@gmail.com>
4891 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4892 * config/nds32/nds32-n13.md: New file.
4893 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4894 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4896 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4897 * config/nds32/nds32.md (pipeline_model): Add n13.
4898 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4899 * config/nds32/pipelines.md: Include n13 settings.
4901 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4902 Chung-Ju Wu <jasonwucj@gmail.com>
4904 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4905 * config/nds32/nds32-n10.md: New file.
4906 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4907 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4909 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4910 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4911 * config/nds32/nds32.md (pipeline_model): Add n10.
4912 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4913 * config/nds32/pipelines.md: Include n10 settings.
4915 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4916 Kito Cheng <kito.cheng@gmail.com>
4917 Chung-Ju Wu <jasonwucj@gmail.com>
4919 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4920 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4921 Add enum values for DSP extension instructions.
4922 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4924 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4925 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4927 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4928 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4929 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4930 * config/nds32/nds32-intrinsic.md: Likewise.
4931 * config/nds32/nds32_intrinsic.h: Likewise.
4932 * config/nds32/nds32-md-auxiliary.c: Likewise.
4933 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4934 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4935 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4936 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4937 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4938 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4939 TYPE_DMAC in switch statement.
4940 * config/nds32/nds32.c: New checking and implementation for DSP
4941 extension instructions.
4942 * config/nds32/nds32.h: Likewise.
4943 * config/nds32/nds32.md: Likewise.
4944 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4945 * config/nds32/predicates.md: Implement new predicates for DSP
4948 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4950 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4951 Reformat alternatives and attributes so it is easier to identify
4952 which constraints/attributes go with which instruction.
4953 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4954 (mov<mode>_softfloat32, FMOVE64): Likewise.
4955 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4956 (mov<mode>_softfloat64, FMOVE64): Likewise.
4958 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4960 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4962 (Basic PowerPC Built-in Functions): The new name of the
4963 subsection previously known as "PowerPC Built-in Functions".
4964 (Basic PowerPC Built-in Functions Available on all Configurations):
4966 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4967 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4968 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4969 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4971 2018-05-11 Martin Jambor <mjambor@suse.cz>
4974 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4977 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4980 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4982 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4984 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4985 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4986 (ix86_handle_option): Handle -mwaitpkg.
4987 * config.gcc: New header.
4988 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4989 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4990 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4992 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4993 OPTION_MASK_ISA_WAITPKG.
4994 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4995 (ix86_option_override_internal): Add PTA_WAITPKG.
4996 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4997 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4998 IX86_BUILTIN_TPAUSE.
4999 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
5000 __builtin_ia32_umwait and __builtin_ia32_tpause.
5001 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
5002 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
5003 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
5004 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
5005 UNSPECV_TPAUSE): New.
5006 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
5007 * config/i386/i386.opt: Add -mwaitpkg.
5008 * config/i386/waitpkgintrin.h: New file.
5009 * config/i386/x86intrin.h: New header.
5010 * doc/invoke.texi: Add -mwaitpkg.
5012 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
5015 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
5017 (cortex-m0): Use armv6s-m isa.
5018 (cortex-m0plus): Likewise.
5019 (cortex-m1): Likewise.
5020 (cortex-m0.small-multiply): Likewise.
5021 (cortex-m0plus.small-multiply): Likewise.
5022 (cortex-m1.small-multiply): Likewise.
5024 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
5025 Jakub Jelinek <jakub@redhat.com>
5027 PR tree-optimization/85692
5028 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
5029 source permute as well.
5031 2018-05-11 Martin Liska <mliska@suse.cz>
5034 * doc/extend.texi: Document LLVM style format for no_sanitize
5037 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5039 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
5040 mode_supports_vsx_dform_quad to mode_supports_dq_form.
5041 (mode_supports_vsx_dform_quad): Likewise.
5042 (mode_supports_vmx_dform): Move these functions to be next to the
5043 other mode_supports functions.
5044 (mode_supports_dq_form): Likewise.
5045 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
5046 mode_supports_dq_form.
5047 (reg_offset_addressing_ok_p): Likewise.
5048 (offsettable_ok_by_alignment): Likewise.
5049 (rs6000_legitimate_offset_address_p): Likewise.
5050 (legitimate_lo_sum_address_p): Likewise.
5051 (rs6000_legitimize_address): Likewise.
5052 (rs6000_legitimize_reload_address): Likewise.
5053 (rs6000_secondary_reload_inner): Likewise.
5054 (rs6000_preferred_reload_class): Likewise.
5055 (rs6000_output_move_128bit): Likewise.
5057 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5059 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
5060 Generate SImode target register for null target.
5061 <case IX86_BUILTIN_XGETBV>: Ditto.
5062 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
5063 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
5065 2018-05-10 Carl Love <cel@us.ibm.com>
5067 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
5068 dcbtt and dcbtstt if operands[2] is 0.
5070 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
5073 * config/i386/sse.md (usadv64qi): New expander.
5075 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
5077 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
5078 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
5079 -maltivec=be support.
5080 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
5081 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
5082 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
5083 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
5084 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
5085 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
5086 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
5087 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
5088 altivec_vsumsws): Adjust.
5089 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
5090 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
5091 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
5093 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
5094 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
5095 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
5096 (altivec_lve<VI_char>x): Delete expand.
5097 (*altivec_lve<VI_char>x_internal): Rename to...
5098 (altivec_lve<VI_char>x): ... this.
5099 (altivec_lvxl_<mode>): Delete expand.
5100 (*altivec_lvxl_<mode>_internal): Rename to ...
5101 (altivec_lvxl_<mode>): ... this.
5102 (altivec_stvxl_<mode>): Delete expand.
5103 (*altivec_stvxl_<mode>_internal): Rename to ...
5104 (altivec_stvxl_<mode>): ... this.
5105 (altivec_stve<VI_char>x): Delete expand.
5106 (*altivec_stve<VI_char>x_internal): Rename to ...
5107 (altivec_stve<VI_char>x): ... this.
5108 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
5109 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
5110 reduc_plus_scal_<mode>): Adjust.
5111 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
5113 (rs6000_cpu_cpp_builtins): Adjust.
5114 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
5115 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
5116 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
5117 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
5118 -maltivec=be support.
5119 (rs6000_split_vec_extract_var): Adjust.
5120 (rs6000_split_v4si_init): Adjust.
5121 (swap_selector_for_mode): Delete.
5122 (altivec_expand_lvx_be, altivec_expand_stvx_be,
5123 altivec_expand_stvex_be): Delete.
5124 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
5125 -maltivec=be support.
5126 (rs6000_gimple_fold_builtin): Ditto.
5127 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5129 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5130 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5131 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5132 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5133 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5134 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5135 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5136 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5137 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5138 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5139 anonymous split): Adjust.
5140 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5141 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5143 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5145 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5146 when --with-gxx-include-dir is also specified.
5147 * configure: Regenerate.
5149 2018-05-09 Jim Wilson <jimw@sifive.com>
5152 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5153 * config/riscv/t-withmultilib: New.
5154 * config/riscv/withmultilib.h: New.
5155 * doc/install.texi: Document RISC-V --with-multilib-list support.
5157 2018-05-09 Richard Biener <rguenther@suse.de>
5159 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5161 (vect_bb_vectorization_profitable_p): Adjust. Compute
5162 actual scalar cost using the cost vector and the add_stmt_cost
5165 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5167 PR rtl-optimization/85645
5168 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5169 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5171 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5173 PR rtl-optimization/85645
5174 * shrink-wrap.c (spread_components): Return a boolean saying if
5175 anything was changed.
5176 (try_shrink_wrapping_separate): Iterate spread_components until
5177 nothing changes anymore.
5179 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5181 PR rtl-optimization/85645
5182 * regrename.c (build_def_use): Also kill the chains that include the
5183 destination of a REG_CFA_REGISTER note.
5185 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5187 PR rtl-optimization/85645
5188 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5189 insn that has a REG_CFA_REGISTER note.
5191 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5193 * cfgexpand.c (expand_clobber): New function.
5194 (expand_gimple_stmt_1): Use it.
5195 * tree-vect-stmts.c (vect_clobber_variable): New function,
5197 (vectorizable_simd_clone_call): ...here.
5198 (vectorizable_store): Emit a clobber either side of an
5199 IFN_STORE_LANES sequence.
5200 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5202 2018-05-09 Tom de Vries <tom@codesourcery.com>
5205 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5206 (define_insn "trap_if_false"): Add exit after trap.
5208 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5210 PR rtl-optimization/85638
5211 * bb-reorder.c: Include common/common-target.h.
5212 (create_forwarder_block): New function extracted from...
5213 (fix_up_crossing_landing_pad): ...here. Rename into...
5214 (dw2_fix_up_crossing_landing_pad): ...this.
5215 (sjlj_fix_up_crossing_landing_pad): New function.
5216 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5217 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5218 from both partitions and exit the loop after one iteration.
5220 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5223 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5225 (Basic PowerPC Built-in Functions): The new name of the
5226 subsection previously known as "PowerPC Built-in Functions".
5227 (Basic PowerPC Built-in Functions Available on all Configurations):
5229 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5231 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5233 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5235 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5238 2018-05-08 Jim Wilson <jimw@sifive.com>
5240 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5241 (LD_EMUL_SUFFIX): New.
5242 (LINK_SPEC): Use it.
5244 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5246 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5248 (Basic PowerPC Built-in Functions): The new name of the
5249 subsection previously known as "PowerPC Built-in Functions".
5250 (Basic PowerPC Built-in Functions Available on all Configurations):
5252 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5254 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5256 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5258 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5261 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5264 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5265 after cmpelim optimization.
5267 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5269 * config.gcc: Support "goldmont".
5270 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5271 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5273 * config/i386/i386.c (m_GOLDMONT): Define.
5274 (processor_target_table): Add "goldmont".
5275 (PTA_GOLDMONT): Define.
5276 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5277 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5278 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5279 (fold_builtin_cpu): Add "goldmont".
5280 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5281 (ix86_option_override_internal): Add "goldmont".
5282 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5283 (processor_type): Add PROCESSOR_GOLDMONT.
5284 * config/i386/i386.md: Add CPU "glm".
5285 * config/i386/glm.md: New file.
5286 * config/i386/x86-tune.def: Add m_GOLDMONT.
5287 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5289 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5292 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5294 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5295 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5296 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5299 * config/i386/i386.c (ix86_fold_builtin): Handle
5300 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5303 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5304 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5306 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5309 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5310 (check_arch): Likewise.
5311 (check_fpu): Return the result rather than printing it.
5312 (end arch): Fix operator precedence.
5313 (end cpu): Likewise.
5314 (END): Print the result from check_fpu.
5316 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5317 Alan Hayward <alan.hayward@arm.com>
5318 David Sherwood <david.sherwood@arm.com>
5320 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5321 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5322 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5323 (*fcmuo<mode>_and): New patterns.
5325 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5327 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5328 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5329 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5330 (cmp_op, sve_imm_con): New code attributes.
5331 (SVE_COND_INT_CMP, imm_con): Delete.
5332 (cmp_op): Remove above unspecs from int attribute.
5333 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5335 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5336 comparison-specific unspecs.
5337 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5338 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5339 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5340 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5341 (*vec_fcm<cmp_op><mode>): Rename to...
5342 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5343 (*vec_fcmuo<mode>): Rename to...
5344 (*fcmuo<mode>): ...this and adjust likewise.
5345 (*pred_fcm<cmp_op><mode>): New pattern.
5346 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5347 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5349 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5351 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5352 (aarch64_emit_sve_predicated_cond): New function.
5353 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5354 (aarch64_emit_unspec_cond_or): Replace with...
5355 (aarch64_emit_sve_or_conds): ...this new function. Use
5356 aarch64_emit_sve_ptrue_op for the individual comparisons and
5357 aarch64_emit_binop to OR them together.
5358 (aarch64_emit_inverted_unspec_cond): Replace with...
5359 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5360 aarch64_emit_sve_ptrue_op for the comparison and
5361 aarch64_emit_unop to invert the result.
5362 (aarch64_expand_sve_vec_cmp_float): Update after the above
5363 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5365 2018-05-07 Nathan Sidwell <nathan@acm.org>
5367 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5368 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5369 (Backwards Compatibility): Likewise.
5371 2018-05-07 Luis Machado <luis.machado@linaro.org>
5375 2018-05-07 Luis Machado <luis.machado@linaro.org>
5377 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5378 <prefetch_dynamic_strides>: New const bool field.
5379 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5380 prefetch_dynamic_strides.
5381 (exynosm1_prefetch_tune): Likewise.
5382 (thunderxt88_prefetch_tune): Likewise.
5383 (thunderx_prefetch_tune): Likewise.
5384 (thunderx2t99_prefetch_tune): Likewise.
5385 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5387 (aarch64_override_options_internal): Update to set
5388 PARAM_PREFETCH_DYNAMIC_STRIDES.
5389 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5390 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5391 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5392 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5393 prefetch-dynamic-strides setting.
5395 2018-05-07 Luis Machado <luis.machado@linaro.org>
5397 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5398 <minimum_stride>: New const int field.
5399 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5400 minimum_stride field.
5401 (exynosm1_prefetch_tune): Likewise.
5402 (thunderxt88_prefetch_tune): Likewise.
5403 (thunderx_prefetch_tune): Likewise.
5404 (thunderx2t99_prefetch_tune): Likewise.
5405 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5406 (aarch64_override_options_internal): Update to set
5407 PARAM_PREFETCH_MINIMUM_STRIDE.
5408 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5409 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5410 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5411 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5412 stride is constant and is below the minimum stride threshold.
5414 2018-05-07 Luis Machado <luis.machado@linaro.org>
5416 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5419 2018-05-07 Luis Machado <luis.machado@linaro.org>
5421 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5422 <prefetch_dynamic_strides>: New const bool field.
5423 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5424 prefetch_dynamic_strides.
5425 (exynosm1_prefetch_tune): Likewise.
5426 (thunderxt88_prefetch_tune): Likewise.
5427 (thunderx_prefetch_tune): Likewise.
5428 (thunderx2t99_prefetch_tune): Likewise.
5429 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5431 (aarch64_override_options_internal): Update to set
5432 PARAM_PREFETCH_DYNAMIC_STRIDES.
5433 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5434 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5435 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5436 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5437 prefetch-dynamic-strides setting.
5439 2018-05-07 Luis Machado <luis.machado@linaro.org>
5441 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5442 <minimum_stride>: New const int field.
5443 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5444 minimum_stride field.
5445 (exynosm1_prefetch_tune): Likewise.
5446 (thunderxt88_prefetch_tune): Likewise.
5447 (thunderx_prefetch_tune): Likewise.
5448 (thunderx2t99_prefetch_tune): Likewise.
5449 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5450 (aarch64_override_options_internal): Update to set
5451 PARAM_PREFETCH_MINIMUM_STRIDE.
5452 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5453 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5454 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5455 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5456 stride is constant and is below the minimum stride threshold.
5458 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5461 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5462 the type is addressable. Don't force op into register if it has
5465 2018-05-05 Roland McGrath <mcgrathr@google.com>
5468 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5469 any section for which we don't know a specific type it should have,
5470 regardless of name. Previously this was done only for the exact
5471 names ".init_array", ".fini_array", and ".preinit_array".
5472 (default_elf_asm_named_section): Add comment about
5473 relationship with default_section_type_flags and SECTION_NOTYPE.
5474 (get_section): Don't consider it a type conflict if one side has
5475 SECTION_NOTYPE and the other doesn't, as long as neither has the
5476 SECTION_BSS et al used in the default_section_type_flags logic.
5478 2018-05-05 Tom de Vries <tom@codesourcery.com>
5481 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5482 (workaround_barsyncs): New function.
5483 (nvptx_reorg): Use workaround_barsyncs.
5484 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5485 (define_expand "nvptx_membar_cta"): New define_expand.
5486 (define_insn "*nvptx_membar_cta"): New insn.
5488 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5490 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5491 To improve optimization opportunities.
5492 * builtin-types.def: The new needed builtin types for the above.
5494 2018-05-04 Richard Biener <rguenther@suse.de>
5496 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5497 * gimple-ssa-store-merging.c
5498 (imm_store_chain_info::output_merged_store): Remove redundant create,
5499 release split_store vector contents on failure.
5500 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5501 scalar stmt vector on cache hit.
5503 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5505 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5507 * config.gcc (powerpc-xilinx-eabi*): Remove.
5508 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5510 (fusion_addis_mem_combo_load): Ditto.
5511 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5513 (rs6000_cpu_cpp_builtins): Ditto.
5514 * config/rs6000/rs6000-linux.c
5515 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5516 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5517 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5519 (rs6000_setup_reg_addr_masks): Ditto.
5520 (rs6000_init_hard_regno_mode_ok): Ditto.
5521 (rs6000_option_override_internal): Ditto.
5522 (legitimate_lo_sum_address_p): Ditto.
5523 (rs6000_legitimize_address): Ditto.
5524 (rs6000_legitimize_reload_address): Ditto.
5525 (rs6000_legitimate_address_p): Ditto.
5526 (abi_v4_pass_in_fpr): Ditto.
5527 (setup_incoming_varargs): Ditto.
5528 (rs6000_gimplify_va_arg): Ditto.
5529 (rs6000_split_multireg_move): Ditto.
5530 (rs6000_savres_strategy): Ditto.
5531 (rs6000_emit_prologue_components): Ditto.
5532 (rs6000_emit_epilogue_components): Ditto.
5533 (rs6000_emit_prologue): Ditto.
5534 (rs6000_emit_epilogue): Ditto.
5535 (rs6000_elf_file_end): Ditto.
5536 (rs6000_function_value): Ditto.
5537 (rs6000_libcall_value): Ditto.
5538 * config/rs6000/rs6000.h: Ditto.
5539 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5540 (TARGET_MINMAX): ... this. New.
5541 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5542 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5543 (*movsi_internal1_single): Delete.
5544 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5545 mfpu=, mxilinx-fpu): Delete.
5546 * config/rs6000/singlefp.h: Delete.
5547 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5548 * config/rs6000/t-rs6000: Ditto.
5549 * config/rs6000/t-xilinx: Delete.
5550 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5551 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5552 (VStype_simple): Delete.
5553 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5554 * config/rs6000/xfpu.h: Delete.
5555 * config/rs6000/xfpu.md: Delete.
5556 * config/rs6000/xilinx.h: Delete.
5557 * config/rs6000/xilinx.opt: Delete.
5558 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5559 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5561 2018-05-04 Tom de Vries <tom@codesourcery.com>
5564 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5567 2018-05-04 Richard Biener <rguenther@suse.de>
5570 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5571 (expand_complex_div_wide): Likewise.
5572 (expand_complex_operations_1): Likewise.
5573 (expand_complex_libcall): Preserve EH info of the original stmt.
5574 (tree_lower_complex): Handle removed blocks.
5575 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5576 on complex multiplication and division libcall builtins.
5578 2018-05-04 Richard Biener <rguenther@suse.de>
5581 * fold-const.c (negate_expr_p): Restrict negation of operand
5582 zero of a division to when we know that can happen without
5584 (fold_negate_expr_1): Likewise.
5586 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5589 * real.h (real_nextafter): Declare.
5590 * real.c (real_nextafter): New function.
5591 * fold-const-call.c (fold_const_nextafter): New function.
5592 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5593 CASE_CFN_NEXTTOWARD.
5594 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5595 even when arg1_mode is different from arg0_mode.
5597 2018-05-03 Nathan Sidwell <nathan@acm.org>
5599 * doc/extend.texi (Deprecated Features): Remove
5601 (Backwards Compatibility): Likewise.
5602 * doc/invoke.texi (C++ Language Options): Likewise.
5603 (C++ Dialect Options): Likewise.
5605 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5608 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5609 _mm512_mask_mullox_epi64): New intrinsics.
5611 2018-05-03 Tom de Vries <tom@codesourcery.com>
5614 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5615 dump files): Add offload-tree.
5617 2018-05-03 Richard Biener <rguenther@suse.de>
5619 PR tree-optimization/85615
5620 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5621 to loops not nested in BBs loop father to avoid creating multi-entry
5624 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5626 PR tree-optimization/70291
5627 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5628 arguments. Change return type to tree. Emit libcall as a new
5629 statement rather than replacing existing one when inplace_p is true.
5630 (expand_complex_multiplication_components): New function.
5631 (expand_complex_multiplication): Expand floating-point complex
5632 multiplication using the above.
5633 (expand_complex_division): Rename inner_type parameter to type.
5634 Update expand_complex_libcall call-site.
5635 (expand_complex_operations_1): Update expand_complex_multiplication
5636 and expand_complex_division call-sites.
5638 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5641 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5642 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5643 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5644 the highest significant bit of the shift count mask is clear. In
5645 check whether and[sq]i3 is needed verify that all significant bits
5646 of the shift count other than the highest are set.
5648 2018-05-02 Tom de Vries <tom@codesourcery.com>
5651 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5652 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5653 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5654 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5655 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5656 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5657 __builtin_goacc_parlevel_size.
5659 2018-05-02 Richard Biener <rguenther@suse.de>
5661 PR tree-optimization/85597
5662 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5663 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5666 2018-05-02 Tom de Vries <tom@codesourcery.com>
5669 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5670 dump files): Add ltrans-tree.
5672 2018-05-02 Tom de Vries <tom@codesourcery.com>
5675 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5676 dump files): Add wpa-ipa.
5678 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5680 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5681 powerpc*-*-linux*paired* target.
5682 * config/rs6000/750cl.h: Delete.
5683 * config/rs6000/paired.h: Delete.
5684 * config/rs6000/paired.md: Delete.
5685 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5687 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5688 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5689 comment. Remove paired float support.
5690 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5691 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5693 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5694 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5696 * config/rs6000/rs6000.c: Remove paired float support.
5697 (paired_expand_vector_init, paired_expand_vector_move,
5698 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5699 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5700 paired_expand_builtin, paired_expand_predicate_builtin,
5701 paired_init_builtins): Delete.
5702 * config/rs6000/rs6000.h: Remove paired float support.
5703 * config/rs6000/rs6000.md: Remove paired float support.
5704 (move_from_CR_ov_bit): Delete.
5705 * config/rs6000/rs6000.opt (mpaired): Delete.
5706 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5707 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5709 2018-05-02 Richard Biener <rguenther@suse.de>
5712 * gimplify.c (gimplify_save_expr): When in SSA form allow
5713 SAVE_EXPRs to compute to SSA vars.
5715 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5718 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5719 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5720 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5721 clobber operands[2], instead use a new pseudo. Formatting fixes.
5723 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5725 PR tree-optimization/85586
5726 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5727 exit early for statements in the same group if the accesses are
5730 2018-05-02 Tom de Vries <tom@codesourcery.com>
5733 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5736 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5738 PR tree-optimization/85143
5739 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5741 2018-05-01 Tom de Vries <tom@codesourcery.com>
5744 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5745 not found" error message.
5747 2018-05-01 Tom de Vries <tom@codesourcery.com>
5750 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5751 * vec.c (test_ordered_remove_if): New function.
5752 (vec_c_tests): Call test_ordered_remove_if.
5753 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5754 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5755 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5756 VEC_ORDERED_REMOVE_IF.
5758 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5760 PR tree-optimization/82665
5761 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5762 pointer subtraction where arguments come from a memchr call.
5764 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5766 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5767 --push-state --as-needed and --pop-state instead of --as-needed and
5768 --no-as-needed if ld supports it.
5769 * configure: Regenerated.
5772 * doc/install.texi2html: Replace _002d with - and _002a with * in
5773 generated html files using sed.
5775 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5778 * gcc-rich-location.c (blank_line_before_p): New function.
5779 (use_new_line): New function.
5780 (gcc_rich_location::add_fixit_insert_formatted): New function.
5781 * gcc-rich-location.h
5782 (gcc_rich_location::add_fixit_insert_formatted): New function.
5784 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5786 * selftest.c (assert_streq): Rename "expected" and "actual" to
5787 "val1" and "val2". Extend NULL-handling to cover both inputs
5788 symmetrically, while still requiring both to be non-NULL for a pass.
5789 * selftest.h (assert_streq): Rename "expected" and "actual" to
5791 (ASSERT_EQ): Likewise.
5792 (ASSERT_EQ_AT): Likewise.
5793 (ASSERT_KNOWN_EQ): Likewise.
5794 (ASSERT_KNOWN_EQ_AT): Likewise.
5795 (ASSERT_NE): Likewise.
5796 (ASSERT_MAYBE_NE): Likewise.
5797 (ASSERT_MAYBE_NE_AT): Likewise.
5798 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5799 the assertion to pass.
5800 (ASSERT_STREQ_AT): Likewise.
5802 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5804 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5805 interaction with -pie.
5807 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5809 * selftest.h: Fix alphabetization of per-source-file selftest
5812 2018-04-30 Jason Merrill <jason@redhat.com>
5814 PR c++/61982 - dead stores to destroyed objects.
5815 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5818 2018-04-30 Jason Merrill <jason@redhat.com>
5820 * tree.c (build_clobber): New.
5821 * tree.h: Declare it.
5822 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5824 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5826 * diagnostic-show-locus.c (layout::layout): Update for
5827 location_get_source_line returning a char_span.
5828 (struct char_span): Move to input.h.
5829 (struct correction): Update for fields in char_span becoming
5831 (struct source_line): Update for location_get_source_line
5832 returning a char_span.
5833 (layout::print_line): Likewise.
5834 * edit-context.c (edited_file::print_content): Likewise.
5835 (edited_file::print_diff_hunk): Likewise.
5836 (edited_file::print_run_of_changed_lines): Likewise.
5837 (edited_file::get_num_lines): Likewise.
5838 (edited_line::edited_line): Likewise.
5839 * final.c (asm_show_source): Likewise.
5840 * input.c (location_get_source_line): Convert return type
5841 from const char * to char_span, losing the final "line_len"
5843 (dump_location_info): Update for the above.
5844 (get_substring_ranges_for_loc): Likewise. Use a char_span
5845 when handling the literal within the line.
5846 (test_reading_source_line): Update for location_get_source_line
5847 returning a char_span.
5848 * input.h (class char_span): Move here from
5849 diagnostic-show-locus.c, converting from a struct to a class.
5850 Make data members private.
5851 (char_span::operator bool): New.
5852 (char_span::length): New.
5853 (char_span::get_buffer): New.
5854 (char_span::operator[]): New.
5855 (char_span::subspan): Make const.
5856 (char_span::xstrdup): New.
5857 (location_get_source_line): Convert return type from const char *
5858 to char_span, losing the final "line_size" param.
5860 2018-04-30 Jan Hubicka <jh@suse.cz>
5862 * lto-wrapper.c (ltrans_priorities): New static var.
5863 (cmp_priority): New.
5864 (run_gcc): Read priorities and if doing parallel build order
5865 the Makefile by them.
5867 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5869 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5871 2018-04-30 Richard Biener <rguenther@suse.de>
5873 * tree-cfg.c (verify_address): Remove base argument, add
5874 flag whether to check TREE_ADDRESSABLE and do that.
5875 (verify_expr): Remove.
5876 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5877 (verify_gimple_assign_single): Likewise.
5878 (verify_gimple_switch): Likewise.
5879 (verify_expr_location_1): Dereference tp once. Add (disabled)
5880 piece from verify_expr.
5881 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5883 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5885 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5887 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5889 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5890 (small_data_pattern): Likewise.
5891 (arc_rewrite_small_data): Likewise.
5892 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5893 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5894 (get_symbol_alignment): New function.
5895 (legitimate_small_data_address_p): Likewise.
5896 (legitimate_scaled_address): Update, call
5897 legitimate_small_data_address_p.
5898 (output_sdata): New static variable.
5899 (arc_print_operand): Update how we handle small data operands.
5900 (arc_print_operand_address): Likewise.
5901 (arc_legitimate_address_p): Update, use
5902 legitimate_small_data_address_p.
5903 (arc_rewrite_small_data_p): Remove.
5904 (arc_rewrite_small_data_1): Likewise.
5905 (arc_rewrite_small_data): Likewise.
5906 (small_data_pattern): Likewise.
5907 (compact_sda_memory_operand): Update to use
5908 legitimate_small_data_address_p and get_symbol_alignment.
5909 (prepare_move_operands): Don't rewite sdata pattern.
5910 (prepare_extend_operands): Remove.
5911 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5913 (zero_extendqisi2): Likewise.
5914 (zero_extendhisi2): Likewise.
5915 (extendqihi2): Likewise.
5916 (extendqisi2): Likewise.
5917 (extendhisi2): Likewise.
5921 * config/arc/constraints.md (Usd): Change it to memory constraint.
5923 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5925 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5926 as source of std instructions.
5927 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5928 allow 6-bit constants as source for store instructions.
5929 (movdi_insn): Update instruction pattern to allow 6-bit constants
5930 as source for store instructions.
5932 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5934 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5936 2018-04-30 Nathan Sidwell <nathan@acm.org>
5937 Sandra Loosemore <sandra@codesourcery.com>
5939 * dumpfile.c (dump_open): Allow '-' for stdout.
5940 * doc/invoke.texi (Developer Options): Document dump filename
5941 determination early. Document stdin/stdout selection.
5943 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5945 Microblaze Target: PIC data text relative
5947 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5948 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5950 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5951 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5952 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5953 New addressing mode for data-text relative position indepenedent code.
5954 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5955 'ADDRESS_SYMBOLIC_TXT_REL'.
5956 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5957 (microblaze_legitimate_pic_operand): Exclude function calls from
5958 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5959 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5961 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5962 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5963 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5964 for 'address + offset'.
5965 (microblaze_expand_prologue): Add new function prologue call for
5967 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5968 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5969 table in case of TARGET_PIC_DATA_TEXT_REL.
5970 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5971 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5972 Add new macros 'UNSPEC_TEXT',
5973 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5974 + exclude function calls from 'UNSPEC_PLT' in case of data text
5976 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5977 new target hook for generating address diff vector tables in case of
5979 * doc/tm.texi : Regenerate.
5980 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5981 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5982 of addr diff vector generation.
5983 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5984 target hook definition.
5985 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5986 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5987 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5988 option for data text relative.
5990 2018-04-30 Richard Biener <rguenther@suse.de>
5992 * tree-chrec.h (evolution_function_is_constant_p): Remove
5994 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5996 2018-04-30 Richard Biener <rguenther@suse.de>
5999 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
6001 2018-04-30 Richard Biener <rguenther@suse.de>
6003 PR tree-optimization/28364
6004 PR tree-optimization/85275
6005 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
6006 copying first exit test.
6008 2018-04-28 Mark Wielaard <mark@klomp.org>
6010 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
6012 (dwarf_AT): Handle DW_AT_addr_base.
6013 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
6015 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
6018 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
6019 (*ashl<dwi>3_doubleword_mask_1): Ditto.
6020 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
6021 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
6023 2018-04-28 Richard Biener <rguenther@suse.de>
6025 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
6026 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
6027 to reflect use. Only add interesting stmts.
6029 2018-04-27 Martin Jambor <mjambor@suse.cz>
6032 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
6033 the jump function allows for passing through aggregate values.
6035 2018-04-27 David Malcolm <dmalcolm@redhat.com>
6037 * input.h (in_system_header_at): Convert from macro to inline
6039 (from_macro_expansion_at): Likewise.
6040 (from_macro_definition_at): Likewise.
6042 2018-04-27 Jeff Law <law@redhat.com>
6044 * config.gcc: Mark tile* targets as deprecated/obsolete.
6046 2018-04-27 Richard Biener <rguenther@suse.de>
6048 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
6051 2018-04-27 Richard Biener <rguenther@suse.de>
6053 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
6055 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
6057 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
6058 with Yd constraint. Set "preferred_for_speed" attribute from
6059 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
6061 (*movdi_internal): Ditto.
6062 (movti_interunit splitters): Remove
6063 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
6064 (movdi_interunit splitters): Ditto.
6065 * config/i386/constraints.md (Ye): Remove.
6066 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
6068 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6071 * config/aarch64/constraints.md (Usg): Limit to 31.
6074 2018-04-27 Jakub Jelinek <jakub@redhat.com>
6076 PR tree-optimization/85529
6077 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
6078 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
6079 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
6080 zero extension or masking of the MSB bit.
6081 (optimize_range_tests): Add FIRST_BB argument, pass it through
6082 to optimize_range_tests_var_bound.
6083 (maybe_optimize_range_tests, reassociate_bb): Adjust
6084 optimize_range_tests callers.
6086 2018-04-26 Richard Biener <rguenther@suse.de>
6087 Jakub Jelinek <jakub@redhat.com>
6089 * cgraph.h (symbol_table): Just declare debug method here.
6090 * symtab.c (symbol_table::debug): Define.
6092 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
6094 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
6096 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
6098 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
6099 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
6100 (*movdi_internal): Substitute Yi and Yj constraint with x
6101 and Ym and Yn constraint with y constraint. Update "isa"
6102 attribute and set "preferred_for_speed" attribute from
6103 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
6104 (*movsi_internal): Ditto.
6105 (*movdf_internal): Ditto.
6106 (*movsf_internal): Ditto.
6107 (*zero_extendsidi2): Ditto.
6108 * config/i386/sse.md (vec_set<mode>_0): Ditto.
6109 (sse2_loadld): Ditto.
6110 (*vec_extract<ssevecmodelower>_0): Ditto.
6111 (*vec_extractv4si_0_zext_sse4): Ditto.
6112 (vec_concatv2di): Ditto.
6113 (*vec_dup<mode>): Ditto.
6114 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
6115 * config/i386/constraints.md (Yi): Remove.
6120 2018-04-26 Nathan Sidwell <nathan@acm.org>
6122 * dumpfile.c (dump_open): New.
6123 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
6124 (dump_finish): Detect stdio/stderr by value not name.
6126 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6128 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6130 2018-04-26 Tom de Vries <tom@codesourcery.com>
6133 * config/nvptx/nvptx.c (verify_neutering_jumps)
6134 (verify_neutering_labels): New function
6135 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6137 2018-04-26 Tom de Vries <tom@codesourcery.com>
6140 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6141 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6144 2018-04-26 Richard Biener <rguenther@suse.de>
6145 Tom de Vries <tom@codesourcery.com>
6148 * lto-streamer-out.c (output_function): Fixup loops if required to match
6149 discovery done in the reader.
6151 2018-04-26 Richard Biener <rguenther@suse.de>
6153 PR tree-optimization/85116
6154 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6155 have a loop exit from the single latch predecessor. Remove
6156 case of header with just condition.
6157 (ch_base::copy_headers): Exclude infinite loops from any
6159 (pass_ch::execute): Record exits.
6161 2018-04-26 Richard Biener <rguenther@suse.de>
6163 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6164 prologue cost vector and pass it to vect_get_load_cost.
6165 (vect_get_peeling_costs_all_drs): Likewise.
6166 (vect_peeling_hash_get_lowest_cost): Likewise.
6167 (vect_enhance_data_refs_alignment): Likewise.
6169 2018-04-26 Richard Biener <rguenther@suse.de>
6172 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6173 checking of integer<->pointer conversions.
6174 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6175 sign-/zero-extending pointer types.
6176 (expand_omp_for_static_chunk): Likewise.
6178 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6179 Jean Lee <xiaoyur347@gmail.com>
6181 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6182 (TARGET_ASAN_SHADOW_OFFSET): Define.
6183 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6184 true for -fsanitize=address.
6186 2018-04-25 Mark Wielaard <mark@klomp.org>
6188 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6191 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6193 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6194 than "alu", remove explicit "memory" and "imm_disp" attributes.
6195 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6198 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6199 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6200 gen_lowpart_no_emit.
6202 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6205 * config/i386/i386.c (ix86_expand_builtin): Change memory
6206 operand to XI, extend p0 to Pmode.
6207 * config/i386/i386.md: Change unspec volatile and operand
6208 1 mode to XI, change operand 0 mode to P.
6210 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6212 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6213 GET_MODE_MASK before any checking.
6214 (nds32_can_use_bset_p): Likewise.
6215 (nds32_can_use_btgl_p): Likewise.
6217 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6219 * config/nds32/nds32-doubleword.md: New define_split pattern for
6220 illegal register number.
6222 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6224 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6226 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6228 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6230 2018-04-25 Richard Biener <rguenther@suse.de>
6232 * lto-streamer.h (LTO_major_version): Bump to 8.
6234 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6236 * BASE-VER: Set to 9.0.0.
6238 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6240 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6241 in __abskf2 and __powikf2.
6243 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6246 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6247 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6248 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6249 Use the above on operand 2. Reindent.
6250 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6252 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6255 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6257 * config/i386/i386.opt (mcet): Removed.
6258 * doc/install.texi: Remove -mcet documentation.
6259 * doc/invoke.texi: Likewise.
6261 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6264 * doc/install.texi: Remove -mcet from bootstrap-cet.
6266 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6269 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6270 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6274 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6275 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6276 containing a CONST_VECTOR.
6278 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6280 * doc/install.texi: Update newlib dependency for nvptx.
6282 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6285 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6286 instead of INTVAL when shifting x left.
6288 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6290 PR tree-optimization/85478
6291 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6292 vect_grouped_store_supported for single element vectors.
6294 2018-04-24 Richard Biener <rguenther@suse.de>
6297 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6298 load cost increase to the case of non-constant step.
6300 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6303 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6304 destination if any_malformed_asm.
6306 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6309 * expr.c (store_field): In the bitfield case, if the value comes from
6310 a function call and is returned in registers by means of a PARALLEL,
6311 do not change the mode of the temporary unless BLKmode and VOIDmode.
6313 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6315 PR rtl-optimization/85423
6316 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6317 dependencies to debug insns when the previous insn is non-debug.
6319 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6321 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6322 enums into a single definition.
6323 (fls): Fix predicates and printing.
6326 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6328 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6329 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6330 and short u6 immediate.
6331 (check_if_valid_sleep_operand): Remove.
6332 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6334 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6336 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6337 flag_always_save_lp condition.
6338 * config/nds32/nds32.opt (malways-save-lp): New option.
6340 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6342 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6343 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6344 * config/nds32/nds32.h
6345 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6346 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6348 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6350 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6351 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6353 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6354 Chung-Ju Wu <jasonwucj@gmail.com>
6356 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6358 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6359 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6361 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6363 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6365 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6367 * config/nds32/nds32-protos.h (nds32_data_alignment,
6368 nds32_local_alignment): Declare.
6369 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6370 nds32_local_alignment): New functions.
6371 (TARGET_CONSTANT_ALIGNMENT): Define.
6372 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6374 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6376 * config/nds32/nds32.c
6377 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6378 (TARGET_MODES_TIEABLE_P): Likewise.
6380 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6382 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6385 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6386 Chung-Ju Wu <jasonwucj@gmail.com>
6388 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6389 for unaligned access.
6390 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6392 * config/nds32/nds32-intrinsic.md: Likewise.
6393 * config/nds32/nds32_intrinsic.h: Likewise.
6394 * config/nds32/nds32.h (nds32_builtins): Likewise.
6395 * config/nds32/nds32.opt (munaligned-access): New option.
6396 * config/nds32/nds32.c (nds32_asm_file_start): Display
6397 flag_unaligned_access status.
6399 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6401 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6402 -mno-relax is present.
6403 * config/riscv/linux.h (LINK_SPEC): Ditto.
6405 2018-04-20 Martin Sebor <msebor@redhat.com>
6408 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6410 (gimple_fold_builtin_stxcpy_chk): Same.
6411 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6413 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6416 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6417 __powikf2 when long double is IEEE 128-bit.
6419 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6421 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6422 step to make sure stack always aligned.
6424 2018-04-20 Carl Love <cel@us.ibm.com>
6427 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6428 size check for arg0.
6430 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6431 Tom de Vries <tom@codesourcery.com>
6434 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6435 Emit insns for calls too.
6436 (nvptx_find_par): Always look for worker-level predecessor insn.
6437 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6439 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6440 (nvptx_process_pars): Propagate frames for calls.
6442 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6445 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6447 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6448 (ix86_handle_option): Don't handle OPT_mibt.
6449 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6451 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6453 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6454 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6455 (ix86_target_macros): Define __CET__ with flag_cf_protection
6456 for -fcf-protection.
6457 * config/i386/i386.c (isa2_opts): Remove -mibt.
6458 * config/i386/i386.h (TARGET_IBT): Removed.
6459 (TARGET_IBT_P): Likewise.
6460 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6461 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6462 * config/i386/i386.opt (mcet): Update help message.
6465 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6466 -mcet as an alias for -mshstk.
6468 2018-04-20 Richard Biener <rguenther@suse.de>
6471 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6472 complexity by forcing a single use of the multiply operand.
6474 2018-04-20 Martin Jambor <mjambor@suse.cz>
6477 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6478 recursion dependency to only apply to non-clones.
6480 2018-04-20 Martin Jambor <mjambor@suse.cz>
6483 * ipa-cp.c (create_specialized_node): Check that clones of
6484 self-recursive edges exist during IPA-CP.
6486 2018-04-19 Toon Moene <toon@moene.org>
6488 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6491 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6493 PR tree-optimization/85467
6494 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6495 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6496 VECTOR_CST element to type.
6498 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6501 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6502 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6503 (builtin_longjmp): Likewise.
6504 (save_stack_nonlocal): New pattern.
6505 (restore_stack_nonlocal): Likewise.
6507 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6510 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6511 Replace ASM_OUTPUT_LABEL with fprintf.
6513 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6516 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6517 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6518 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6519 define __IBT__ and __SHSTK__ for -fcf-protection.
6520 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6522 (ix86_trampoline_init): Likewise.
6523 (x86_output_mi_thunk): Likewise.
6524 (ix86_notrack_prefixed_insn_p): Likewise.
6525 (ix86_option_override_internal): Don't disallow -fcf-protection.
6526 * config/i386/i386.md (rdssp<mode>): Also enable for
6528 (incssp<mode>): Likewise.
6529 (nop_endbr): Likewise.
6530 * config/i386/i386.opt (mcet): Change help message to built-in
6534 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6535 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6536 enable CET built-in functions.
6538 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6540 * common/config/i386/i386-common.c
6541 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6542 OPTION_MASK_ISA_MOVDIRI_UNSET,
6543 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6544 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6545 * config.gcc (movdirintrin.h): New header.
6546 * config/i386/cpuid.h (bit_MOVDIRI,
6547 bit_MOVDIR64B): New bits.
6548 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6550 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6551 (VOID, PVOID, PCVOID)): New function types.
6552 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6553 __builtin_ia32_directstoreu_u64,
6554 __builtin_ia32_movdir64b): New builtins.
6555 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6556 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6558 (ix86_valid_target_attribute_inner_p): Ditto.
6559 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6560 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6561 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6562 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6563 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6564 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6565 (movdiri<mode>, movdir64b_<mode>): New.
6566 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6567 * config/i386/immintrin.h: Include movdirintrin.h.
6568 * config/i386/movdirintrin.h: New file.
6569 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6571 2018-04-19 Richard Biener <rguenther@suse.de>
6574 * cfg.c (clear_bb_flags): When loop state says we have
6575 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6577 2018-04-19 Richard Biener <rguenther@suse.de>
6579 PR tree-optimization/84737
6580 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6581 copying restrict info.
6582 (vect_setup_realignment): Use it.
6583 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6584 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6585 the first DR to all generated stores.
6586 (vectorizable_load): Likewise for loads.
6588 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6590 PR tree-optimization/85446
6591 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6592 the integral and pointer types to have the same precision.
6594 * doc/install.texi: Document --disable-cet being the default and
6597 2018-04-18 Martin Liska <mliska@suse.cz>
6599 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6602 2018-04-18 Martin Liska <mliska@suse.cz>
6605 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6608 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6609 arguments if they are comparable.
6611 2018-04-18 Martin Liska <mliska@suse.cz>
6614 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6617 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6620 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6623 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6624 ENDBR after calling __morestack.
6626 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6629 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6630 by using gcc_base_ver to generate a gcc_driver_version, and use
6631 it when generating GCC_DRIVER_NAME.
6632 * configure: Regenerate.
6634 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6637 * config.gcc: Obsolete powerpc*-*-*spe*.
6639 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6642 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6643 (stabstr_D): Change type of unum from unsigned int to
6644 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6647 2018-04-17 Jim Wilson <jimw@sifive.com>
6650 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6651 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6652 Set arg_pointer_offset after using pretend_args_size.
6654 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6656 PR rtl-optimization/85431
6657 * dse.c (record_store): Ignore zero width stores.
6660 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6661 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6662 __builtin_stack_restore rather than after it.
6663 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6664 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6665 argument instead of virtual_dynamic_stack_rtx.
6667 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6669 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6671 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6672 Add note to error message to explain internal mapping of overloaded
6673 built-in function name to non-overloaded built-in function name.
6674 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6677 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6680 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6681 where the inputs overlap with the output.
6683 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6685 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6686 (=v, v) alternative and explicit "memory" attribute.
6687 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6688 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6690 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6691 "sselog1" type instead of "sselog".
6692 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6693 "sselog". Remove explicit "memory" attribute.
6694 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6695 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6697 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6698 "sselog1" type instead of "sselog". Remove explicit "memory"
6700 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6701 use "sselog1" type instead of "sselog". Remove explicit "memory"
6703 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6704 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6706 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6707 "sselog1" type instead of "sselog". Remove explicit "memory"
6709 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6710 use "sselog1" type instead of "sselog". Remove explicit "memory"
6714 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6717 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6720 2018-04-17 Martin Jambor <mjambor@suse.cz>
6723 * ipa-cp.c (create_specialized_node): Call
6724 expand_all_artificial_thunks if necessary.
6726 2018-04-17 Martin Liska <mliska@suse.cz>
6729 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6730 in message, remote space in between '_G' and '('.
6732 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6735 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6736 avx512f_vmcmp<mode>3<round_saeonly_name>,
6737 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6738 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6739 avx512f_rndscale<mode><round_saeonly_name>,
6740 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6741 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6742 Use %<iptr>2 instead of %2 for -masm=intel.
6743 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6744 avx512f_vcvttss2usi<round_saeonly_name>,
6745 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6747 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6748 avx512f_vcvttsd2usi<round_saeonly_name>,
6749 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6750 Use %q1 instead of %1 for -masm=intel.
6751 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6752 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6753 of %3 for -masm=intel.
6754 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6756 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6758 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6760 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6761 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6763 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6765 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6766 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6767 %g1 and one with %0 and %1.
6768 (avx512er_vmrcp28<mode><round_saeonly_name>,
6769 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6771 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6772 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6773 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6774 of %0 and %{%4%} for -masm=intel.
6775 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6776 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6777 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6778 order of %0 and %{%5%}%{z%} for -masm=intel.
6780 2018-04-17 Jan Hubicka <jh@suse.cz>
6783 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6785 2018-04-17 Martin Liska <mliska@suse.cz>
6788 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6789 for target_clone error message. Make default implementation
6790 clone to be a local declaration.
6791 (separate_attrs): Add new argument and check for an empty
6793 (expand_target_clones): Handle it.
6794 (ipa_target_clone): Make redirection just for target_clones
6797 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6798 Tom de Vries <tom@codesourcery.com>
6801 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6802 tiled basic blocks without omp continue statements.
6804 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6807 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6808 vec_extract expression as having side effects to make sure it gets
6811 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6814 * config/i386/i386.c (get_builtin_code_for_version): Check
6817 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6820 * gcc/config.gcc: Support "skylake".
6821 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6823 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6824 (processor_target_table): Add "skylake".
6825 (ix86_option_override_internal): Add "skylake".
6826 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6827 PROCESSOR_CANNONLAKE.
6828 (get_builtin_code_for_version): Fix priority for
6829 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6830 PROCESSOR_SKYLAKE-AVX512.
6831 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6832 (processor_type): Add PROCESSOR_SKYLAKE.
6834 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6835 Jason Merrill <jason@redhat.com>
6838 * convert.c (convert_to_integer_1): Use direct recursion for
6839 enumeral types and types with a precision less than the number
6840 of bits in their mode.
6842 2018-04-16 Julia Koval <julia.koval@intel.com>
6845 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6846 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6848 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6851 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6852 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6853 and -mno-direct-move.
6855 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6858 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6859 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6860 Ensure negative shifts result in {0}.
6862 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6864 PR rtl-optimization/79916
6865 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6866 regs (if any) to define how to gnerate SD moves when LRA is in
6869 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6871 PR rtl-optimization/85393
6872 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6873 * except.c (expand_dw2_landing_pad_for_region): Make static.
6874 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6875 a label and unconditional jump to old_bb, rather than
6876 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6879 PR rtl-optimization/85376
6880 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6881 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6882 instead of a specific value.
6884 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6885 Bin Cheng <bin.cheng@arm.com>
6887 PR tree-optimization/82965
6888 PR tree-optimization/83991
6889 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6890 by_profile_only parameter.
6891 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6892 information if the loop was predicted to iterate too many times.
6893 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6895 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6898 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6901 2018-04-13 Martin Liska <mliska@suse.cz>
6902 Jakub Jelinek <jakub@redhat.com>
6905 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6906 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6907 * builtins.c (expand_builtin_memory_copy_args): Use
6908 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6909 handle dest_addr == pc_rtx.
6911 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6914 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6915 asked to not generate direct moves.
6916 (fix_trunc<mode>si2_stfiwx): Similar.
6917 (fix_trunc<mode>si2_internal): Similar.
6919 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6922 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6923 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6924 lookup if dest in some wider mode is known to be const0_rtx and
6925 if so, record permanent equivalence for it to be ZERO_EXTEND of
6926 the narrower mode destination.
6928 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6930 * lto-streamer-out.c (output_function): Revert 259346.
6931 * omp-expand.c (expand_oacc_for): Likewise.
6933 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6935 PR rtl-optimization/85354
6936 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6937 * sel-sched.c (sel_global_init): ... here.
6939 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6942 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6943 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6944 mode for PE-COFF targets.
6945 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6946 (i386_pe_asm_lto_end): Likewise.
6947 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6948 (TARGET_ASM_LTO_END): Likewise.
6949 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6950 (i386_pe_asm_lto_start): New function.
6951 (i386_pe_asm_lto_end): Likewise.
6953 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6954 Richard Biener <rguenther@suse.de>
6957 * lto-streamer-out.c (output_function): Fix CFG loop state before
6959 * omp-expand.c (expand_oacc_for): Handle calls to internal
6960 functions like regular functions.
6962 2018-04-12 Richard Biener <rguenther@suse.de>
6965 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6966 for the early LTO debug to properly generate references to it
6967 during DIE emission. Do not re-use that for the skeleton for
6969 (dwarf2out_early_finish): Likewise.
6971 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6974 * config/i386/sse.md
6975 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6976 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6977 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6978 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6979 and output is a reg, avoid creating invalid lowpart subreg, but
6980 instead split into a 512-bit move. Don't split if not AVX512VL,
6981 input is xmm16+ reg and output is a mem.
6982 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6983 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6984 xmm16+ reg and output is a mem.
6986 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6988 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6989 also for flag_dwarf2_cfi_asm.
6991 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6993 PR rtl-optimization/85342
6994 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6995 a bool scalar var inside of the loop instead. Don't try to update
6996 recog_data.operand after failed apply_change_group.
6998 2018-04-12 Tom de Vries <tom@codesourcery.com>
7001 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
7002 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
7003 array with flexible array member as array without given dimension.
7004 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
7005 argument for undefined param to true.
7007 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
7010 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7011 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
7012 from PowerPC section.
7013 * config/rs6000/sysv4.opt (mcall-): Improve help text.
7014 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
7015 help text that is too long.
7016 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
7017 help text that is too long.
7018 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
7019 help text that is too long.
7021 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
7023 * config/alpha/alpha.md (stack_probe_internal): Rename
7024 from "probe_stack". Update all callers.
7026 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7028 PR rtl-optimization/84566
7029 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
7030 sched_macro_fuse_insns.
7032 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7035 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
7036 (compute_block_dependences): ... from here.
7038 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7040 PR tree-optimization/85331
7041 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
7042 from int to HOST_WIDE_INT.
7044 2018-04-11 Martin Jambor <mjambor@suse.cz>
7047 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
7048 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
7049 not the same as the source val.
7050 (cgraph_edge_brings_value_p): New parameter.
7051 (gather_edges_for_value): Pass destination value to
7052 cgraph_edge_brings_value_p.
7053 (perhaps_add_new_callers): Likewise.
7054 (get_info_about_necessary_edges): Likewise and exclude values brought
7055 only by self-recursive edges.
7056 (create_specialized_node): Redirect only clones of self-calling edges.
7057 (+self_recursive_pass_through_p): New function.
7058 (find_more_scalar_values_for_callers_subset): Use it.
7059 (find_aggregate_values_for_callers_subset): Likewise.
7060 (known_aggs_to_agg_replacement_list): Removed.
7061 (decide_whether_version_node): Re-calculate known constants for all
7062 remaining context clones.
7064 2018-04-11 Richard Biener <rguenther@suse.de>
7067 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
7068 from early DWARF output.
7069 (dwarf2out_early_finish): Output line info unconditionally into
7070 early DWARF and add reference to it.
7072 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7075 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
7076 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
7077 other than V2DFmode using iptr mode attribute.
7078 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
7080 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
7082 PR rtl-optimization/84659
7083 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
7085 2018-04-11 Jakub Jelinek <jakub@redhat.com>
7088 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
7090 (output_loc_list): Pass address of a dummy size variable even in the
7091 locview handling loop.
7092 (index_location_lists): Add comment on why skip_loc_list_entry can't
7095 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
7098 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
7101 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
7104 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
7105 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
7106 and -mstring-compare-inline-limit.
7108 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7111 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
7112 for stack clash protection in a register whenever we need it to be in
7115 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7117 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
7118 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
7120 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
7123 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
7125 (mlong-double-): Ditto.
7126 * config/rs6000/sysv4.opt (msdata=): Ditto.
7127 (mtls-size=): Ditto.
7129 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7131 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7132 erroneous entries for
7133 "vector int vec_ldl (int, long int *)", and
7134 "vector unsigned int vec_ldl (int, unsigned long int *)".
7135 Add comments and entries for
7136 "vector bool char vec_ldl (int, bool char *)",
7137 "vector bool short vec_ldl (int, bool short *)",
7138 "vector bool int vec_ldl (int, bool int *)",
7139 "vector bool long long vec_ldl (int, bool long long *)",
7140 "vector pixel vec_ldl (int, pixel *)",
7141 "vector long long vec_ldl (int, long long *)",
7142 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7143 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7144 type tree bool_long_long_type_node and correct definition of
7145 bool_V2DI_type_node to make reference to this new type tree.
7146 (rs6000_mangle_type): Replace erroneous reference to
7147 bool_long_type_node with bool_long_long_type_node.
7148 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7149 comments to emphasize sign distinctions for char and int types and
7150 replace RS6000_BTI_bool_long constant with
7151 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7152 use of RS6000_BTI_pixel.
7153 (bool_long_type_node): Remove this macro definition.
7154 (bool_long_long_type_node): New macro definition
7156 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7158 PR rtl-optimization/85300
7159 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7160 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7161 simplify_unary_operation fails.
7163 2018-04-10 Martin Liska <mliska@suse.cz>
7165 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7166 cgraph_edge and ipa_ref.
7168 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7172 * config/i386/sse.md
7173 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7174 computation of the VEC_MERGE selector from mask.
7175 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7176 Fix decoding of the VEC_MERGE selector into mask.
7178 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7180 PR tree-optimization/85286
7181 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7183 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7185 * final.c (final_1): Set insn_last_address as well as
7186 insn_current_address.
7188 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7191 * explow.c (emit_stack_probe): Call validize_mem on memory location
7192 before passing it to gen_probe_stack. Create address operand and
7193 legitimize it for the probe_stack_address case.
7195 2018-04-09 Jan Hubicka <jh@suse.cz>
7198 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7199 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7200 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7202 (free_lang_data): Rebuild type inheritance graph.
7204 2018-04-09 Martin Sebor <msebor@redhat.com>
7206 * invoke.texi (-finline-small-functions): Mention other optimization
7208 (-findirect-inlining, -fpartial-inlining): Same.
7209 (-finline-functions-called-once): Same.
7210 (-freorder-blocks-and-partition): Same.
7212 2018-04-09 Jan Hubicka <jh@suse.cz>
7215 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7216 jumps; choose last target that matches the criteria (i.e.
7217 no partition changes for non-crossing jumps).
7218 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7219 support for redirecting crossing jumps to non-crossing.
7221 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7223 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7224 also for naked functions.
7226 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7228 * config/arc/arc.md (add_shift): New pattern.
7229 (add_shift2): Likewise.
7230 (sub_shift): Likewise.
7231 (sub_shift_cmp0_noout): Likewise.
7232 (compare_si_ashiftsi): Likewise.
7233 (xbfu_cmp0_noout): New combine pattern.
7234 (xbfu_cmp0"): Likewise.
7235 (movsi_set_cc_insn): Place the predicable variant first.
7236 (commutative_binary_cmp0_noout): Remove clobber.
7237 (commutative_binary_cmp0): New pattern.
7238 (noncommutative_binary_cmp0): Likewise.
7239 (noncommutative_binary_cmp0_noout): Likewise.
7240 (noncommutative_binary_comparison_result_used): Removed.
7241 (rsub_cmp0): New pattern.
7242 (rsub_cmp0_noout): Likewise.
7243 (extzvsi): Changed, keep only meaningful variants.
7244 (SQH, SEZ): New iterators.
7245 (SQH_postfix): New mode attribute.
7246 (SEZ_prefix): New code attribute.
7247 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7248 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7249 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7251 (noncommutative_operator): Check the availability of barrel
7254 2018-04-09 Richard Biener <rguenther@suse.de>
7256 PR tree-optimization/85284
7257 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7258 Only use the niter constraining form of simple_iv when the exit
7261 2018-04-09 Tom de Vries <tom@codesourcery.com>
7264 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7265 (define_expand "*memory_barrier"): New define_expand.
7266 (define_insn "memory_barrier"): New insn.
7268 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7270 PR rtl-optimization/80463
7271 PR rtl-optimization/83972
7272 PR rtl-optimization/83480
7274 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7275 correct producer for the insn.
7276 (tidy_control_flow): Fixup seqnos in case of debug insns.
7278 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7280 PR rtl-optimization/83913
7282 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7283 different sched-times when merging exprs.
7285 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7287 PR rtl-optimization/83962
7289 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7290 tidy_fallthru_edge and tidy_control_flow.
7292 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7294 PR rtl-optimization/83530
7296 * sel-sched.c (force_next_insn): New global variable.
7297 (remove_insn_for_debug): When force_next_insn is true, also leave only
7298 next insn in the ready list.
7299 (sel_sched_region): When the region wasn't scheduled, make another pass
7300 over it with force_next_insn set to 1.
7302 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7304 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7306 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7307 for interrupt control.
7308 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7309 functions for interrupt control.
7310 * config/nds32/nds32-intrinsic.md: Likewise.
7311 * config/nds32/nds32_intrinsic.h: Likewise.
7312 * config/nds32/nds32.h (nds32_builtins): Likewise.
7314 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7316 * config/nds32/nds32.c (nds32_init_machine_status,
7317 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7318 strict_aligned_p field.
7319 (nds32_expand_to_rtl_hook): New function.
7320 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7321 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7323 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7324 Chung-Ju Wu <jasonwucj@gmail.com>
7326 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7327 * config/nds32/nds32-n7.md: New file.
7328 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7329 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7331 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7332 * config/nds32/nds32.md (pipeline_model): Add n7.
7333 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7334 * config/nds32/pipelines.md: Include n7 settings.
7336 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7337 Chung-Ju Wu <jasonwucj@gmail.com>
7339 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7340 * config/nds32/nds32-e8.md: New file.
7341 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7342 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7344 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7345 * config/nds32/nds32.md (pipeline_model): Add e8.
7346 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7347 * config/nds32/pipelines.md: Include e8 settings.
7349 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7350 Chung-Ju Wu <jasonwucj@gmail.com>
7352 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7353 * config/nds32/nds32-n8.md: New file.
7354 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7355 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7357 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7358 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7359 * config/nds32/nds32.md (pipeline_model): Add n8.
7360 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7361 * config/nds32/pipelines.md: Include n8 settings.
7363 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7364 Chung-Ju Wu <jasonwucj@gmail.com>
7366 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7367 * config/nds32/nds32-n9-2r1w.md: New file.
7368 * config/nds32/nds32-n9-3r2w.md: New file.
7369 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7370 nds32_register_ports): New or modify for cpu n9.
7371 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7373 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7374 * config/nds32/nds32-utils.c: New file.
7375 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7376 TARGET_MUL_SLOW): Define.
7377 * config/nds32/nds32.md (pipeline_model): New attribute.
7378 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7379 New options that support cpu n9.
7380 * config/nds32/pipelines.md: Include n9 settings.
7381 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7383 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7385 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7386 information if necessary.
7387 (output_cond_branch_compare_zero): Likewise.
7388 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7389 (nds32_target_alignment): Refine for alignment.
7390 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7391 (FUNCTION_BOUNDARY): Modify.
7392 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7394 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7396 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7398 * config/nds32/constants.md (unspec_volatile_element): Add values for
7399 TLB operation and data prefetch.
7400 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7401 functions for TLB operation and data prefetch.
7402 * config/nds32/nds32-intrinsic.md: Likewise.
7403 * config/nds32/nds32_intrinsic.h: Likewise.
7404 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7405 (nds32_print_operand): Likewise.
7406 * config/nds32/nds32.h (nds32_builtins): Likewise.
7408 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7409 Andrew Pinski <pinsika@gcc.gnu.org>
7412 * match.pd: Use constant_boolean_node of correct type instead of
7413 boolean_true_node or boolean_false_node for simplifying
7414 pointer comparisons to zero.
7416 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7418 PR tree-optimization/80021
7419 * tree.c (verify_type_variant): Make error call in verify_variant_match
7420 translatable and remove final full stop.
7422 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7424 * config/nds32/constants.md (unspec_volatile_element): Add
7425 UNSPEC_VOLATILE_EH_RETURN.
7426 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7427 nds32_output_stack_pop): Support dwarf exception handling process.
7428 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7429 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7430 exception handling process.
7431 (nds32_compute_stack_frame): Likewise.
7432 (nds32_return_addr_rtx): Likewise.
7433 (nds32_initial_elimination_offset): Likewise.
7434 (nds32_expand_prologue): Likewise.
7435 (nds32_expand_epilogue): Likewise.
7436 (nds32_dynamic_chain_address): New function.
7437 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7439 (DYNAMIC_CHAIN_ADDRESS): Define.
7440 (EH_RETURN_DATA_REGNO): Define.
7441 (EH_RETURN_STACKADJ_RTX): Define.
7442 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7443 patterns for dwarf exception handling.
7445 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7447 * config/nds32/nds32.h: Clean up obsolete macros.
7449 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7451 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7452 Add enum values for particular instructions.
7453 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7454 particular intrinsic functions.
7455 * config/nds32/nds32-intrinsic.md: Likewise.
7456 * config/nds32/nds32_intrinsic.h: Likewise.
7457 * config/nds32/nds32.h (nds32_builtins): Likewise.
7458 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7459 (btst, ave): New patterns for particular instructions.
7461 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7463 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7464 Add enum values for atomic load/store and memory sync.
7465 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7467 * config/nds32/nds32-intrinsic.md: Likewise.
7468 * config/nds32/nds32_intrinsic.h: Likewise.
7469 * config/nds32/nds32.h (nds32_builtins): Likewise.
7471 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7473 PR tree-optimization/85257
7474 * fold-const.c (native_encode_vector): If not all elts could fit
7475 and off is -1, return 0 rather than offset.
7476 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7477 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7478 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7479 adjust buffer in native_interpret_expr call.
7481 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7483 * config/nds32/constants.md (unspec_volatile_element): Add cache
7484 control enum values.
7485 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7486 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7487 * config/nds32/nds32.c (nds32_cctl_names): New.
7488 (nds32_print_operand): Handle cache control register names.
7489 * config/nds32/nds32.h (nds32_builtins): New enum values.
7490 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7492 * config/nds32/nds32.md (type): Add mmu.
7493 * config/nds32/pipelines.md (simple_insn): Add mmu.
7495 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7497 * config/nds32/nds32.md (type): Remove call.
7498 * config/nds32/pipelines.md (simple_insn): Likewise.
7500 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7502 * config/nds32/constants.md (unspec_volatile_element): Add
7503 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7504 UNSPEC_VOLATILE_FMFCFG.
7505 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7506 description for fmfcfg and fmfcsr.
7507 (bdesc_1arg): Add fmtcsr.
7508 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7509 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7510 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7511 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7512 unspec_fmfcfg): New patterns.
7513 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7514 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7515 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7516 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7517 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7518 __nds32__fmfcfg): Define.
7520 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7522 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7523 intrinsic register names.
7524 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7525 intrinsic register enum values and macros.
7527 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7529 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7530 for load/store addressing form.
7531 (nds32_print_operand_address): Likewise.
7533 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7536 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7537 based on LABEL_REF. Remove useless assertion.
7538 (pic_address_needs_scratch): Fix formatting.
7539 (sparc_legitimize_pic_address): Minor tweaks.
7540 (sparc_delegitimize_address): Adjust assertion accordingly.
7541 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7542 into symbolic_operand.
7543 (movsi_high_pic_label_ref): Likewise.
7544 (movsi_lo_sum_pic_label_ref): Likewise.
7545 (movdi_pic_label_ref): Likewise.
7546 (movdi_high_pic_label_ref): Likewise.
7547 (movdi_lo_sum_pic_label_ref): Likewise.
7549 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7551 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7552 custom LIB_SPEC setup.
7554 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7555 Kito Cheng <kito.cheng@gmail.com>
7557 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7558 * config/riscv/freebsd.h: New.
7560 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7562 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7563 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7566 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7567 Kito Cheng <kito.cheng@gmail.com>
7569 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7570 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7571 * config/nds32/nds32-protos.h (nds32_output_call,
7572 nds32_output_return): Declare.
7573 * config/nds32/nds32.md: Refine all the call and return patterns.
7575 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7578 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7579 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7581 PR rtl-optimization/84872
7582 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7583 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7586 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7588 * expr.c (copy_blkmode_to_reg): Revert 254862.
7589 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7591 2018-04-06 Richard Biener <rguenther@suse.de>
7594 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7595 after seeing a component reference with an adjacent field. Treat
7596 refs to arrays at struct end of external decls similar to
7597 refs to unconstrained commons.
7599 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7602 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7603 look through SAVE_EXPRs with non-side-effects argument. Adjust
7605 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7608 2018-04-06 Richard Biener <rguenther@suse.de>
7611 * alias.c (find_base_term): New wrapper around find_base_term
7612 unwinding CSELIB_VAL_PTR changes.
7613 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7616 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7618 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7620 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7621 constant definitions.
7622 ("nop"): lr 0,0 -> nopr r0
7623 ("nop_lr0", "nop_lr1"): New insn definitions.
7625 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7627 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7628 NDS32_V3PUSH_AVAILABLE_P macro.
7630 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7631 Chung-Ju Wu <jasonwucj@gmail.com>
7633 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7634 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7635 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7636 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7637 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7638 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7639 * config/nds32/constraints.md: New constraints and checking for hard
7640 float configuration.
7641 * config/nds32/iterators.md: New mode iterator and attribute for hard
7642 float configuration.
7643 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7645 * config/nds32/nds32-fpu.md: New file.
7646 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7647 deal with hard float code generation.
7648 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7650 (abi_type, float_reg_number): New enum type.
7651 * config/nds32/nds32-predicates.c: New predicates for hard float.
7652 * config/nds32/nds32-protos.h: Declare functions for hard float.
7653 * config/nds32/nds32.c: Implementation for hard float configuration.
7654 * config/nds32/nds32.h: Definitions for hard float configuration.
7655 * config/nds32/nds32.md: Include hard float machine description and
7656 modify patterns for hard float configuration.
7657 * config/nds32/nds32.opt: New options for hard float configuration.
7658 * config/nds32/predicates.md: New predicates for hard float
7661 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7663 * common/config/nds32/nds32-common.c
7664 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7666 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7669 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7670 CONSTRUCTOR_ELT (ctor, ...)->value.
7672 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7675 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7677 2018-04-05 Tom de Vries <tom@codesourcery.com>
7680 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7683 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7684 Kito Cheng <kito.cheng@gmail.com>
7686 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7687 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7688 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7689 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7691 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7692 Kito Cheng <kito.cheng@gmail.com>
7694 * config/nds32/constraints.md (Ufe): New memory constraint.
7695 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7696 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7697 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7699 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7700 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7702 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7704 * config/nds32/nds32.md: Use optimize_size in the condition for
7705 alu-shift instructions.
7707 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7709 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7711 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7713 * config/nds32/nds32.md (negsi2): Refine pattern.
7715 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7716 Chung-Ju Wu <jasonwucj@gmail.com>
7718 * config/nds32/iterators.md (shift_rotate): New code iterator.
7719 (shift): New code attribute.
7720 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7721 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7722 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7723 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7724 bit-wise operations.
7725 (andsi3, *andsi3): Ditto.
7726 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7727 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7728 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7729 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7730 nds32_ior_operand, nds32_xor_operand): New predicates.
7732 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7734 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7735 (addsi3, subsi3): ... this.
7737 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7739 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7741 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7743 * config/nds32/nds32.md: Adjust indention.
7745 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7747 * config/nds32/nds32.md (feature): New attribute.
7749 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7751 * config/nds32/nds32.md (subtype): New attribute.
7753 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7756 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7757 expansion to perform a bitwise AND of the argument followed by a
7758 boolean negation of the result.
7760 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7762 PR rtl-optimization/84878
7763 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7764 the basic block. Assert the use reference is not artificial and that
7765 it has an associated insn.
7767 2018-04-04 Michael Matz <matz@suse.de>
7769 * builtins.c (compute_objsize): Pass correct operand
7770 to array_at_struct_end_p.
7772 2018-04-04 Richard Biener <rguenther@suse.de>
7775 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7776 from contexts for DINFO_LEVEL_TERSE and below.
7778 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7780 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7781 resiter_operand condition.
7782 * config/nds32/nds32.md (*move<mode>): Ditto.
7784 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7785 Monk Chiang <sh.chiang04@gmail.com>
7787 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7789 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7791 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7793 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7794 Kito Cheng <kito.cheng@gmail.com>
7796 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7797 nds32_cond_code_str, output_cond_branch,
7798 output_cond_branch_compare_zero, nds32_expand_cbranch,
7799 nds32_expand_cstore, nds32_expand_movcc,
7800 nds32_output_cbranchsi4_equality_zero,
7801 nds32_output_cbranchsi4_equality_reg,
7802 nds32_output_cbranchsi4_equality_reg_or_const_int,
7803 nds32_output_cbranchsi4_greater_less_zero: New functions.
7804 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7805 nds32_expand_cstore, nds32_expand_movcc,
7806 nds32_output_cbranchsi4_equality_zero,
7807 nds32_output_cbranchsi4_equality_reg,
7808 nds32_output_cbranchsi4_equality_reg_or_const_int,
7809 nds32_output_cbranchsi4_greater_less_zero): Declare.
7810 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7811 nds32_rimm11s_operand): New predicates.
7812 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7813 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7816 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7818 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7819 * config/nds32/nds32.md: Ditto.
7820 * config/nds32/pipelines.md: Ditto.
7822 2018-04-04 Richard Biener <rguenther@suse.de>
7824 PR tree-optimization/85168
7825 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7826 propagating abnormals.
7828 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7830 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7832 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7833 Kito Cheng <kito.cheng@gmail.com>
7835 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7836 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7837 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7838 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7839 * config/nds32/nds32.md (sibcall_internal): New.
7840 (sibcall_register): Remove.
7841 (sibcall_immediate): Remove.
7842 (sibcall_value_internal): New.
7843 (sibcall_value_register): Remove.
7844 (sibcall_value_immediate): Remove.
7845 * config/nds32/predicates.md (nds32_general_register_operand): New.
7846 (nds32_call_address_operand): New.
7848 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7850 PR rtl-optimization/85167
7851 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7852 bb_defs if *split_p, instead preinitialize it to NULL.
7854 PR tree-optimization/85156
7855 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7856 evaluating the argument multiple times.
7858 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7860 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7862 (_mm_cvtpd_ps): Likewise.
7863 (_mm_cvttpd_epi32): Likewise.
7864 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7865 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7866 vector, pixel, and bool following altivec.h include.
7868 2018-04-03 Martin Sebor <msebor@redhat.com>
7870 * doc/extend.texi (Common Function Attributes): Clarify.
7871 (const attribute): Likewise.
7872 (pure attribute): Likewise.
7874 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7877 * config/i386/i386.c (ix86_expand_vector_set): Use
7878 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7880 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7882 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7883 instructions when changing rounding bits to preserve precision bits
7884 in the x87 control word.
7886 2018-04-03 Martin Liska <mliska@suse.cz>
7888 PR tree-optimization/82491
7889 * rtl.h (strip_offset_and_add): Replace += suboffset with
7890 poly_uint64 () + suboffset.
7892 2018-03-29 Martin Liska <mliska@suse.cz>
7893 Martin Jambor <mjambor@suse.cz>
7896 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7897 param_type is not an integral or pointer type.
7899 2018-04-03 Richard Biener <rguenther@suse.de>
7901 * sese.h (recompute_all_dominators): Remove.
7903 2018-04-02 Martin Sebor <msebor@redhat.com>
7905 * doc/invoke.texi (-Wrestrict): Fix typos.
7907 2018-04-02 Jim Wilson <jimw@sifive.com>
7909 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7910 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7911 (<optab>di3, <optab>si3_extend): Likewise.
7912 (<optab>si3_mask, <optab>si3_mask_1): New.
7913 (<optab>di3_mask, <optab>di3_mask_1): New.
7914 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7915 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7916 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7918 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7920 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7923 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7925 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7926 (nds32_canonicalize_comparison): New function.
7928 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7929 Kito Cheng <kito.cheng@gmail.com>
7930 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7932 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7933 * config/nds32/constants.md (unspec_volatile_element): Add
7934 UNSPEC_VOLATILE_RELAX_GROUP.
7935 * config/nds32/nds32-relax-opt.c: New file.
7936 * config/nds32/nds32-predicates.c
7937 (nds32_symbol_load_store_p): New function.
7938 * config/nds32/nds32-protos.h
7939 (nds32_symbol_load_store_p): Declare function.
7940 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7941 * config/nds32/nds32.c
7942 (nds32_register_pass): New function to register pass.
7943 (nds32_register_passes): New function to register passes.
7944 * config/nds32/nds32.md (relax_group): New pattern.
7945 * config/nds32/nds32.opt (mrelax-hint): New option.
7946 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7948 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7950 * config/nds32/t-nds32: Modify files dependency.
7952 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7954 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7955 (PROFILE_HOOK): Define its implementation.
7957 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7959 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7960 type and 32-bit size.
7962 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7965 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7966 (V_128_256): New mode iterator.
7967 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7968 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7969 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7971 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7972 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7974 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7977 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7978 NaN inputs correctly.
7980 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7983 * config/rs6000/vsx.md (??r): New mode attribute.
7984 (*vsx_mov<mode>_64bit): Use it.
7985 (*vsx_mov<mode>_32bit): Likewise.
7987 2018-03-30 Martin Sebor <msebor@redhat.com>
7989 PR tree-optimization/84818
7990 * builtins.c (check_access): Use warning_n.
7992 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7995 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7997 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
8000 2018-03-30 Julia Koval <julia.koval@intel.com>
8003 * x86-tune.def (movx, partial_reg_dependency): Enable for
8006 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
8009 * lra-constraints.c (process_alt_operands): Move setting
8010 this_alternative_matches below.
8012 2018-03-29 Martin Liska <mliska@suse.cz>
8015 * doc/invoke.texi: Document how LTO works with debug info.
8016 Describe auto-load support of binutils. Mention 'x86-64'
8017 as valid option value of -march option.
8019 2018-03-29 Jakub Jelinek <jakub@redhat.com>
8021 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
8024 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
8025 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
8026 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
8029 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8032 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
8033 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
8034 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
8035 for RS6000_BTM_POWERPC64.
8036 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
8037 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
8038 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
8043 2018-03-28 Carl Love <cel@us.ibm.com>
8046 2017-09-27 Carl Love <cel@us.ibm.com>
8048 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
8049 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
8050 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
8053 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8055 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
8056 instead of __vector bool.
8057 (_mm_max_pu8): Likewise.
8058 (_mm_min_pi16): Likewise.
8060 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
8063 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
8064 (DIVWEUO): Likewise.
8066 (DIVDEUO): Likewise.
8067 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
8068 DIVWEUO and DIVDEUO.
8069 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
8070 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
8071 (div_extend): Likewise.
8072 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
8074 (__builtin_divweuo): Likewise.
8075 (__builtin_divdeo): Likewise.
8076 (__builtin_divdeuo): Likewise.
8078 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8081 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
8082 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
8084 PR tree-optimization/82004
8085 * gimple-match-head.c (optimize_pow_to_exp): New function.
8086 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
8087 Don't fold to exp if optimize_pow_to_exp is false.
8089 2018-03-28 Martin Liska <mliska@suse.cz>
8092 * calls.c (initialize_argument_information): Fix trailing space.
8093 * common.opt: Fix typo and provide better explanation for
8094 -fsanitize-coverage option.
8095 * config/i386/i386.opt: Fix typo.
8097 2018-03-28 Jakub Jelinek <jakub@redhat.com>
8098 Martin Liska <mliska@suse.cz>
8101 * gimplify.c (asan_poison_variable): Don't do the check for
8102 gimplify_omp_ctxp here.
8103 (gimplify_decl_expr): Do it here.
8104 (gimplify_target_expr): Likewise.
8106 2018-03-28 Martin Liska <mliska@suse.cz>
8109 * config/i386/i386.c (ix86_function_arg_advance): Do not call
8110 chkp_type_bounds_count if MPX is not enabled.
8112 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
8114 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
8116 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
8119 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
8120 function to create the function decl for complex long double
8121 multiply and divide for -mabi=ieeelongdouble.
8122 (init_float128_ieee): Call it.
8124 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8127 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8128 -fcf-protection=branch -mibt.
8129 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8131 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8134 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8135 arm_word_relocations.
8137 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8140 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8141 extern array declarations.
8143 2018-03-27 Richard Biener <rguenther@suse.de>
8146 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8147 explicit single_use checks.
8149 2018-03-27 Richard Biener <rguenther@suse.de>
8151 PR tree-optimization/85082
8152 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8155 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8157 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8158 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8159 Turn on fasynchronous-unwind-tables and funwind-tables.
8161 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8164 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8165 (*bmi_blsr_<mode>_ccz): Ditto.
8167 2018-03-26 Tom de Vries <tom@codesourcery.com>
8169 PR tree-optimization/85063
8170 * omp-general.c (offloading_function_p): New function. Factor out
8172 * omp-offload.c (pass_omp_target_link::gate): ... here.
8173 * omp-general.h (offloading_function_p): Declare.
8174 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8175 with attribute omp declare target for offloading functions.
8177 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8179 PR tree-optimization/84005
8180 * tree-data-ref.h (get_base_for_alignment): Declare.
8181 * tree-data-ref.c (get_base_for_alignment_1): New function.
8182 (get_base_for_alignment): Likewise.
8183 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8184 get_base_for_alignment to find a suitable base object, instead
8185 of always using drb->base_address.
8187 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8190 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8191 known size by default.
8193 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8196 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8197 and non BLKmode operands.
8199 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8202 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8203 Clean up attributes.
8205 2018-03-23 Richard Biener <rguenther@suse.de>
8208 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8209 we are going to emit early debug for LTO.
8211 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8214 * function.c (match_asm_constraints_1): Don't optimize if input
8215 doesn't satisfy general_operand predicate for output's mode.
8218 * alias.c (write_dependence_p): Don't require for x_canonicalized
8219 non-VOIDmode if x has VOIDmode.
8222 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8223 just don't try to optimize it rather than assert it never happens.
8225 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8227 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8228 macro expansions for definition of ST_INTERNAL_<mode> and
8229 LD_INTERNAL_<mode> builtins.
8230 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8232 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8234 (altivec_expand_st_builtin): Likewise.
8235 (altivec_expand_builtin): Remove calls to deleted functions.
8236 (rs6000_address_for_altivec): Delete this function.
8237 * config/rs6000/vector.md: Remove expands for
8238 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8240 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8243 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8244 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8245 re-computing once computed.
8246 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8247 (arm_init_machine_status): Initialize
8248 machine->static_chain_stack_bytes.
8250 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8253 * doc/extend.texi: Add four new prototypes for vec_ld.
8254 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8255 definitions for more logical presentation.
8256 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8257 entries for V1TI variants of __builtin_altivec_ld builtin.
8258 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8259 handling of V1TI variant of LVX icode pattern.
8260 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8261 (rs6000_gimple_fold_builtin): Likewise.
8262 (altivec_init_builtins): Add code to define
8263 __builtin_altivec_lvx_v1ti function.
8265 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8268 * function.c (match_asm_constraints_1): Don't do the optimization
8269 if input isn't a REG, SUBREG, MEM or constant.
8271 2018-03-22 Tom de Vries <tom@codesourcery.com>
8273 PR tree-optimization/84956
8274 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8275 bb_has_abnormal_pred.
8277 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8280 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8281 DECL_INITIAL (decl) to decl at the end.
8282 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8285 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8287 * doc/extend.texi (__builtin_tgmath): Document when complex
8288 integer types are treated as _Complex _Float64.
8290 2018-03-21 Tom de Vries <tom@codesourcery.com>
8292 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8294 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8296 PR tree-optimization/84960
8297 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8298 if it is ENTRY block, move them into single succ of ENTRY in that case.
8300 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8302 PR tree-optimization/84811
8303 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8304 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8305 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8306 (known_subrange_p): Update accordingly. Cast each value involved
8307 in the size comparison, rather than casting the result of the
8310 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8312 PR tree-optimization/84982
8313 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8314 by flipping the least significant bit rather than all bits from
8315 bitpos to bitpos + bitsize - 1.
8317 2018-03-21 Nathan Sidwell <nathan@acm.org>
8319 * doc/extend.texi (Deprecated Features): Remove mention of
8320 long-deleted deprecations.
8322 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8325 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8326 * configure: Regenerate.
8328 2018-03-21 Tom de Vries <tom@codesourcery.com>
8330 PR tree-optimization/83126
8331 * tree-parloops.c (num_phis): New function.
8332 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8334 2018-03-21 Nathan Sidwell <nathan@acm.org>
8336 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8337 mention anon-struct/union members and trailing attributes.
8339 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8341 PR tree-optimization/84969
8342 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8343 builtin memset partitions if they set different rhs values.
8345 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8347 PR rtl-optimization/84989
8348 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8349 VEC_DUPLICATE with scalar result mode.
8351 2018-03-21 Martin Liska <mliska@suse.cz>
8354 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8355 not intended return statement.
8357 2018-03-21 Martin Liska <mliska@suse.cz>
8360 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8361 (chkp_find_bound_slots_1): Limit number of iterations.
8363 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8366 * Minor grammar fixes for x86 options.
8368 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8371 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8372 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8374 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8377 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8378 (altivec_lvx_<mode>_1op): Likewise.
8379 (altivec_stvx_<mode>_2op): Likewise.
8380 (altivec_stvx_<mode>_1op): Likewise.
8381 (altivec_lvx_<VM2:mode>): New define_expand.
8382 (altivec_stvx_<VM2:mode>): Likewise.
8383 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8384 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8385 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8386 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8387 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8388 (rs6000_gen_lvx): Likewise.
8389 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8390 (altivec_expand_stv_builtin): Likewise.
8391 (altivec_expand_builtin): Likewise.
8392 * config/rs6000/vector.md: Likewise.
8394 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8397 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8400 2018-03-20 Richard Biener <rguenther@suse.de>
8403 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8404 sign-conversions as zero, fall back to standard scalar_stmt
8407 2018-03-20 Martin Liska <mliska@suse.cz>
8410 * predict.c (rebuild_frequencies): Handle case when we have
8411 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8413 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8416 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8417 flag_section_anchors.
8418 * varasm.c (use_blocks_for_decl_p): Remove hack for
8419 dw2_force_const_mem.
8422 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8424 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8425 be created, use lowpart_subreg of operands[0] rather than operands[0]
8427 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8428 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8429 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8430 and n constraint instead of aarch64_shift_imm_di and Usd.
8431 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8432 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8434 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8437 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8438 to favor GPR over NEON registers.
8439 (<shift>di3_neon): Likewise.
8441 2018-03-20 Tom de Vries <tom@codesourcery.com>
8444 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8445 (nvptx_process_pars): Emit bar.sync asap and alap.
8447 2018-03-20 Tom de Vries <tom@codesourcery.com>
8450 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8451 seen_label if seen_label is already set.
8453 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8456 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8457 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8458 Use 1U instead of 1. Formatting fixes.
8461 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8462 instead of TREE_TYPE (s1) for the return value.
8464 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8466 PR tree-optimization/84946
8467 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8468 bitsize + bitsize in poly_uint64 rather than poly_int64.
8471 * dwarf2asm.c: Include fold-const.c.
8472 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8473 of decl rather than decl itself.
8475 PR rtl-optimization/84643
8476 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8478 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8481 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8482 calling assemble_variable.
8484 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8487 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8488 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8490 2018-03-19 Jim Wilson <jimw@sifive.com>
8493 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8494 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8495 (riscv_first_stack_step): Likewise.
8496 (riscv_option_override): Use STACK_BOUNDARY instead of
8498 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8500 (BIGGEST_ALIGNMENT): Set to 128.
8501 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8502 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8505 2018-03-19 Richard Biener <rguenther@suse.de>
8507 PR tree-optimization/84933
8508 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8509 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8511 2018-03-19 Richard Biener <rguenther@suse.de>
8513 PR tree-optimization/84859
8514 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8515 (cond_if_else_store_replacement): Perform sinking operation on
8516 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8517 Generalize what a BB with a single eligible store is.
8519 2018-03-19 Richard Biener <rguenther@suse.de>
8521 PR tree-optimization/84929
8522 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8523 chrec_is_positive against non-chrec arg.
8525 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8528 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8530 2018-03-18 Martin Liska <mliska@suse.cz>
8532 PR rtl-optimization/84635
8533 * regrename.c (build_def_use): Use matches_mode only when
8536 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8538 PR tree-optimization/84913
8539 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8540 vectorize chains of COND_EXPRs.
8542 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8544 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8546 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8548 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8550 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8552 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8554 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8555 Kito Cheng <kito.cheng@gmail.com>
8557 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8558 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8559 (nds32_adjust_reg_alloc_order): New function.
8560 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8562 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8564 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8565 nds32_print_operand, nds32_print_operand_address): Use
8566 HOST_WIDE_INT_PRINT_DEC instead.
8568 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8570 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8572 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8575 * config/i386/i386.c (initial_ix86_tune_features,
8576 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8578 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8579 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8580 rather than 1u << ix86_tune. Formatting fix.
8581 (ix86_option_override_internal): Change ix86_arch_mask from
8582 unsigned int to unsigned HOST_WIDE_INT, initialize to
8583 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8584 (ix86_function_specific_restore): Likewise.
8586 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8589 * postreload.c (reload_combine_recognize_pattern): Perform
8590 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8591 truncate_int_for_mode the result for the destination's mode.
8594 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8595 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8597 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8600 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8603 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8604 regno_allocno_class_array and sorted_pseudos.
8605 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8606 insns where regno is used.
8608 2018-03-16 Martin Liska <mliska@suse.cz>
8611 * multiple_target.c (create_dispatcher_calls): Redirect
8612 reference in the symbol table.
8614 2018-03-16 Martin Liska <mliska@suse.cz>
8617 * multiple_target.c (create_dispatcher_calls): Redirect also
8620 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8624 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8625 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8626 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8628 2018-03-16 Julia Koval <julia.koval@intel.com>
8630 * doc/invoke.texi (Skylake Server): Add CLWB.
8631 Cannonlake): Remove CLWB.
8633 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8635 PR tree-optimization/84841
8636 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8638 (FLOAT_ONE_CONST_TYPE): Define.
8639 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8640 (sort_by_operand_rank): Put entries with higher constant_type last
8641 rather than first to match comments.
8643 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8645 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8648 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8651 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8652 check and use address of target always.
8654 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8657 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8658 (indirect_thunk_bnd_needed): Likewise.
8659 (indirect_thunks_used): Likewise.
8660 (indirect_thunks_bnd_used): Likewise.
8661 (indirect_return_needed): New.
8662 (indirect_return_bnd_needed): Likewise.
8663 (output_indirect_thunk_function): Add a bool argument for
8665 (output_indirect_thunk_function): Don't generate alias for
8666 function return thunk.
8667 (ix86_code_end): Call output_indirect_thunk_function to generate
8668 function return thunks.
8669 (ix86_output_function_return): Set indirect_return_bnd_needed
8670 and indirect_return_needed instead of indirect_thunk_bnd_needed
8671 and indirect_thunk_needed.
8673 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8675 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8676 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8677 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8679 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8680 Paul Hua <paul.hua.gm@gmail.com>
8683 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8685 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8687 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8688 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8691 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8694 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8695 instead of GET_MODE_SIZE when comparing Units.
8697 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8700 * varasm.c (hash_section): Return an unchangeble hash value
8701 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8702 Return !aarch64_can_use_per_function_literal_pools_p ().
8704 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8707 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8708 rather than address of cmode as last argument to prepare_cmp_insn.
8710 2018-03-15 Julia Koval <julia.koval@intel.com>
8712 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8713 F_AVX512VNNI, F_AVX512BITALG): New.
8715 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8718 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8719 insn for floating-point loads and stores.
8721 2018-03-14 Carl Love <cel@us.ibm.com>
8723 * config/rs6000/rs6000-c.c: Add macro definitions for
8724 ALTIVEC_BUILTIN_VEC_PERMXOR.
8725 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8726 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8727 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8728 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8730 * config/doc/extend.texi: Add prototypes for vec_permxor.
8732 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8735 * diagnostic-show-locus.c (class layout_point): Convert m_line
8736 from int to linenum_type.
8737 (line_span::comparator): Use linenum "compare" function when
8738 comparing line numbers.
8739 (test_line_span): New function.
8740 (layout_range::contains_point): Convert param "row" from int to
8742 (layout_range::intersects_line_p): Likewise.
8743 (layout::will_show_line_p): Likewise.
8744 (layout::print_source_line): Likewise.
8745 (layout::should_print_annotation_line_p): Likewise.
8746 (layout::print_annotation_line): Likewise.
8747 (layout::print_leading_fixits): Likewise.
8748 (layout::annotation_line_showed_range_p): Likewise.
8749 (struct line_corrections): Likewise for field m_row.
8750 (line_corrections::line_corrections): Likewise for param "row".
8751 (layout::print_trailing_fixits): Likewise.
8752 (layout::get_state_at_point): Likewise.
8753 (layout::get_x_bound_for_row): Likewise.
8754 (layout::print_line): Likewise.
8755 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8756 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8757 * input.c (selftest::test_linenum_comparisons): New function.
8758 (selftest::input_c_tests): Call it.
8759 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8760 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8761 * selftest.h (ASSERT_GT): New macro.
8762 (ASSERT_GT_AT): New macro.
8763 (ASSERT_LT): New macro.
8764 (ASSERT_LT_AT): New macro.
8766 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8768 PR rtl-optimization/84780
8769 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8771 2018-03-14 Martin Liska <mliska@suse.cz>
8773 * tree.c (record_node_allocation_statistics): Use
8774 get_stats_node_kind.
8775 (get_stats_node_kind): New function extracted from
8776 record_node_allocation_statistics.
8777 (free_node): Use get_stats_node_kind.
8779 2018-03-14 Richard Biener <rguenther@suse.de>
8781 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8782 that the value-set of ANTIC_IN doesn't grow.
8785 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8787 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8788 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8790 2018-03-14 Julia Koval <julia.koval@intel.com>
8792 * config.gcc (icelake-client, icelake-server): New.
8794 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8795 (initial_ix86_arch_features): Ditto.
8796 (PTA_SKYLAKE): Add SGX.
8797 (PTA_ICELAKE): Remove.
8798 (PTA_ICELAKE_CLIENT): New.
8799 (PTA_ICELAKE_SERVER): New.
8800 (ix86_option_override_internal): Split up icelake on icelake client and
8802 (get_builtin_code_for_version): Ditto.
8803 (fold_builtin_cpu): Ditto.
8804 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8805 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8806 * config/i386/i386.h (processor_type): Ditto.
8807 * doc/invoke.texi: Ditto.
8809 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8812 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8813 INTEGER_CST offset, add it together with bitpos / 8 and
8814 sign extend based on POINTER_SIZE.
8818 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8821 * config/i386/constraints.md (Yc): New register constraint.
8822 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8823 Use Yc constraint for alternative 2 of operand 0. Remove
8824 preferred_for_speed attribute.
8826 2018-03-14 Richard Biener <rguenther@suse.de>
8828 PR tree-optimization/84830
8829 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8830 with the old one to avoid oscillations.
8832 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8835 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8837 (assign_by_spills): Return a flag of reload assignment failure.
8838 Do not process the reload assignment failures. Do not spill other
8839 reload pseudos if they has the same reg class. Update n if
8841 (lra_assign): Add a return arg. Set up from the result of
8842 assign_by_spills call.
8843 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8844 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8845 usage_insns if it is not NULL.
8846 (spill_hard_reg_in_range): New function.
8847 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8848 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8849 function prototypes.
8850 (lra_assign): Change prototype.
8851 * lra.c (lra): Add code to deal with fails by splitting hard reg
8854 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8856 * config/riscv/riscv.opt (mrelax): New option.
8857 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8858 "norelax" when riscv_mrelax is disabled.
8859 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8861 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8864 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8865 reassociation for int modes.
8867 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8869 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8870 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8872 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8873 * config/aarch64/aarch64-sve.md
8874 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8875 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8876 (*extend<mode><Vwide>2): Rename to...
8877 (aarch64_sve_extend<mode><Vwide>2): ...this.
8878 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8879 renaming the old pattern to...
8880 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8882 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8883 define_expand, renaming the old pattern to...
8884 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8885 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8886 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8887 account when deciding which SVE instruction the optab should use.
8888 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8890 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8892 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8893 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8894 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8895 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8896 (tlsdesc_small_<mode>): Turn a define_expand and use
8897 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8898 (tlsdesc_small_advsimd_<mode>): ...this.
8899 (tlsdesc_small_sve_<mode>): New pattern.
8901 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8903 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8904 (UNSPEC_UMUL_HIGHPART): New constants.
8905 (MUL_HIGHPART): New int iteraor.
8906 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8907 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8909 (*<su>mul<mode>3_highpart): New define_insn.
8911 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8914 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8917 2018-03-13 Martin Liska <mliska@suse.cz>
8920 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8922 (sem_item_optimizer::~sem_item_optimizer): Release it.
8923 (sem_item_optimizer::merge_classes): Register variable aliases.
8924 (sem_item_optimizer::fixup_pt_set): New function.
8925 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8926 * ipa-icf.h: Declare new variables and functions.
8928 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8931 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8932 integer_pow2p@2 and test integer_pow2p in condition.
8933 (A < 0 ? C : 0): Similarly for @1.
8936 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8937 characters starting at p contain '\0' character, don't look beyond
8941 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8942 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8945 * reg-stack.c (change_stack): Change update_end var from int to
8946 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8947 also call set_block_for_insn on the newly added insns and rescan.
8950 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8951 on the last operand.
8954 * tree.c (stabilize_reference_1): Return save_expr (e) for
8955 STATEMENT_LIST even if it doesn't have side-effects.
8957 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8959 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8961 2018-03-12 Renlin Li <renlin.li@arm.com>
8963 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8964 aarch64_output_scalar_simd_mov_immediate.
8966 2018-03-12 Martin Sebor <msebor@redhat.com>
8968 PR tree-optimization/83456
8969 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8970 for perfectly overlapping calls to memcpy.
8971 (gimple_fold_builtin_memory_chk): Same.
8972 (gimple_fold_builtin_strcpy): Handle no-warning.
8973 (gimple_fold_builtin_stxcpy_chk): Same.
8974 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8976 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8978 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8979 parameter. Use it for SFmode.
8980 (rs6000_function_arg_advance_1): Adjust.
8981 (rs6000_function_arg): Adjust.
8982 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8984 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8986 PR rtl-optimization/84169
8987 PR rtl-optimization/84780
8988 * combine.c (can_combine_p): Check for a 2-insn combination whether
8989 the destination register is used between the two insns, too.
8991 2018-03-12 Richard Biener <rguenther@suse.de>
8993 PR tree-optimization/84803
8994 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8995 for refs DR analysis didn't process.
8997 2018-03-12 Richard Biener <rguenther@suse.de>
8999 PR tree-optimization/84777
9000 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
9001 force-vectorize loops ignore whether we are optimizing for size.
9003 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
9005 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
9006 (TARGET_MD_ASM_ADJUST): Define.
9008 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
9009 Kito Cheng <kito.cheng@gmail.com>
9010 Chung-Ju Wu <jasonwucj@gmail.com>
9012 * config/nds32/nds32.c (nds32_compute_stack_frame,
9013 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
9014 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
9015 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
9016 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
9017 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
9018 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
9019 * config/nds32/nds32.md (prologue, epilogue): Use macro
9020 NDS32_V3PUSH_AVAILABLE_P to do checking.
9022 2018-03-11 Jakub Jelinek <jakub@redhat.com>
9025 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
9026 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
9027 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
9028 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
9029 addition of most attributes on !orig_type_die or the attribute not
9030 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
9032 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9033 Chung-Ju Wu <jasonwucj@gmail.com>
9035 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
9037 * config/nds32/nds32.opt (mvh): New option.
9039 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9040 Chung-Ju Wu <jasonwucj@gmail.com>
9042 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
9044 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
9045 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
9048 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
9049 Chung-Ju Wu <jasonwucj@gmail.com>
9051 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
9053 * config/nds32/nds32-multiple.md (strlensi): New pattern.
9054 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
9056 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
9057 Kito Cheng <kito.cheng@gmail.com>
9058 Chung-Ju Wu <jasonwucj@gmail.com>
9060 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
9061 UNSPEC_FFMISM and UNSPEC_FLMISM.
9062 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
9063 for ffb, ffmism and flmism.
9064 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
9065 (unspec_ffmism): Ditto.
9066 (unspec_flmism): Ditto.
9067 (nds32_expand_builtin_impl): Check if string extension is available.
9068 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
9069 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
9071 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
9074 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9077 * lra-assigns.c (assign_by_spills): Return a flag of reload
9078 assignment failure. Do not process the reload assignment
9079 failures. Do not spill other reload pseudos if they has the same
9081 (lra_assign): Add a return arg. Set up from the result of
9082 assign_by_spills call.
9083 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9084 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9085 usage_insns if it is not NULL.
9086 (spill_hard_reg_in_range): New function.
9087 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9088 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9089 function prototypes.
9090 (lra_assign): Change prototype.
9091 * lra.c (lra): Add code to deal with fails by splitting hard reg
9094 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
9097 * config/i386/i386.opt: Replace Enforcment with Enforcement.
9099 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
9102 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
9103 (dw_val_node): Add val_symbolic_view.
9104 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
9105 (symview_upper_bound): New.
9106 (new_line_info_table): Initialize symviews_since_reset.
9107 (dwarf2out_source_line): Count symviews_since_reset and set
9108 symview_upper_bound.
9109 (dw_val_equal_p): Handle symview.
9110 (add_AT_symview): New.
9111 (print_dw_val): Handle symview.
9112 (attr_checksum, attr_checksum_ordered): Likewise.
9113 (same_dw_val_p, size_of_die): Likewise.
9114 (value_format, output_die): Likewise.
9115 (add_high_low_attributes): Use add_AT_symview for entry_view.
9116 (dwarf2out_finish): Reset symview_upper_bound, clear
9119 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
9122 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
9123 Add strict argument and use it.
9124 (rs6000_split_multireg_move): Update for new strict argument.
9125 (mem_operand_gpr): Disallow all non-offsettable addresses.
9126 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9128 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9131 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9132 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9133 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9136 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9137 decl, use remap_type if we want to use the type.
9139 2018-03-09 Martin Sebor <msebor@redhat.com>
9141 PR tree-optimization/84526
9142 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9144 (builtin_access::generic_overlap): Be prepared to handle non-array
9147 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9149 PR rtl-optimization/84682
9150 * lra-constraints.c (process_address_1): Check is_address flag
9151 for address constraints.
9152 (process_alt_operands): Likewise.
9153 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9154 preprocess_constraints.
9155 * recog.h (preprocess_constraints): Add oploc parameter.
9157 * recog.c (preprocess_constraints): Test address_operand for
9158 CT_ADDRESS constraints.
9160 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9163 * lra-assigns.c (assign_by_spills): Return a flag of reload
9164 assignment failure. Do not process the reload assignment
9165 failures. Do not spill other reload pseudos if they has the same
9167 (lra_assign): Add a return arg. Set up from the result of
9168 assign_by_spills call.
9169 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9170 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9171 usage_insns if it is not NULL.
9172 (spill_hard_reg_in_range): New function.
9173 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9174 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9175 function prototypes.
9176 (lra_assign): Change prototype.
9177 * lra.c (lra): Add code to deal with fails by splitting hard reg
9180 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9183 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9184 Accept complain bool parameter. Only emit errors if it is true.
9185 (arm_parse_cpu_option_name): Likewise.
9186 (arm_target_thumb_only): Adjust callers of the above.
9187 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9188 prototype to take a default true bool parameter.
9189 (arm_parse_arch_option_name): Likewise.
9191 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9192 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9196 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9197 * configure: Regenerate.
9198 * configure.ac ("linker --version-script option"): New.
9199 ("linker soname option"): New.
9201 2018-03-09 Richard Biener <rguenther@suse.de>
9203 PR tree-optimization/84775
9204 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9205 immediate uses of predicate stmts and mark them modified.
9208 PR tree-optimization/84178
9209 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9211 (version_loop_for_if_conversion): Delay update_ssa call.
9212 (tree_if_conversion): Delay update_ssa until after predicate
9215 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9218 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9219 when the function accesses prior frames.
9221 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9224 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9225 gen_llsym, otherwise call maybe_gen_llsym.
9228 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9229 has ',' character inside of it.
9231 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9234 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9235 as clobbering CC_REGNUM.
9237 2018-03-08 Richard Biener <rguenther@suse.de>
9240 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9241 (follow_copies_to_constant): Do not follow SSA names registered
9244 2018-03-08 Richard Biener <rguenther@suse.de>
9246 PR tree-optimization/84178
9247 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9249 (version_loop_for_if_conversion): Delay update_ssa call.
9250 (tree_if_conversion): Delay update_ssa until after predicate
9253 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9255 PR tree-optimization/84178
9256 * tree-if-conv.c (release_bb_predicate): Remove the
9257 the assertion that the stmts have NULL use_ops.
9258 Discard the statements, asserting that they haven't
9259 yet been added to a BB.
9261 2018-03-08 Richard Biener <rguenther@suse.de>
9263 PR tree-optimization/84746
9264 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9265 (phi_translate): Pass in destination ANTIC_OUT set.
9266 (phi_translate_1): Likewise. For a simplified result lookup
9267 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9268 (phi_translate_set): Adjust.
9269 (do_pre_regular_insertion): Likewise.
9270 (do_pre_partial_partial_insertion): Likewise.
9272 2018-03-08 Martin Liska <mliska@suse.cz>
9274 PR gcov-profile/84735
9275 * doc/gcov.texi: Document usage of profile files.
9276 * gcov-io.h: Document changes in the format.
9278 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9282 * dwarf2out.c (struct dw_line_info_table): Update comments for
9284 (FORCE_RESET_NEXT_VIEW): New.
9285 (FORCE_RESETTING_VIEW_P): New.
9286 (RESETTING_VIEW_P): Check for -1 too.
9287 (ZERO_VIEW_P): Likewise.
9288 (new_line_info_table): Force-reset next view.
9289 (dwarf2out_begin_function): Likewise.
9290 (dwarf2out_source_line): Simplify zero_view_p initialization.
9291 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9292 view directly. Omit view when omitting .loc at line 0.
9294 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9296 PR tree-optimization/84740
9297 * tree-switch-conversion.c (process_switch): Call build_constructors
9298 only if info.phi_count is non-zero.
9300 PR tree-optimization/84739
9301 * tree-tailcall.c (find_tail_calls): Check call arguments against
9302 DECL_ARGUMENTS (current_function_decl) rather than
9303 DECL_ARGUMENTS (func) when checking for tail recursion.
9305 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9307 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9308 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9309 Volker Reichelt's entry and add entries for people that perform
9310 GCC fuzzy testing and report numerous bugs.
9312 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9315 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9316 readonly data in sdata, if that is disabled.
9317 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9318 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9319 -mreadonly-in-sdata option.
9321 2018-03-07 Martin Sebor <msebor@redhat.com>
9323 PR tree-optimization/84468
9324 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9325 basic block when looking for nul assignment.
9327 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9330 * except.h (output_function_exception_table): Adjust prototype.
9331 * except.c (output_function_exception_table): Remove FNNAME parameter
9332 and add SECTION parameter. Ouput one part of the table at a time.
9333 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9334 the first part of the exception table and emit unwind directives.
9335 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9336 (i386_pe_seh_cold_init): Likewise.
9337 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9338 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9339 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9340 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9341 * config/i386/winnt.c: Include except.h.
9342 (struct seh_frame_state): Add reg_offset, after_prologue and
9343 in_cold_section fields.
9344 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9345 (i386_pe_seh_cold_init): New function.
9346 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9347 to seh->in_cold_section.
9348 (seh_emit_push): Record the offset of the push.
9349 (seh_emit_save): Record the offet of the save.
9350 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9351 Test seh->after_prologue to disregard the epilogue.
9352 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9353 (i386_pe_end_cold_function): New function.
9355 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9358 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9359 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9362 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9364 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9365 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9368 * multiple_target.c: Include tree-inline.h and intl.h.
9369 (expand_target_clones): Diagnose and fail if node->definition and
9370 !tree_versionable_function_p (node->decl).
9372 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9374 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9376 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9377 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9378 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9380 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9383 * combine.c (try_combine): Use reg_or_subregno instead of handling
9384 just paradoxical SUBREGs and REGs.
9386 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9388 * config/arc/arc.c (arc_finalize_pic): Remove function.
9389 (arc_must_save_register): We use single base PIC register, remove
9390 checks to save/restore the PIC register.
9391 (arc_expand_prologue): Likewise.
9392 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9394 (arc_verify_short): Likewise.
9395 (arc_attr_type): Likewise.
9396 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9397 (walk_stores): Likewise.
9398 (arc_address_cost): Make it static.
9399 (arc_verify_short): Likewise.
9400 (branch_dest): Likewise.
9401 (arc_attr_type): Likewise.
9402 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9403 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9404 (arc_final_prescan_insn): Remove inserting the nops due to
9405 hardware hazards. It is done in reorg step.
9406 (insn_length_variant_t): Remove.
9407 (insn_length_parameters_t): Likewise.
9408 (arc_insn_length_parameters): Likewise.
9409 (arc_get_insn_variants): Likewise.
9410 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9412 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9415 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9418 PR tree-optimization/84687
9419 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9421 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9423 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9425 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9426 Rename to ppc_speculation_barrier.
9427 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9428 __builtin_ppc_speculation_barrier.
9430 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9433 * combine.c (combine_simplify_rtx): Don't try to simplify if
9434 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9437 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9439 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9440 to 32 bytes when compiling for POWER9.
9442 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9445 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9446 regparm >= 3 with no arg reg available also for calls with
9447 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9450 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9452 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9454 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9457 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9459 2018-03-05 Richard Biener <rguenther@suse.de>
9461 PR tree-optimization/84486
9462 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9463 When inserting a __builtin_assume_aligned call set the LHS
9464 SSA name alignment info accordingly.
9466 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9468 PR tree-optimization/84114
9469 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9470 Avoid reassociation of FLOAT_MODE addition.
9472 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9474 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9475 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9476 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9477 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9478 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9479 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9480 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9482 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9483 __builtin_ia32_wbinvd): New builtins.
9484 (SPECIAL_ARGS2): New.
9485 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9486 (SPECIAL_ARGS2): New.
9487 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9488 (ix86_valid_target_attribute_inner_p): Ditto.
9489 (ix86_init_mmx_sse_builtins): Add special_args2.
9490 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9491 TARGET_WBNOINVD_P): New.
9492 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9493 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9494 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9495 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9496 * config/i386/pconfigintrin.h: New file.
9497 * config/i386/wbnoinvdintrin.h: Ditto.
9498 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9500 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9502 2018-03-05 Richard Biener <rguenther@suse.de>
9504 PR tree-optimization/84670
9505 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9507 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9508 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9509 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9510 doesn't grow if all successors (recursively) were visited at least
9513 2018-03-05 Richard Biener <rguenther@suse.de>
9515 PR tree-optimization/84650
9516 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9517 if executed in the loop pipeline.
9519 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9521 * doc/configfiles.texi (Configuration Files): Move info about
9522 conditionalizing $target-protos.h to...
9523 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9524 differs from $target-protos.h.
9526 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9527 Chung-Ju Wu <jasonwucj@gmail.com>
9529 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9530 * config/nds32/nds32-multiple.md (setmemsi): Define.
9531 * config/nds32/nds32-memory-manipulation.c
9532 (nds32_gen_dup_4_byte_to_word_value): New.
9533 (emit_setmem_word_loop): New.
9534 (emit_setmem_byte_loop): New.
9535 (nds32_expand_setmem_loop): New.
9536 (nds32_expand_setmem_loop_v3m): New.
9537 (nds32_expand_setmem_unroll): New.
9538 (nds32_expand_setmem): New.
9540 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9541 Chung-Ju Wu <jasonwucj@gmail.com>
9543 * config/nds32/nds32-memory-manipulation.c
9544 (nds32_emit_load_store): New.
9545 (nds32_emit_post_inc_load_store): New.
9546 (nds32_emit_mem_move): New.
9547 (nds32_emit_mem_move_block): New.
9548 (nds32_expand_movmemsi_loop_unknown_size): New.
9549 (nds32_expand_movmemsi_loop_known_size): New.
9550 (nds32_expand_movmemsi_loop): New.
9551 (nds32_expand_movmemsi_unroll): New.
9552 (nds32_expand_movmemqi): Rename ...
9553 (nds32_expand_movmemsi): ... to this.
9554 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9555 (movmemsi): ... to this.
9556 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9557 (nds32_expand_movmemsi): ... to this.
9559 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9560 Monk Chiang <sh.chiang04@gmail.com>
9561 Chung-Ju Wu <jasonwucj@gmail.com>
9563 * config/nds32/nds32-protos.h
9564 (nds32_expand_load_multiple): New arguments.
9565 (nds32_expand_store_multiple): Ditto.
9566 (nds32_valid_multiple_load_store): Rename ...
9567 (nds32_valid_multiple_load_store_p): ... to this.
9568 * config/nds32/nds32-memory-manipulation.c
9569 (nds32_expand_load_multiple): Refine implementation.
9570 (nds32_expand_store_multiple): Ditto.
9571 * config/nds32/nds32-multiple.md
9572 (load_multiple): Update nds32_expand_load_multiple interface.
9573 (store_multiple): Update nds32_expand_store_multiple interface.
9574 * config/nds32/nds32-predicates.c
9575 (nds32_valid_multiple_load_store): Rename ...
9576 (nds32_valid_multiple_load_store_p): ... to this and refine
9578 * config/nds32/predicates.md
9579 (nds32_load_multiple_and_update_address_operation): New predicate.
9580 (nds32_store_multiple_and_update_address_operation): New predicate.
9582 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9583 Chung-Ju Wu <jasonwucj@gmail.com>
9585 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9586 (combo): New attribute.
9587 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9589 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9591 * config/nds32/nds32.opt: Change -mcmodel= default value.
9593 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9594 Monk Chiang <sh.chiang04@gmail.com>
9595 Chung-Ju Wu <jasonwucj@gmail.com>
9597 * config/nds32/constants.md (unspec_element): New enum.
9598 * config/nds32/constraints.md (Umw): New constraint.
9599 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9600 * config/nds32/nds32-intrinsic.md: Likewise.
9601 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9602 (nds32_valid_smw_lwm_base_p): New.
9603 (nds32_output_smw_single_word): New.
9604 (nds32_output_lmw_single_word): New.
9605 (nds32_expand_unaligned_load): New.
9606 (nds32_expand_unaligned_store): New.
9607 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9608 (nds32_output_smw_single_word): Declare.
9609 (nds32_output_lmw_single_word): Declare.
9610 (nds32_expand_unaligned_load): Declare.
9611 (nds32_expand_unaligned_store): Declare.
9612 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9613 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9614 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9615 NDS32_BUILTIN_UASTORE_DW.
9616 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9619 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9620 Kito Cheng <kito.cheng@gmail.com>
9621 Chung-Ju Wu <jasonwucj@gmail.com>
9623 * config/nds32/nds32-intrinsic.c
9624 (nds32_expand_builtin_null_ftype_reg): Delete.
9625 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9626 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9627 (nds32_read_argument): New.
9628 (nds32_legitimize_target): Ditto.
9629 (nds32_legitimize_argument): Ditto.
9630 (nds32_check_constant_argument): Ditto.
9631 (nds32_expand_unop_builtin): Ditto.
9632 (nds32_expand_unopimm_builtin): Ditto.
9633 (nds32_expand_binop_builtin): Ditto.
9634 (nds32_builtin_decl_impl): Ditto.
9635 (builtin_description): Ditto.
9636 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9637 (nds32_init_builtins_impl): Ditto.
9638 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9639 (nds32_builtin_decl): New.
9640 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9641 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9643 2018-03-02 Jeff Law <law@redhat.com>
9645 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9646 (redundant_insn, fill_simple_delay_slots): Likewise.
9647 (fill_slots_from_thread): Likewise.
9648 * resource.c (mark_referenced_resources): Likewise.
9649 (mark_set_resources, find_dead_or_set_registers): Likewise.
9651 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9653 * substring-locations.h (format_warning_va): Formatting fix for
9655 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9657 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9658 * substring-locations.c: Include intl.h.
9659 (format_warning_va): Turned into small wrapper around
9660 format_warning_n_va, renamed to ...
9661 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9662 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9664 (format_warning_at_substring_n): New function.
9665 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9666 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9667 format_warning_at_substring with just a shorter name instead of
9668 const function pointer.
9669 (fmtwarn_n): New function.
9670 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9671 appropriate, get rid of all the fmtstr temporaries, move conditionals
9672 with G_() wrapped string literals directly into fmtwarn arguments,
9673 cast dir.len to (int), formatting fixes.
9675 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9677 * doc/invoke.texi: Remove "Cilk Plus" references.
9679 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9680 Richard Biener <rguenther@suse.de>
9683 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9684 for error or warning attributes if CALL_FROM_THUNK_P is set.
9687 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9690 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9691 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9694 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9695 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9697 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9698 instead of -1U in last predictors element's probability member.
9700 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9703 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9704 arguments if they are comparable.
9706 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9708 PR tree-optimization/84634
9709 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9710 masks and masked_loop_p with a single loop_masks, making sure it's
9711 null for bb vectorization.
9713 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9715 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9716 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9717 loop->force_vectorize to check whether there is no alias.
9719 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9722 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9724 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9726 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9727 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9728 instead of a loop around prev_real_insn.
9729 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9733 * config/i386/i386.c (ix86_print_operand): Use conditional
9734 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9737 2018-03-02 Richard Biener <rguenther@suse.de>
9739 PR tree-optimization/84427
9740 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9741 (bitmap_set_subtract_values): Rewrite to handle multiple
9744 (prune_clobbered_mems): Likewise.
9745 (phi_translate): Take edge instead of pred/phiblock.
9746 (phi_translate_1): Likewise.
9747 (phi_translate_set): Likewise. Insert all translated
9748 exprs for a value into the set, keeping possibly multiple
9749 expressions per value.
9750 (compute_antic_aux): Adjust for phi_translate changes.
9751 When intersecting union the expressions and prune those
9752 not in the final value set, keeping possibly multiple
9753 expressions per value. Do not use value-insertion
9754 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9755 all expressions. Add verification that the value-sets
9756 only shrink during iteration.
9757 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9758 (do_pre_regular_insertion): Likewise.
9759 (do_pre_partial_partial_insertion): Likewise.
9761 2018-03-02 Richard Biener <rguenther@suse.de>
9764 * config/darwin.c (saved_debug_info_level): New static global.
9765 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9766 (darwin_asm_lto_end): Restore debug info generation settings.
9768 2018-03-01 Martin Liska <mliska@suse.cz>
9771 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9774 2018-03-01 Richard Biener <rguenther@suse.de>
9777 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9778 type annotation with LTO when debug was disabled at compile-time.
9780 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9782 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9784 (mips_final_postscan_insn): Likewise.
9786 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9788 PR rtl-optimization/84528
9789 * alias.c (init_alias_target): Add commentary.
9790 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9791 a unique base value if the frame pointer is not eliminated
9792 to the stack pointer.
9794 2018-03-01 Tom de Vries <tom@codesourcery.com>
9796 PR rtl-optimization/83327
9797 * lra-int.h (hard_regs_spilled_into): Declare.
9798 * lra.c (hard_regs_spilled_into): Define.
9799 (init_reg_info): Init hard_regs_spilled_into.
9800 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9801 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9802 (process_bb_lives): Handle hard_regs_spilled_into.
9803 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9804 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9806 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9808 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9809 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9810 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9811 * config/rs6000/aix72.h: New file.
9813 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9815 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9816 instead of warning_at with conditional singular and plural messages
9820 * stor-layout.c (update_alignment_for_field): For
9821 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9822 && !DECL_PACKED (field), do the alignment update, just use
9823 only desired_align instead of MAX (type_align, desired_align)
9825 (place_field): Don't do known_align < desired_align handling
9826 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9827 is non-NULL, instead do it after rli->prev_field handling and
9828 only if not within a bitfield word. For DECL_PACKED (field)
9829 use type_align of BITS_PER_UNIT.
9831 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9833 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9834 superfluous parentheses and trailing spaces.
9836 2018-02-28 Richard Biener <rguenther@suse.de>
9838 PR tree-optimization/84584
9839 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9840 SCoPs with fake exit edge.
9842 2018-02-28 Martin Liska <mliska@suse.cz>
9845 * timevar.c (timer::print): Fix format to properly print 100%
9848 2018-02-28 Richard Biener <rguenther@suse.de>
9851 * genmatch.c (capture_info::walk_match): Do not mark
9852 captured expressions without operands as expr_p given
9853 they act more like predicates and should be subject to
9854 "lost tail" side-effect preserving.
9856 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9858 PR rtl-optimization/81611
9859 * auto-inc-dec.c (attempt_change): Move dead note from
9860 mem_insn if it's the next use of regno
9861 (find_address): Take address use of reg holding
9862 non-incremented value. Add parm to limit search to the named
9864 (merge_in_block): Attempt to use a mem insn that is the next
9865 use of the original regno.
9867 2018-02-27 Martin Sebor <msebor@redhat.com>
9870 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9871 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9873 2018-02-27 Martin Sebor <msebor@redhat.com>
9875 PR translation/84207
9876 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9877 n argument to unsigned HOST_WIDE_INT.
9878 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9879 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9880 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9881 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9883 2018-02-27 Richard Biener <rguenther@suse.de>
9885 PR tree-optimization/84512
9886 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9887 Do not use the estimate returned from record_stmt_cost for
9888 the scalar iteration cost but sum properly using add_stmt_cost.
9890 2018-02-27 Richard Biener <rguenther@suse.de>
9892 PR tree-optimization/84466
9893 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9894 Adjust last change to less strictly validate use operands.
9896 2018-02-27 Martin Liska <mliska@suse.cz>
9898 PR gcov-profile/84548
9899 * gcov.c (process_file): Allow partial overlap and consider it
9900 also as group functions.
9901 (output_lines): Properly calculate range of lines for a group.
9903 2018-02-27 Martin Liska <mliska@suse.cz>
9905 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9906 'ggc' suffixes. Change first column width.
9907 (timer::print): Fix formatting of the column.
9909 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9911 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9912 preserve inline entry blocks for the sake of debug inline
9913 entry point markers alone.
9914 (remove_unused_locals): Suggest in comments a better place to
9915 force the preservation of inline entry blocks that are
9916 otherwise unused, but do not preserve them.
9918 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9920 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9922 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9925 * config/i386/constraints.md (Bs): Replace
9926 ix86_indirect_branch_register with
9927 TARGET_INDIRECT_BRANCH_REGISTER.
9929 * config/i386/i386.md (indirect_jump): Likewise.
9930 (tablejump): Likewise.
9931 (*sibcall_memory): Likewise.
9932 (*sibcall_value_memory): Likewise.
9933 Peepholes of indirect call and jump via memory: Likewise.
9934 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9935 (*sibcall_value_GOT_32): Likewise.
9936 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9937 (GOT_memory_operand): Likewise.
9938 (call_insn_operand): Likewise.
9939 (sibcall_insn_operand): Likewise.
9940 (GOT32_symbol_operand): Likewise.
9941 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9943 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9945 PR rtl-optimization/83496
9946 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9947 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9948 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9949 redundant insn, if any.
9950 (relax_delay_slots): Likewise.
9951 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9953 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9955 PR tree-optimization/83965
9956 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9957 that grouped statements are part of a reduction chain. Return
9958 true if the statement is not marked as a reduction itself but
9960 (vect_recog_dot_prod_pattern): Don't check whether the statement
9961 is part of a group here.
9962 (vect_recog_sad_pattern): Likewise.
9963 (vect_recog_widen_sum_pattern): Likewise.
9965 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9968 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9970 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9973 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9975 (ix86_output_indirect_function_return): New prototype.
9976 (ix86_split_simple_return_pop_internal): Likewise.
9977 * config/i386/i386.c (indirect_return_via_cx): New.
9978 (indirect_return_via_cx_bnd): Likewise.
9979 (indirect_thunk_name): Handle return va CX_REG.
9980 (output_indirect_thunk_function): Create alias for
9981 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9982 (ix86_output_indirect_jmp): Remove the bool argument.
9983 (ix86_output_indirect_function_return): New function.
9984 (ix86_split_simple_return_pop_internal): Likewise.
9985 * config/i386/i386.md (*indirect_jump): Don't pass false
9986 to ix86_output_indirect_jmp.
9987 (*tablejump_1): Likewise.
9988 (simple_return_pop_internal): Change it to define_insn_and_split.
9989 Call ix86_split_simple_return_pop_internal to split it for
9991 (simple_return_indirect_internal): Call
9992 ix86_output_indirect_function_return instead of
9993 ix86_output_indirect_jmp.
9995 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9998 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9999 memset and value initialization afterwards.
10001 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
10003 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
10005 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10008 * common/config/aarch64/aarch64-common.c
10009 (aarch_option_optimization_table[]): Switch
10010 off fomit-frame-pointer
10012 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10013 Chung-Ju Wu <jasonwucj@gmail.com>
10015 * config/nds32/nds32-multiple.md (load_multiple): Disallow
10017 (store_multiple): Ditto.
10019 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
10021 * config.gcc: Add --with-cpu support for nds32 target.
10022 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
10023 * config/nds32/nds32.opt: Add -mcpu= option.
10025 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
10027 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
10028 isel=yes): Warn for these deprecated options.
10030 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
10032 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
10033 ISA_2_5_MASKS_EMBEDDED.
10035 2018-02-23 Jakub Jelinek <jakub@redhat.com>
10037 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
10038 p->max as pointers rather than using iterative_hash_expr.
10040 2018-02-23 Carl Love <cel@us.ibm.com>
10042 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
10043 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
10045 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
10046 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
10047 P8V_BUILTIN_VEC_VUNSIGNED2.
10049 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
10052 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
10053 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
10055 * lra-constraints.c (curr_insn_transform): Set up
10056 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
10058 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
10059 * lra-eliminations.c (spill_pseudos): Ditto.
10060 (process_insn_for_elimination): Ditto.
10061 * lra-lives.c (reg_early_clobber_p): Use the new macros.
10062 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
10063 LRA_NON_CLOBBERED_ALT.
10065 2018-02-22 Martin Sebor <msebor@redhat.com>
10067 PR tree-optimization/84480
10068 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
10069 to maybe_diag_stxncpy_trunc. Call it.
10070 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
10071 from gimple_fold_builtin_strcpy. Print inlining stack.
10072 (handle_builtin_stxncpy): Print inlining stack.
10073 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
10075 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
10078 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
10079 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
10080 and -fcheck-pointer-bounds are used together.
10081 (indirect_thunk_prefix): New enum.
10082 (indirect_thunk_need_prefix): New function.
10083 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
10084 "_nt" instead of "_bnd" for NOTRACK prefix.
10085 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
10086 (output_indirect_thunk_function): Likewise.
10088 (ix86_code_end): Update output_indirect_thunk_function calls.
10089 (ix86_output_indirect_branch_via_reg): Replace
10090 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
10091 (ix86_output_indirect_branch_via_push): Likewise.
10092 (ix86_output_function_return): Likewise.
10093 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
10094 incompatible with -fcf-protection=branch and
10095 -fcheck-pointer-bounds.
10097 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10100 * config/aarch64/aarch64.c (aarch64_print_address_internal):
10101 Change gcc_assert call to output_operand_lossage.
10103 2018-02-22 Steve Ellcey <sellcey@cavium.com>
10105 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
10107 2018-02-22 DJ Delorie <dj@redhat.com>
10108 Sebastian Perta <sebastian.perta@renesas.com>
10109 Oleg Endo <olegendo@gcc.gnu.org>
10111 * config/rx/rx.c (rx_rtx_costs): New function.
10112 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
10114 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
10116 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
10118 2018-02-22 Martin Liska <mliska@suse.cz>
10121 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
10122 Add "native" as a possible value.
10124 2018-02-22 Martin Liska <mliska@suse.cz>
10127 * config/i386/i386.c (ix86_option_override_internal):
10128 Add "native" as a possible value for -march and -mtune.
10130 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10133 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10134 to all type variants.
10136 PR tree-optimization/84503
10137 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10138 width as info->bitpos + info->bitsize - start.
10139 (merged_store_group::merge_overlapping): Simplify width computation.
10140 (check_no_overlap): New function.
10141 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10142 start + width and last_order of the group, fail if check_no_overlap
10144 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10145 to group if check_no_overlap fails.
10147 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10149 * config/rs6000/altivec.md: Delete contraint arguments to
10150 define_expand, define_split, and define_peephole2, and in
10151 define_insn_and_split if always unused.
10152 * config/rs6000/darwin.md: Ditto.
10153 * config/rs6000/dfp.md: Ditto.
10154 * config/rs6000/rs6000.md: Ditto.
10155 * config/rs6000/sync.md: Ditto.
10156 * config/rs6000/vector.md: Ditto.
10157 * config/rs6000/vsx.md: Ditto.
10159 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10161 * config/rs6000/altivec.md: Write output control strings as braced
10162 blocks instead of double-quoted strings.
10163 * config/rs6000/darwin.md: Ditto.
10164 * config/rs6000/rs6000.md: Ditto.
10165 * config/rs6000/vector.md: Ditto.
10166 * config/rs6000/vsx.md: Ditto.
10168 2018-02-21 Jason Merrill <jason@redhat.com>
10170 PR c++/84314 - ICE with templates and fastcall attribute.
10171 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10173 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10175 * ipa-cp.c (determine_versionability): Fix comment typos.
10177 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10180 * ipa-cp.c (determine_versionability): Do not version functions caling
10183 2018-02-21 Martin Liska <mliska@suse.cz>
10186 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10187 Add "native" as a possible value.
10188 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10189 the macro when native cpu detection is available.
10191 2018-02-21 Martin Liska <mliska@suse.cz>
10194 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10195 Add "native" as a possible value.
10196 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10197 when native cpu detection is available.
10199 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10200 Martin Sebor <msebor@redhat.com>
10202 PR tree-optimization/84478
10203 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10205 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10206 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10207 support which is conservatively correct, for 2 only stay conservative
10208 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10209 argument to the 2 argument get_range_strlen, adjust 6 arg
10210 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10212 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10213 (gimple_fold_builtin_strlen): Pass true as last argument to
10216 2018-02-20 Martin Sebor <msebor@redhat.com>
10218 PR middle-end/84095
10219 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10220 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10221 (builtin_memref::builtin_memref): Factor out parts into
10222 set_base_and_offset and call it.
10224 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10226 PR middle-end/84406
10227 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10228 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10229 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10230 search at the associated MODE_INT.
10232 2018-02-20 Jeff Law <law@redhat.com>
10234 PR middle-end/82123
10235 PR tree-optimization/81592
10236 PR middle-end/79257
10237 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10238 for range data rather than using global data.
10239 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10240 range data rather than using global data.
10241 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10242 pass it to children as needed.
10243 (struct directive::fmtresult): Similarly.
10244 (struct directive::set_width): Similarly.
10245 (struct directive::set_precision): Similarly.
10246 (format_integer, format_directive, parse_directive): Similarly.
10247 (format_none): Accept unnamed vr_values parameter.
10248 (format_percent, format_floating, format_character): Similarly.
10249 (format_string, format_plain): Similarly.
10250 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10251 the EVRP range analyzer for range data rather than using global data.
10252 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10253 gimple-ssa-evrp-analyze.h
10254 (class sprintf_dom_walker): Add after_dom_children member function.
10255 Add evrp_range_analyzer member.
10256 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10257 range analyzer as needed.
10258 (sprintf_dom_walker::after_dom_children): New member function.
10259 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10261 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10262 (evrp_range_analyzer::pop_to_marker): Likewise.
10264 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10266 PR tree-optimization/84419
10267 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10268 with the required type if its current type is compatible but
10271 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10273 PR middle-end/82004
10274 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10275 after vectorization.
10277 2018-02-20 Martin Liska <mliska@suse.cz>
10280 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10281 possible values if we don't have a hint.
10283 2018-02-20 Martin Liska <mliska@suse.cz>
10287 * final.c (shorten_branches): Build align_tab array with one
10289 * opts.c (finish_options): Add alignment option limit check.
10290 (MAX_CODE_ALIGN): Likewise.
10291 (MAX_CODE_ALIGN_VALUE): Likewise.
10292 * doc/invoke.texi: Document maximum allowed option value for
10293 all -falign-* options.
10295 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10298 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10299 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10300 * var-tracking.c (emit_note_insn_var_location): Remove all references
10301 to NOTE_INSN_CALL_ARG_LOCATION.
10302 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10303 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10304 Use copy_rtx_if_shared.
10305 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10306 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10307 (dwarf2out_var_location): Remove handling of
10308 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10310 * final.c (final_scan_insn): Remove all references to
10311 NOTE_INSN_CALL_ARG_LOCATION.
10312 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10313 before dumping final insns.
10314 * except.c (emit_note_eh_region_end): Remove all references to
10315 NOTE_INSN_CALL_ARG_LOCATION.
10316 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10317 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10318 * config/arc/arc.c (hwloop_optimize): Likewise.
10319 * config/arm/arm.c (create_fix_barrier): Likewise.
10320 * config/s390/s390.c (s390_chunkify_start): Likewise.
10321 * config/sh/sh.c (find_barrier): Likewise.
10322 * config/i386/i386.c (rest_of_insert_endbranch,
10323 ix86_seh_fixup_eh_fallthru): Likewise.
10324 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10325 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10326 * config/frv/frv.c (frv_function_prologue): Likewise.
10327 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10329 (note_outside_basic_block_p): Remove all references to
10330 NOTE_INSN_CALL_ARG_LOCATION.
10331 * gengtype.c (adjust_field_rtx_def): Likewise.
10332 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10334 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10335 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10338 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10341 PR tree-optimization/84452
10342 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10343 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10346 2018-02-19 Martin Liska <mliska@suse.cz>
10349 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10351 2018-02-19 Martin Liska <mliska@suse.cz>
10352 Richard Sandiford <richard.sandiford@linaro.org>
10354 PR tree-optimization/82491
10355 * gimple-fold.c (get_base_constructor): Make earlier bail out
10358 2018-02-19 Carl Love <cel@us.ibm.com>
10360 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10361 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10363 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10364 P8V_BUILTIN_VEC_NEG.
10366 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10368 * config/rl78/rl78.md (movdf): New define expand.
10370 2018-02-19 Martin Liska <mliska@suse.cz>
10373 * doc/invoke.texi: Fix typo.
10374 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10376 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10378 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10379 handle rs6000_single_float and rs6000_double_float specially for
10382 2018-02-16 Jeff Law <law@redhat.com>
10384 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10385 (pop_regs): Use it.
10387 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10390 * ipa-inline.c (inline_small_functions): Fix a typo.
10392 2018-02-16 Nathan Sidwell <nathan@acm.org>
10394 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10396 2018-02-16 Carl Love <cel@us.ibm.com>
10398 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10399 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10400 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10401 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10402 expansion to P8V_BUILTIN_VEC_FLOAT2.
10404 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10406 PR rtl-optimization/70023
10407 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10408 src_regno into account.
10410 2018-02-16 Carl Love <cel@us.ibm.com>
10412 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10413 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10414 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10415 * config/rs6000/rs6000.c: Remove case statements for
10416 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10417 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10418 and P9V_BUILTIN_VEC_VINSERT4B.
10419 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10420 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10421 * config/rs6000/vsx.md:
10422 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10425 2018-02-16 Carl Love <cel@us.ibm.com>
10427 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10429 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10431 * config/rs6000/rs6000-c.c: Add the definitions for
10432 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10433 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10434 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10435 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10436 definition for insert4b and define insn *insert3b_internal.
10437 * doc/extend.texi: Add documentation for vec_extract4b.
10439 2018-02-16 Nathan Sidwell <nathan@acm.org>
10441 * doc/extend.texi (Backwards Compatibility): Mention friend
10442 injection. Note for-scope is deprecated.
10443 * doc/invoke.texi (-ffriend-injection): Deprecate.
10445 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10447 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10448 that moved to I2, also allow destinations that are a paradoxical
10449 subreg (instead of a normal reg).
10451 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10454 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10457 2018-02-16 Richard Biener <rguenther@suse.de>
10459 PR tree-optimization/84037
10460 PR tree-optimization/84016
10462 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10463 Adjust vec_construct for the fact we need additional higher latency
10464 128bit inserts for AVX256 and AVX512 vector builds.
10465 (ix86_add_stmt_cost): Scale vector construction cost for
10468 2018-02-16 Richard Biener <rguenther@suse.de>
10470 PR tree-optimization/84417
10471 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10472 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10473 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10475 2018-02-16 Martin Liska <mliska@suse.cz>
10478 * internal-fn.def (ASAN_CHECK): Set proper flags.
10479 (ASAN_MARK): Likewise.
10481 2018-02-16 Julia Koval <julia.koval@intel.com>
10483 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10484 from PTA_CANNONLAKE.
10486 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10489 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10490 Use ++iter rather than iter++ for std::list iterators.
10491 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10492 defer deleting them until all nodes in the forest are processed. Do
10493 free even leaf nodes. Change to_process into auto_vec.
10496 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10497 * vec.h (vec_default_construct): Use memset instead of placement new
10498 if BROKEN_VALUE_INITIALIZATION is defined.
10499 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10500 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10503 PR rtl-optimization/83723
10504 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10505 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10506 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10508 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10510 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10512 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10514 PR rtl-optimization/81443
10515 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10516 from inner REGs to paradoxical SUBREGs.
10518 2018-02-16 Richard Biener <rguenther@suse.de>
10520 PR tree-optimization/84399
10521 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10522 For operands we can analyze at their definition make sure we can
10523 analyze them at each use as well.
10525 2018-02-16 Richard Biener <rguenther@suse.de>
10527 PR tree-optimization/84190
10528 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10529 volatile accesses if the decl isn't volatile.
10531 2018-02-15 Jason Merrill <jason@redhat.com>
10533 PR c++/84314 - ICE with templates and fastcall attribute.
10534 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10535 TYPE_CANONICAL on an existing type.
10537 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10539 PR tree-optimization/84383
10540 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10541 dstoff nor call operand_equal_p if dstbase is NULL.
10543 PR tree-optimization/84334
10544 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10545 also a CONSTANT_CLASS_P, punt.
10547 2018-02-14 Jim Wilson <jimw@sifive.com>
10549 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10550 first SMALL_OPERAND check. New local min_second_step. Move assert
10551 to where locals are set. Add TARGET_RVC support.
10552 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10554 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10556 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10558 2018-02-14 Martin Sebor <msebor@redhat.com>
10560 PR tree-optimization/83698
10561 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10562 arrays constrain the offset range to their bounds.
10563 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10564 (builtin_access::overlap): Avoid setting the size of overlap if it's
10566 (maybe_diag_overlap): Also consider arrays when deciding what values
10567 of offsets to include in diagnostics.
10569 2018-02-14 Martin Sebor <msebor@redhat.com>
10572 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10573 that correspond to the kind of a declaration.
10575 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10578 * config/pa/pa.md: Load address of PIC label using the linkage table
10579 if the label is nonlocal.
10581 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10583 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10584 warning message if user requests -maltivec=be.
10585 * doc/invoke.texi: Document deprecation of -maltivec=be.
10587 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10590 * config/rs6000/rs6000-c.c: Update definitions for
10591 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10592 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10594 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10597 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10598 add _get_ssp intrinsics. Remove argument from
10599 __builtin_ia32_rdssp[d|q].
10600 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10601 * config/i386/i386-builtin.def: Remove argument from
10602 __builtin_ia32_rdssp[d|q].
10603 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10604 ix86_expand_special_args_builtin for _rdssp[d|q].
10605 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10606 Clear register before usage.
10607 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10608 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10610 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10612 PR tree-optimization/84357
10613 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10614 operand 1 of an ARRAY_REF too.
10616 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10619 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10620 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10622 (set_of_reg): New struct.
10623 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10624 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10625 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10627 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10628 Split into bitclr, bitset, bitinvert patterns if appropriate.
10629 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10630 use rx_fuse_in_memory_bitop.
10631 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10632 to named insn, correct maximum insn length.
10634 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10637 * machmode.def: Define a complex mode for PARTIAL_INT.
10638 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10640 * doc/rtl.texi: Document CSPImode.
10641 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10643 (msp430_hard_regno_nregs_with_padding): Likewise.
10645 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10648 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10650 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10652 PR rtl-optimization/84169
10653 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10654 we generated a parallel as new i3 and we split that to new i2 and i3
10655 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10656 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10657 those to i2, not i1. Partially rewrite this scan code.
10659 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10662 * stor-layout.c (place_field): For variable length fields, adjust
10663 offset_align afterwards not just based on the field's alignment,
10664 but also on the size.
10666 PR middle-end/84309
10667 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10668 of exps and logs in the use_exp2 case.
10670 2018-02-13 Jeff Law <law@redhat.com>
10672 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10673 entry for "vector".
10675 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10678 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10682 * common.opt (gas-loc-support, gas-locview-support): New.
10683 (ginline-points, ginternal-reset-location-views): New.
10684 * doc/invoke.texi: Document them. Use @itemx where intended.
10685 (gvariable-location-views): Adjust.
10686 * target.def (reset_location_view): New.
10687 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10688 (TARGET_RESET_LOCATION_VIEW): New.
10689 * doc/tm.texi: Rebuilt.
10690 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10691 (dwarf2out_default_as_locview_support): New.
10692 (output_asm_line_debug_info): Use option variables.
10693 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10694 (output_loc_list): Likewise.
10695 (add_high_low_attributes): Check option variables.
10696 Don't output entry view attribute in strict mode.
10697 (gen_inlined_subroutine_die): Check option variables.
10698 (dwarf2out_inline_entry): Likewise.
10699 (init_sections_and_labels): Likewise.
10700 (dwarf2out_early_finish): Likewise.
10701 (maybe_reset_location_view): New, from...
10702 (dwarf2out_var_location): ... here. Call it.
10703 * debug.h (dwarf2out_default_as_loc_support): Declare.
10704 (dwarf2out_default_as_locview_support): Declare.
10705 * hooks.c (hook_int_rtx_insn_0): New.
10706 * hooks.h (hook_int_rtx_insn_0): Declare.
10707 * toplev.c (process_options): Take -gas-loc-support and
10708 -gas-locview-support from dwarf2out. Enable
10709 -gvariable-location-views by default only with locview
10710 assembler support. Enable -ginternal-reset-location-views by
10711 default only if the target defines the corresponding hook.
10712 Enable -ginline-points by default if location views are
10713 enabled; force it disabled if statement frontiers are
10715 * tree-inline.c (expand_call_inline): Check option variables.
10716 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10718 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10720 PR tree-optimization/84321
10721 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10722 handling. Also check whether the anti-range contains any values
10723 that satisfy the mask; switch to a VR_RANGE if not.
10725 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10728 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10730 2018-02-13 Martin Jambor <mjambor@suse.cz>
10733 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10734 of call statements, also set location of a load to a temporary.
10736 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10738 * config/rl78/rl78.c (add_vector_labels): New function.
10739 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10740 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10741 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10742 which checks that no arguments are passed.
10743 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10744 * doc/extend.texi: Documentation for the new attribute.
10746 2018-02-13 Andreas Schwab <schwab@suse.de>
10748 * config/riscv/linux.h (CPP_SPEC): Define.
10750 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10753 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10754 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10755 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10756 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10757 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10758 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10759 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10760 temporarily for AES and PCLMUL builtins.
10762 PR tree-optimization/84339
10763 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10764 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10767 PR middle-end/84309
10768 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10769 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10770 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10772 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10774 * omp-simd-clone.h: New file.
10775 * omp-simd-clone.c: Include omp-simd-clone.h.
10776 (expand_simd_clones): No longer static.
10777 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10778 cgraph.h and omp-simd-clone.h.
10779 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10780 (vect_recog_widen_shift_pattern): Formatting fix.
10781 (vect_pattern_recog_1): Don't check optab for calls.
10784 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10785 operands[2] into a REG before using gen_lowpart on it.
10787 2018-02-12 Jeff Law <law@redhat.com>
10790 * config/sh/sh.c (find_barrier): Consider a sibling call
10793 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10794 successfully back substituting a reg.
10796 2018-02-12 Richard Biener <rguenther@suse.de>
10798 PR tree-optimization/84037
10799 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10800 parameter, move visited init to caller.
10801 (vect_slp_analyze_operations): Separate cost from validity
10802 check, initialize visited once for all instances.
10803 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10805 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10807 (vect_model_promotion_demotion_cost): Likewise.
10808 (vectorizable_bswap): Guard cost modeling with !slp_node
10809 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10811 (vectorizable_call): Likewise.
10812 (vectorizable_conversion): Likewise.
10813 (vectorizable_assignment): Likewise.
10814 (vectorizable_shift): Likewise.
10815 (vectorizable_operation): Likewise.
10816 (vectorizable_store): Likewise.
10817 (vectorizable_load): Likewise.
10818 (vectorizable_condition): Likewise.
10819 (vectorizable_comparison): Likewise.
10821 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10824 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10825 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10826 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10828 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10830 PR middle-end/83665
10831 * params.def (inline-min-speedup): Increase from 8 to 15.
10832 (max-inline-insns-auto): Decrease from 40 to 30.
10833 * ipa-split.c (consider_split): Add some buffer for function to
10834 be considered inlining candidate.
10835 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10838 2018-02-12 Richard Biener <rguenther@suse.de>
10840 PR tree-optimization/84037
10841 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10842 matched stmts if we cannot swap the non-matched ones.
10844 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10846 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10847 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10848 _mm_maskz_scalef_round_ss): New intrinsics.
10849 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10850 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10851 __builtin_ia32_scalefss_round): Remove.
10852 (__builtin_ia32_scalefsd_mask_round,
10853 __builtin_ia32_scalefss_mask_round): New intrinsics.
10854 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10855 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10856 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10857 "<round_constraint>")): Changed to ...
10858 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10859 "<round_scalar_constraint>")): ... this.
10860 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10861 %0, %1, %2<round_op3>}"): Changed to ...
10862 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10863 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10864 %2<round_scalar_mask_op3>}"): ... this.
10865 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10867 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10869 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10870 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10871 (_mm_maskz_sqrt_round_ss): New intrinsics.
10872 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10873 (__builtin_ia32_sqrtsd_mask_round)
10874 (__builtin_ia32_sqrtss_mask_round): New builtins.
10875 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10876 (__builtin_ia32_sqrtss_round): Remove.
10877 (__builtin_ia32_sqrtsd_mask_round)
10878 (__builtin_ia32_sqrtss_mask_round): New builtins.
10879 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10880 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10881 ((match_operand:VF_128 1 "vector_operand"
10882 "xBm,<round_constraint>")): Changed to ...
10883 ((match_operand:VF_128 1 "vector_operand"
10884 "xBm,<round_scalar_constraint>")): ... this.
10885 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10886 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10887 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10888 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10889 %<iptr>1<round_scalar_mask_op3>}): ... this.
10890 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10891 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10893 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10896 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10897 Cast vec_cmpeq result to correct type.
10898 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10899 Cast vec_cmpgt result to correct type.
10901 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10903 * final.c (final_scan_insn_1): Renamed from...
10904 (final_scan_insn): ... this. New wrapper, to recover
10905 seen from the outermost call in recursive ones.
10906 * config/sparc/sparc.c (output_return): Drop seen from call.
10907 (output_sibcall): Likewise.
10908 * config/visium/visium.c (output_branch): Likewise.
10910 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10912 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10915 2018-02-10 Alan Modra <amodra@gmail.com>
10918 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10919 Specify LR as an input.
10921 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10924 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10925 remove_member_access_dummy_vars): New functions.
10926 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10927 lower_omp_1, execute_lower_omp): Use them.
10929 PR rtl-optimization/84308
10930 * shrink-wrap.c (spread_components): Release todo vector.
10932 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10934 PR rtl-optimization/57193
10935 * ira-color.c (struct allocno_color_data): Add member
10936 conflict_allocno_hard_prefs.
10937 (update_conflict_allocno_hard_prefs): New.
10938 (bucket_allocno_compare_func): Add a preference based on
10939 conflict_allocno_hard_prefs.
10940 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10941 (color_allocnos): Remove a dead code. Initiate
10942 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10944 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10947 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10948 constraint from =wa to wa. Avoid a subreg on the output operand,
10949 instead use a pseudo and subreg it in a move.
10950 (p9_xxbrd_<mode>): Changed to ...
10951 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10952 (p9_xxbrd_v2df): New expander.
10953 (p9_xxbrw_<mode>): Changed to ...
10954 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10955 (p9_xxbrw_v4sf): New expander.
10957 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10959 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10961 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10964 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10965 multiply in 32-bit mode.
10966 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10967 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10970 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10972 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10973 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10974 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10975 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10977 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10980 * dwarf2out.c (is_trivial_indirect_ref): New function.
10981 (dwarf2out_late_global_decl): Do not generate a location
10982 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10983 and that are not defined in the current unit.
10985 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10987 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10988 instead of a libcall for UNORDERED.
10990 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10993 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10994 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10996 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10999 * config/s390/s390.c (s390_set_current_function): Invoke
11000 s390_indirect_branch_settings also if fndecl didn't change.
11002 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11004 * config/rs6000/rs6000.md (blockage): Set length to zero.
11006 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
11008 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
11010 2018-02-09 Jakub Jelinek <jakub@redhat.com>
11013 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
11014 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
11018 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
11019 PARALLEL incoming that failed vt_get_decl_and_offset check.
11021 PR middle-end/84237
11022 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
11023 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
11025 (get_variable_section): For decls in named .bss* sections pass true as
11026 second argument to bss_initializer_p.
11028 2018-02-09 Marek Polacek <polacek@redhat.com>
11029 Jakub Jelinek <jakub@redhat.com>
11032 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
11033 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
11034 Sync some changes from cxx_fold_indirect_ref.
11036 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
11038 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
11040 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
11041 (BLOCK_INLINE_ENTRY_LABEL): New.
11042 (dwarf2out_var_location): Disregard inline entry markers.
11043 (inline_entry_data): New struct.
11044 (inline_entry_data_hasher): New hashtable type.
11045 (inline_entry_data_hasher::hash): New.
11046 (inline_entry_data_hasher::equal): New.
11047 (inline_entry_data_table): New variable.
11048 (add_high_low_attributes): Add DW_AT_entry_pc and
11049 DW_AT_GNU_entry_view attributes if a pending entry is found
11050 in inline_entry_data_table. Add old entry_pc attribute only
11051 if debug nonbinding markers are disabled.
11052 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
11053 markers are enabled.
11054 (block_within_block_p, dwarf2out_inline_entry): New.
11055 (dwarf2out_finish): Check that no entries remained in
11056 inline_entry_data_table.
11057 * final.c (reemit_insn_block_notes): Handle inline entry notes.
11058 (final_scan_insn, notice_source_line): Likewise.
11059 (rest_of_clean_state): Skip inline entry markers.
11060 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
11062 * gimple.c (gimple_build_debug_inline_entry): New.
11063 * gimple.h (enum gimple_debug_subcode): Add
11064 GIMPLE_DEBUG_INLINE_ENTRY.
11065 (gimple_build_debug_inline_entry): Declare.
11066 (gimple_debug_inline_entry_p): New.
11067 (gimple_debug_nonbind_marker_p): Adjust.
11068 * insn-notes.def (INLINE_ENTRY): New.
11069 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
11070 inline entry marker notes.
11071 (print_insn): Likewise.
11072 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
11073 (INSN_DEBUG_MARKER_KIND): Likewise.
11074 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
11075 * tree-inline.c (expand_call_inline): Build and insert
11076 debug_inline_entry stmt.
11077 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
11078 inline entry blocks early, if nonbind markers are enabled.
11079 (dump_scope_block): Dump fragment info.
11080 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
11081 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
11082 (gimple_build_debug_inline_entry): New.
11083 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
11084 Enable/disable inline entry points too.
11085 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
11086 (DEBUG_INSN): Describe inline entry markers.
11088 * common.opt (gvariable-location-views): New.
11089 (gvariable-location-views=incompat5): New.
11090 * config.in: Rebuilt.
11091 * configure: Rebuilt.
11092 * configure.ac: Test assembler for view support.
11093 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
11094 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
11095 * dwarf2out.c (var_loc_view): New typedef.
11096 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
11097 (dwarf2out_locviews_in_attribute): New.
11098 (dwarf2out_locviews_in_loclist): New.
11099 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
11100 (enum dw_line_info_opcode): Add LI_adv_address.
11101 (struct dw_line_info_table): Add view.
11102 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
11103 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
11104 (zero_view_p): New variable.
11105 (ZERO_VIEW_P): New macro.
11106 (output_asm_line_debug_info): New.
11107 (struct var_loc_node): Add view.
11108 (add_AT_view_list, AT_loc_list): New.
11109 (add_var_loc_to_decl): Add view param. Test it against last.
11110 (new_loc_list): Add view params. Record them.
11111 (AT_loc_list_ptr): Handle loc and view lists.
11112 (view_list_to_loc_list_val_node): New.
11113 (print_dw_val): Handle dw_val_class_view_list.
11114 (size_of_die): Likewise.
11115 (value_format): Likewise.
11116 (loc_list_has_views): New.
11117 (gen_llsym): Set vl_symbol too.
11118 (maybe_gen_llsym, skip_loc_list_entry): New.
11119 (dwarf2out_maybe_output_loclist_view_pair): New.
11120 (output_loc_list): Output view list or entries too.
11121 (output_view_list_offset): New.
11122 (output_die): Handle dw_val_class_view_list.
11123 (output_dwarf_version): New.
11124 (output_compilation_unit_header): Use it.
11125 (output_skeleton_debug_sections): Likewise.
11126 (output_rnglists, output_line_info): Likewise.
11127 (output_pubnames, output_aranges): Update version comments.
11128 (output_one_line_info_table): Output view numbers in asm comments.
11129 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11130 Call maybe_gen_llsym.
11131 (loc_list_from_tree_1): Adjust.
11132 (add_AT_location_description): Create view list attribute if
11133 needed, check it's absent otherwise.
11134 (convert_cfa_to_fb_loc_list): Adjust.
11135 (maybe_emit_file): Call output_asm_line_debug_info for test.
11136 (dwarf2out_var_location): Reset views as needed. Precompute
11137 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11138 attribute. Set view.
11139 (new_line_info_table): Reset next view.
11140 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11141 (dwarf2out_source_line): Likewise. Output view resets and labels to
11142 the assembler, or select appropriate line info opcodes.
11143 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11144 (optimize_string_length): Catch it. Adjust.
11145 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11146 dw_val_class_view_list, and remove it if no longer needed.
11147 (hash_loc_list): Hash view numbers.
11148 (loc_list_hasher::equal): Compare them.
11149 (optimize_location_lists): Check whether a view list symbol is
11150 needed, and whether the locview attribute is present, and
11151 whether they match. Remove the locview attribute if no longer
11153 (index_location_lists): Call skip_loc_list_entry for test.
11154 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11155 Use output_dwarf_version.
11156 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11157 (struct dw_val_node): Add val_view_list.
11158 * final.c (SEEN_NEXT_VIEW): New.
11159 (set_next_view_needed): New.
11160 (clear_next_view_needed): New.
11161 (maybe_output_next_view): New.
11162 (final_start_function): Rename to...
11163 (final_start_function_1): ... this. Take pointer to FIRST,
11164 add SEEN parameter. Emit param bindings in the initial view.
11165 (final_start_function): Reintroduce SEEN-less interface.
11166 (final): Rename to...
11167 (final_1): ... this. Take SEEN parameter. Output final pending
11168 next view at the end.
11169 (final): Reintroduce seen-less interface.
11170 (final_scan_insn): Output pending next view before switching
11171 sections or ending a block. Mark the next view as needed when
11172 outputting variable locations. Notify debug backend of section
11173 changes, and of location view changes.
11174 (rest_of_handle_final): Adjust.
11175 * toplev.c (process_options): Autodetect value for debug variable
11176 location views option. Warn on incompat5 without -gdwarf-5.
11177 * doc/invoke.texi (gvariable-location-views): New.
11178 (gvariable-location-views=incompat5): New.
11179 (gno-variable-location-views): New.
11181 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11183 PR tree-optimization/84136
11184 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11185 that the result of find_edge is non-NULL.
11187 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11190 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11191 storing integer register in SImode. Fix cost of 256 and 512
11192 byte aligned SSE register store.
11194 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11196 * config/i386/i386.c (ix86_multiplication_cost): Fix
11197 multiplication cost for TARGET_AVX512DQ.
11199 2018-02-08 Marek Polacek <polacek@redhat.com>
11201 PR tree-optimization/84238
11202 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11205 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11207 PR tree-optimization/84265
11208 * tree-vect-stmts.c (vectorizable_store): Don't treat
11209 VMAT_CONTIGUOUS accesses as grouped.
11210 (vectorizable_load): Likewise.
11212 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11214 PR tree-optimization/81635
11215 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11216 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11217 (test_round_for_mask): New functions.
11218 (wide_int_cc_tests): Call test_round_for_mask.
11219 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11220 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11221 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11222 range returned by get_range_info.
11224 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11227 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11228 * symtab.c: Include builtins.h
11229 (symtab_node::output_to_lto_symbol_table_p): Move here
11230 from lto-streamer-out.c:output_symbol_p.
11231 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11232 (output_symbol_p): Move all logic to symtab.c
11233 (produce_symtab): Update.
11235 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11237 * config/s390/s390-opts.h (enum indirect_branch): Define.
11238 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11239 (s390_indirect_branch_via_thunk)
11240 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11241 (enum s390_indirect_branch_type): Define.
11242 * config/s390/s390.c (struct s390_frame_layout, struct
11243 machine_function): Remove.
11244 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11245 (indirect_branch_table_label_no, indirect_branch_table_name):
11247 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11248 (enum s390_indirect_branch_option): Define.
11249 (s390_return_addr_from_memory): New function.
11250 (s390_handle_string_attribute): New function.
11251 (s390_attribute_table): Add new attribute handler.
11252 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11253 (s390_indirect_branch_via_thunk): New function.
11254 (s390_indirect_branch_via_inline_thunk): New function.
11255 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11256 sibling call optimization for non z10 compiles.
11257 (s390_emit_call): Force indirect branch target to be a single
11258 register. Add r1 clobber for non-z10 compiles.
11259 (s390_emit_epilogue): Emit return jump via return_use expander.
11260 (s390_reorg): Handle JUMP_INSNs as execute targets.
11261 (s390_option_override_internal): Perform validity checks for the
11262 new command line options.
11263 (s390_indirect_branch_attrvalue): New function.
11264 (s390_indirect_branch_settings): New function.
11265 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11266 (s390_output_indirect_thunk_function): New function.
11267 (s390_code_end): Implement target hook.
11268 (s390_case_values_threshold): Implement target hook.
11269 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11271 * config/s390/s390.h (struct s390_frame_layout)
11272 (struct machine_function): Move here from s390.c.
11273 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11274 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11275 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11276 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11277 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11278 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11279 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11280 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11281 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11282 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11283 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11284 (mnemonic attribute): Add values which aren't recognized
11286 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11287 pattern for branch conversion. Fix mnemonic attribute.
11288 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11289 indirect branch via thunk if requested.
11290 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11291 ("*indirect_jump"): Disable for branch conversion using out of
11293 ("indirect_jump_via_thunk<mode>_z10")
11294 ("indirect_jump_via_thunk<mode>")
11295 ("indirect_jump_via_inlinethunk<mode>_z10")
11296 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11297 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11298 ("casesi_jump_via_inlinethunk<mode>_z10")
11299 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11300 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11301 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11302 ("*indirect2_jump"): Disable for branch conversion.
11303 ("casesi_jump"): Turn into expander and expand patterns for branch
11305 ("return_use"): New expander.
11306 ("*return"): Emit return via thunk and rename it to ...
11307 ("*return<mode>"): ... this one.
11308 * config/s390/s390.opt: Add new options and and enum for the
11311 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11313 * lra-constraints.c (match_reload): Unconditionally use
11314 gen_lowpart_SUBREG, rather than selecting between that
11315 and equivalent gen_rtx_SUBREG code.
11317 2018-02-08 Richard Biener <rguenther@suse.de>
11319 PR tree-optimization/84233
11320 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11321 changed flag instead of boguously re-using phi_inserted.
11323 2018-02-08 Martin Jambor <mjambor@suse.cz>
11325 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11326 static local variables.
11328 2018-02-08 Richard Biener <rguenther@suse.de>
11330 PR tree-optimization/84278
11331 * tree-vect-stmts.c (vectorizable_store): When looking for
11332 smaller vector types to perform grouped strided loads/stores
11333 make sure the mode is supported by the target.
11334 (vectorizable_load): Likewise.
11336 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11338 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11339 Increase LDP/STP opportunities by adding adjacent callee-saves.
11341 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11343 PR rtl-optimization/84068
11344 PR rtl-optimization/83459
11345 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11347 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11349 PR tree-optimization/84224
11350 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11351 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11352 non-zero arguments.
11354 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11357 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11358 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11359 count, remove one USE.
11361 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11363 * doc/install.texi (Configuration): Document the
11364 --with-long-double-format={ibm,ieee} PowerPC configuration
11368 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11369 Convert from define_expand to be define_insn_and_split. Rework
11370 float/double/_Float128 conversions to QI/HI/SImode to work with
11371 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11372 conversions to QI/HImode types did a store and then a load to
11373 truncate the value. For conversions to VSX registers, don't split
11374 the insn, instead emit the code directly. Use the code iterator
11375 any_fix to combine signed and unsigned conversions.
11376 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11377 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11378 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11379 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11380 (fix_<mode>di2_hw): Likewise.
11381 (fixuns_<mode>di2_hw): Likewise.
11382 (fix_<mode>si2_hw): Likewise.
11383 (fixuns_<mode>si2_hw): Likewise.
11384 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11385 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11386 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11387 fix<uns>_trunc<SFDF:mode>si2_p8.
11388 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11389 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11390 (fix<uns>_<mode>_mem): Likewise.
11391 (fctiw<u>z_<mode>_mem): Likewise.
11392 (fix<uns>_<mode>_mem): Likewise.
11393 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11394 the register allocator from doing a direct move to the GPRs to do
11395 a store, and instead use the ISA 3.0 store byte/half-word from
11396 vector register instruction. For IEEE 128-bit floating point,
11397 also optimize stores of 32-bit ints.
11398 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11400 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11402 * genextract.c (push_pathstr_operand): New function to support
11404 (walk_rtx): Call push_pathstr_operand.
11405 (print_path): Support [a-zA-Z].
11407 2018-02-07 Richard Biener <rguenther@suse.de>
11409 PR tree-optimization/84037
11410 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11411 (cse_and_gimplify_to_preheader): Declare.
11412 (vect_get_place_in_interleaving_chain): Likewise.
11413 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11415 (_loop_vec_info::~_loop_vec_info): Delete it.
11416 (cse_and_gimplify_to_preheader): New function.
11417 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11418 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11419 (vectorizable_load): Likewise. For grouped stores always base
11420 the IV on the first element.
11421 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11422 condition before gimplifying.
11424 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11426 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11427 *DIV_EXPR and *MOD_EXPR.
11429 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11432 * config/i386/i386.c (ix86_option_override_internal): Mask out
11433 the CF_SET bit when checking -fcf-protection.
11435 2018-02-07 Tom de Vries <tom@codesourcery.com>
11438 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11441 2018-02-07 Richard Biener <rguenther@suse.de>
11443 PR tree-optimization/84204
11444 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11447 PR tree-optimization/84205
11448 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11449 special-case isl_ast_op_zdiv_r.
11451 PR tree-optimization/84223
11452 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11453 Only add conditions from within the region.
11454 (gather_bbs::after_dom_children): Adjust.
11456 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11459 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11460 * config/avr/avr.md: Only post-reload split REG-REG moves if
11461 either register is GENERAL_REG_P.
11463 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11465 PR tree-optimization/84235
11466 * tree-ssa-scopedtables.c
11467 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11468 if the subtraction is performed in floating point type where NaNs are
11469 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11470 build 1. Formatting fix.
11472 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11475 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11476 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11477 and skip it regardless of bb boundaries. Use CALL_P macro,
11478 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11479 unnecessarily, formatting fix.
11481 2018-02-06 Michael Collison <michael.collison@arm.com>
11483 * config/arm/thumb2.md:
11484 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11485 (*thumb_mov_notscc): Ditto.
11487 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11490 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11491 unsigned_fix, not "s".
11493 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11495 * configure.ac (gcc_fn_eh_frame_ro): New function.
11496 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11497 correct .eh_frame permissions.
11498 * configure: Regenerate.
11500 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11502 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11503 irrelevant options.
11505 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11507 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11508 Display warning message for -mno-speculate-indirect-jumps.
11510 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11512 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11514 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11516 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11518 PR tree-optimization/84225
11519 * tree-eh.c (find_trapping_overflow): Only call
11520 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11522 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11525 * config/i386/i386.c: Reimplement the check of possible options
11526 -mibt/-mshstk conbination. Change error messages.
11527 * doc/invoke.texi: Fix a typo: remove extra '='.
11529 2018-02-06 Marek Polacek <polacek@redhat.com>
11531 PR tree-optimization/84228
11532 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11534 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11537 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11538 emitted arch directives.
11539 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11540 __ARM_FEATURE_COPROC before changing architectures.
11542 2018-02-06 Richard Biener <rguenther@suse.de>
11544 * config/i386/i386.c (print_reg): Fix typo.
11545 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11547 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11549 * configure: Regenerate.
11551 2018-02-05 Martin Sebor <msebor@redhat.com>
11553 PR tree-optimization/83369
11554 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11557 2018-02-05 Martin Liska <mliska@suse.cz>
11559 * doc/invoke.texi: Cherry-pick upstream r323995.
11561 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11563 * ira.c (ira_init_register_move_cost): Adjust comment.
11565 2018-02-05 Martin Liska <mliska@suse.cz>
11567 PR gcov-profile/84137
11568 * doc/gcov.texi: Fix typo in documentation.
11570 2018-02-05 Martin Liska <mliska@suse.cz>
11572 PR gcov-profile/83879
11573 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11574 using dlopen functionality.
11576 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11578 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11579 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11580 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11581 _mm_maskz_range_round_ss): New intrinsics.
11582 (__builtin_ia32_rangesd128_round)
11583 (__builtin_ia32_rangess128_round): Remove.
11584 (__builtin_ia32_rangesd128_mask_round,
11585 __builtin_ia32_rangess128_mask_round): New builtins.
11586 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11587 __builtin_ia32_rangess128_round): Remove.
11588 (__builtin_ia32_rangesd128_mask_round,
11589 __builtin_ia32_rangess128_mask_round): New builtins.
11590 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11591 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11592 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11593 "<round_saeonly_constraint>")): Changed to ...
11594 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11595 "<round_saeonly_scalar_constraint>")): ... this.
11596 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11597 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11598 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11599 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11600 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11602 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11604 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11606 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11607 Remove all values except native, 8540 and 8548.
11609 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11611 * config/i386/i386.c (ix86_output_function_return): Pass
11612 INVALID_REGNUM, instead of -1, as invalid register number to
11613 indirect_thunk_name and output_indirect_thunk.
11615 2018-02-02 Julia Koval <julia.koval@intel.com>
11617 * config.gcc: Add -march=icelake.
11618 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11619 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11620 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11621 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11622 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11623 (processor_target_table): Add icelake.
11624 (ix86_option_override_internal): Handle new PTAs.
11625 (get_builtin_code_for_version): Handle icelake.
11626 (M_INTEL_COREI7_ICELAKE): New.
11627 (fold_builtin_cpu): Handle icelake.
11628 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11629 * doc/invoke.texi: Add -march=icelake.
11631 2018-02-02 Julia Koval <julia.koval@intel.com>
11633 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11634 to wide_int_bitmask.
11635 * wide-int-bitmask.h: New.
11637 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11640 * config/i386/i386.md: Replace Pmode with word_mode in
11641 builtin_setjmp_setup and builtin_longjmp to support x32.
11643 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11647 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11649 (rs6000_supported_cpu_names): New static variable.
11650 (linux_cpu_translation_table): Likewise.
11651 (elf_platform) <cpu>: Define new static variable and use it.
11652 Translate kernel AT_PLATFORM name to canonical name if needed.
11653 Error if platform name is unknown.
11655 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11658 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11660 2018-02-01 Jeff Law <law@redhat.com>
11663 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11664 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11665 the scratch if RELEASE_VIA_POP is false.
11666 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11667 If we have to save a temporary register, decrement SIZE appropriately.
11668 Pass new arguments to release_scratch_register_on_entry.
11669 (ix86_adjust_stack_and_probe): Likewise.
11670 (ix86_emit_probe_stack_range): Pass new arguments to
11671 release_scratch_register_on_entry.
11673 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11675 PR rtl-optimization/84157
11676 * combine.c (change_zero_ext): Use REG_P predicate in
11677 front of HARD_REGISTER_P predicate.
11679 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11681 * config/avr/avr.c (avr_option_override): Move disabling of
11682 -fdelete-null-pointer-checks to...
11683 * common/config/avr/avr-common.c (avr_option_optimization_table):
11686 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11688 PR tree-optimization/81635
11689 * tree-data-ref.c (split_constant_offset_1): For types that
11690 wrap on overflow, try to use range info to prove that wrapping
11693 2018-02-01 Renlin Li <renlin.li@arm.com>
11696 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11697 TAILCALL_ADDR_REGS.
11698 (aarch64_register_move_cost): Likewise.
11699 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11700 TAILCALL_ADDR_REGS.
11701 (REG_CLASS_NAMES): Likewise.
11702 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11703 TAILCALL_ADDR_REGS. Remove IP registers.
11704 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11706 2018-02-01 Richard Biener <rguenther@suse.de>
11708 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11709 for specifying RPO order and allow NULL for that.
11710 * domwalk.c (dom_walker::dom_walker): Likewise.
11711 (dom_walker::walk): Handle NULL RPO order.
11712 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11714 (rewrite_update_dom_walker): Likewise.
11715 (mark_def_dom_walker): Likewise.
11717 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11719 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11720 (aarch64_maybe_expand_sve_subreg_move): Declare.
11721 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11722 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11724 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11725 that are semantically a reverse operation.
11726 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11727 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11728 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11730 (aarch64_can_change_mode_class): For big-endian, forbid changes
11731 between two SVE modes if they have different element sizes.
11733 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11735 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11736 the TImode handling for big-endian targets.
11738 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11740 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11741 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11743 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11744 Remove BSWAP handing for big-endian targets and use the form of
11745 LD1RQ appropariate for the mode.
11747 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11749 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11750 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11751 duplicated element.
11753 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11756 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11757 check for operands that need to go through aarch64_sve_reload_be.
11759 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11761 PR tree-optimization/81661
11762 PR tree-optimization/84117
11763 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11764 * tree-eh.c: Include gimplify.h.
11765 (find_trapping_overflow, replace_trapping_overflow,
11766 rewrite_to_non_trapping_overflow): New functions.
11767 * tree-vect-loop.c: Include tree-eh.h.
11768 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11769 * tree-data-ref.c: Include tree-eh.h.
11770 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11772 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11774 PR rtl-optimization/84123
11775 * combine.c (change_zero_ext): Check if hard register satisfies
11776 can_change_dest_mode before calling gen_lowpart_SUBREG.
11778 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11781 * ira.c (ira_init_register_move_cost): Remove assert.
11783 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11785 PR rtl-optimization/84071
11786 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11787 * doc/tm.texi: Regenerate.
11789 2018-01-31 Richard Biener <rguenther@suse.de>
11791 PR tree-optimization/84132
11792 * tree-data-ref.c (analyze_miv_subscript): Properly
11793 check whether evolution_function_is_affine_multivariate_p
11794 before calling gcd_of_steps_may_divide_p.
11796 2018-01-31 Julia Koval <julia.koval@intel.com>
11799 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11800 * config/i386/i386.md (rdpid_rex64) New.
11801 (rdpid): Make 32bit only.
11803 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11806 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11807 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11809 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11812 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11814 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11816 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11818 PR rtl-optimization/84071
11819 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11820 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11822 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11824 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11825 (arc_attribute_table): Add 'aux' attribute.
11826 (arc_in_small_data_p): Consider aux like variables.
11827 (arc_is_aux_reg_p): New function.
11828 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11829 (arc_get_aux_arg): New function.
11830 (prepare_move_operands): Handle aux-register access.
11831 (arc_handle_aux_attribute): New function.
11832 * doc/extend.texi (ARC Variable attributes): Add subsection.
11834 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11836 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11837 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11838 (arc_attribute_table): Add 'uncached' attribute.
11839 (arc_print_operand): Print '.di' flag for uncached memory
11841 (arc_in_small_data_p): Do not consider for small data the uncached
11843 (arc_is_uncached_mem_p): New function.
11844 * config/arc/predicates.md (compact_store_memory_operand): Check
11845 for uncached memory accesses.
11846 (nonvol_nonimm_operand): Likewise.
11847 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11849 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11852 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11853 falign-loops=): Add Optimization flag.
11855 2018-01-30 Jeff Law <law@redhat.com>
11858 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11859 INT_REGISTERS_SAVED. Check it prior to calling
11860 get_scratch_register_on_entry.
11861 (ix86_adjust_stack_and_probe): Similarly.
11862 (ix86_emit_probe_stack_range): Similarly.
11863 (ix86_expand_prologue): Corresponding changes.
11865 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11868 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11869 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11871 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11874 * lra-constraints.c (curr_insn_transform): Process AND in the
11877 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11879 PR rtl-optimization/83986
11880 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11881 dependence against last_pending_memory_flush in addition to
11882 pending_jump_insns.
11884 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11886 PR tree-optimization/81611
11887 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11890 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11893 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11896 2018-01-30 Richard Biener <rguenther@suse.de>
11897 Jakub Jelinek <jakub@redhat.com>
11899 PR tree-optimization/84111
11900 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11901 inner loops added during recursion, as they don't have up-to-date
11904 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11907 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11908 (can_inline_edge_by_limits_p): ... here.
11909 (can_early_inline_edge_p, check_callers,
11910 update_caller_keys, update_callee_keys, recursive_inlining,
11911 add_new_edges_to_heap, speculation_useful_p,
11912 inline_small_functions,
11913 inline_small_functions, flatten_function,
11914 inline_to_all_callers_1): Update.
11916 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11918 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11921 2018-01-30 Richard Biener <rguenther@suse.de>
11923 PR tree-optimization/83008
11924 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11925 invariant and constant vector uses in stmts when they need
11926 more than one stmt.
11928 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11931 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11932 * configure: Regenerate.
11934 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11936 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11938 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11939 Use gen_rtx_REG rather than gen_lowpart.
11941 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11943 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11944 rather than 0 when creating partial subregs.
11946 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11948 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11951 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11954 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11955 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11956 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11957 flags. This restores the settings used before the 2017-07-24.
11958 Turning off pre increment/decrement/modify allows IVOPTS to
11959 optimize DF/SF loops where the index is an int.
11961 2018-01-29 Richard Biener <rguenther@suse.de>
11962 Kelvin Nilsen <kelvin@gcc.gnu.org>
11965 * tree-vect-stmts.c (vectorizable_call): Don't call
11966 targetm.vectorize_builtin_md_vectorized_function if callee is
11969 2018-01-22 Carl Love <cel@us.ibm.com>
11971 * doc/extend.tex: Fix typo in second arg in
11972 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11974 2018-01-29 Richard Biener <rguenther@suse.de>
11976 PR tree-optimization/84086
11977 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11978 (flush_ssaname_freelist): When SSA names were released reset
11979 the SCEV hash table.
11981 2018-01-29 Richard Biener <rguenther@suse.de>
11983 PR tree-optimization/84057
11984 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11985 removed paths when removing edges.
11987 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11989 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11990 -mfunction-return=@var{choice}.
11992 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11994 PR diagnostic/84034
11995 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11996 Handle CR like TAB.
11997 (layout::print_source_line): Likewise.
11998 (test_get_line_width_without_trailing_whitespace): Add test cases.
12000 2018-01-27 Jakub Jelinek <jakub@redhat.com>
12002 PR middle-end/84040
12003 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
12006 2018-01-26 Jim Wilson <jimw@sifive.com>
12008 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
12010 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
12013 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12015 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
12016 and CMP + SUB-immediate -> SUBS.
12018 2018-01-26 Martin Sebor <msebor@redhat.com>
12020 PR tree-optimization/83896
12021 * tree-ssa-strlen.c (get_string_len): Rename...
12022 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
12023 Avoid assuming length is constant.
12024 (handle_char_store): Use HOST_WIDE_INT for string length.
12026 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
12029 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
12030 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
12032 2018-01-26 Richard Biener <rguenther@suse.de>
12034 PR rtl-optimization/84003
12035 * dse.c (record_store): Only record redundant stores when
12036 the earlier store aliases at least all accesses the later one does.
12038 2018-01-26 Jakub Jelinek <jakub@redhat.com>
12040 PR rtl-optimization/83985
12041 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
12042 REG_CFA_RESTORE insns.
12043 (delete_unmarked_insns): Don't ignore separate shrink wrapping
12044 REG_CFA_RESTORE insns here.
12047 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
12048 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
12050 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12052 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
12053 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
12054 (arc_init): Likewise.
12055 (arc_override_options): Likewise.
12056 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
12058 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
12060 * config/arc/arc.h (TARGET_DBNZ): Define.
12061 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
12062 properly set the tune attribute.
12063 (dbnz): Use TARGET_DBNZ guard.
12064 * config/arc/arc.opt (mtune): Add core3 option.
12066 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12068 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
12069 recognize new pic like addresses.
12070 (arc_delegitimize_address): Clean up.
12072 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12074 * config/arc/arc-arches.def: Option mrf16 valid for all
12076 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
12077 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
12078 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
12079 * config/arc/arc-tables.opt: Regenerate.
12080 * config/arc/arc.c (arc_conditional_register_usage): Handle
12081 reduced register file case.
12082 (arc_file_start): Set must have build attributes.
12083 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
12084 mrf16 option value.
12085 * config/arc/arc.opt (mrf16): Add new option.
12086 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
12087 * config/arc/genmultilib.awk: Handle new mrf16 option.
12088 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
12089 * config/arc/t-multilib: Regenerate.
12090 * doc/invoke.texi (ARC Options): Document mrf16 option.
12092 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12094 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
12095 * config/arc/arc.c (arc_handle_secure_attribute): New function.
12096 (arc_attribute_table): Add 'secure_call' attribute.
12097 (arc_print_operand): Print secure call operand.
12098 (arc_function_ok_for_sibcall): Don't optimize tail calls when
12100 (arc_is_secure_call_p): New function. * config/arc/arc.md
12101 (call_i): Add support for sjli instruction.
12102 (call_value_i): Likewise.
12103 * config/arc/constraints.md (Csc): New constraint.
12105 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
12106 John Eric Martin <John.Martin@emmicro-us.com>
12108 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
12109 * config/arc/arc.c (_arc_jli_section): New struct.
12110 (arc_jli_section): New type.
12111 (rc_jli_sections): New static variable.
12112 (arc_handle_jli_attribute): New function.
12113 (arc_attribute_table): Add jli_always and jli_fixed attribute.
12114 (arc_file_end): New function.
12115 (TARGET_ASM_FILE_END): Define.
12116 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
12117 (arc_add_jli_section): New function.
12118 (jli_call_scan): Likewise.
12119 (arc_reorg): Call jli_call_scan.
12120 (arc_output_addsi): Remove 'S' from printing asm operand.
12121 (arc_is_jli_call_p): New function.
12122 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
12124 (movhi_insn): Likewise.
12125 (movsi_insn): Likewise.
12126 (movsi_set_cc_insn): Likewise.
12127 (loadqi_update): Likewise.
12128 (load_zeroextendqisi_update): Likewise.
12129 (load_signextendqisi_update): Likewise.
12130 (loadhi_update): Likewise.
12131 (load_zeroextendhisi_update): Likewise.
12132 (load_signextendhisi_update): Likewise.
12133 (loadsi_update): Likewise.
12134 (loadsf_update): Likewise.
12135 (movsicc_insn): Likewise.
12136 (bset_insn): Likewise.
12137 (bxor_insn): Likewise.
12138 (bclr_insn): Likewise.
12139 (bmsk_insn): Likewise.
12140 (bicsi3_insn): Likewise.
12141 (cmpsi_cc_c_insn): Likewise.
12142 (movsi_ne): Likewise.
12143 (movsi_cond_exec): Likewise.
12144 (clrsbsi2): Likewise.
12145 (norm_f): Likewise.
12155 (call_i): Remove 'S' asm letter, add jli instruction.
12156 (call_value_i): Likewise.
12157 * config/arc/arc.op (mjli-always): New option.
12158 * config/arc/constraints.md (Cji): New constraint.
12159 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12161 (subsf3_fpx): Likewise.
12162 (mulsf3_fpx): Likewise.
12163 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12165 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12166 function attrbutes.
12167 * doc/invoke.texi (ARC): Document mjli-always option.
12169 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12171 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12172 avoid addition with 0 and use incw and decw where possible.
12174 2018-01-26 Richard Biener <rguenther@suse.de>
12176 PR tree-optimization/81082
12177 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12178 association if it requires casting to unsigned.
12179 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12180 from fold_plusminus_mult_expr to catch important cases late when
12181 range info is available.
12183 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12185 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12186 * configure.ac (hidden_linkonce): New test.
12187 * configure: Regenerate.
12188 * config.in: Regenerate.
12190 2018-01-26 Julia Koval <julia.koval@intel.com>
12192 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12193 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12194 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12195 _mm_mask_bitshuffle_epi64_mask): Fix type.
12196 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12197 USI_FTYPE_V4DI_V4DI_USI): Remove.
12198 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12199 __builtin_ia32_vpshufbitqmb256_mask,
12200 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12201 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12202 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12204 2018-01-26 Alan Modra <amodra@gmail.com>
12207 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12208 UNSPEC_VBPERMQ. Sort other unspecs.
12210 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12212 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12214 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12216 PR middle-end/83055
12217 * predict.c (drop_profile): Do not push/pop cfun; update also
12219 (handle_missing_profiles): Fix logic looking for zero profiles.
12221 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12223 PR middle-end/83977
12224 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12225 on functions with #pragma omp declare simd or functions with simd
12227 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12228 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12229 Remove trailing \n from warning_at calls.
12231 2018-01-25 Tom de Vries <tom@codesourcery.com>
12234 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12235 for neutered workers.
12237 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12240 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12241 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12243 2018-01-24 Jeff Law <law@redhat.com>
12246 * i386.c (get_probe_interval): Move to earlier point.
12247 (ix86_compute_frame_layout): If -fstack-clash-protection and
12248 the frame is larger than the probe interval, then use pushes
12249 to save registers rather than reg->mem moves.
12250 (ix86_expand_prologue): Remove conditional for int_registers_saved
12253 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12256 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12257 min/max for never referenced object.
12259 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12261 PR middle-end/83977
12262 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12264 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12265 attributes from DECL_ATTRIBUTES (decl) without affecting
12266 DECL_ATTRIBUTES (current_function_decl).
12267 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12268 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12270 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12272 PR tree-optimization/83979
12273 * fold-const.c (fold_comparison): Use constant_boolean_node
12274 instead of boolean_{true,false}_node.
12276 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12278 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12281 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12283 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12284 Simplify the clause that sets the length attribute.
12285 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12286 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12287 clause that sets the length attribute.
12288 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12290 2018-01-24 Tom de Vries <tom@codesourcery.com>
12293 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12294 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12295 Add strict parameter.
12296 (prevent_branch_around_nothing): Insert dummy insn between branch to
12297 label and label with no ptx insn inbetween.
12298 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12300 2018-01-24 Tom de Vries <tom@codesourcery.com>
12303 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12304 for neutered threads in warp.
12305 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12307 2018-01-24 Richard Biener <rguenther@suse.de>
12309 PR tree-optimization/83176
12310 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12313 2018-01-24 Richard Biener <rguenther@suse.de>
12315 PR tree-optimization/82819
12316 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12317 code generating pluses that are no-ops in the target precision.
12319 2018-01-24 Richard Biener <rguenther@suse.de>
12321 PR middle-end/84000
12322 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12324 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12326 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12327 to merge probabilities.
12328 * predict.c (probably_never_executed): Also mark as cold functions
12329 with global 0 profile and guessed local profile.
12330 * profile-count.c (profile_probability::combine_with_count): New
12332 * profile-count.h (profile_probability::operator*,
12333 profile_probability::operator*=, profile_probability::operator/,
12334 profile_probability::operator/=): Reduce precision to adjusted
12335 and set value to guessed on contradictory divisions.
12336 (profile_probability::combine_with_freq): Remove.
12337 (profile_probability::combine_wiht_count): Declare.
12338 (profile_count::force_nonzero):: Set to adjusted.
12339 (profile_count::probability_in):: Set quality to adjusted.
12340 * tree-ssa-tail-merge.c (replace_block_by): Use
12341 combine_with_count.
12343 2018-01-23 Andrew Waterman <andrew@sifive.com>
12344 Jim Wilson <jimw@sifive.com>
12346 * config/riscv/riscv.c (riscv_stack_boundary): New.
12347 (riscv_option_override): Set riscv_stack_boundary. Handle
12348 riscv_preferred_stack_boundary_arg.
12349 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12350 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12351 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12352 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12353 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12354 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12356 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12359 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12360 of struct ix86_frame.
12361 (ix86_expand_epilogue): Likewise. Add a local variable for
12362 the reg_save_offset field in struct ix86_frame.
12364 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12366 PR tree-optimization/82604
12367 * tree-loop-distribution.c (enum partition_kind): New enum item
12368 PKIND_PARTIAL_MEMSET.
12369 (partition_builtin_p): Support above new enum item.
12370 (generate_code_for_partition): Ditto.
12371 (compute_access_range): Differentiate cases that equality can be
12372 proven at all loops, the innermost loops or no loops.
12373 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12374 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12375 (finalize_partitions, distribute_loop): Don't fuse partition of
12376 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12377 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12378 parloop is enabled.
12380 2018-01-23 Martin Liska <mliska@suse.cz>
12382 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12383 order to ignore the predictor.
12384 (PRED_POLYMORPHIC_CALL): Likewise.
12385 (PRED_RECURSIVE_CALL): Likewise.
12387 2018-01-23 Martin Liska <mliska@suse.cz>
12389 * tree-profile.c (tree_profiling): Print function header to
12390 aware reader which function we are working on.
12391 * value-prof.c (gimple_find_values_to_profile): Do not print
12392 not interesting value histograms.
12394 2018-01-23 Martin Liska <mliska@suse.cz>
12396 * profile-count.h (enum profile_quality): Add
12397 profile_uninitialized as the first value. Do not number values
12398 as they are zero based.
12399 (profile_count::verify): Update sanity check.
12400 (profile_probability::verify): Likewise.
12402 2018-01-23 Nathan Sidwell <nathan@acm.org>
12404 * doc/invoke.texi (ffor-scope): Deprecate.
12406 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12408 PR tree-optimization/83510
12409 * domwalk.c (set_all_edges_as_executable): New function.
12410 (dom_walker::dom_walker): Convert bool param
12411 "skip_unreachable_blocks" to enum reachability. Move setup of
12412 edge flags to set_all_edges_as_executable and only do it when
12413 reachability is REACHABLE_BLOCKS.
12414 * domwalk.h (enum dom_walker::reachability): New enum.
12415 (dom_walker::dom_walker): Convert bool param
12416 "skip_unreachable_blocks" to enum reachability.
12417 (set_all_edges_as_executable): New decl.
12418 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12419 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12421 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12422 but converting true to REACHABLE_BLOCKS.
12423 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12425 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12426 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12427 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12429 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12430 if check_all_array_refs will be called.
12432 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12434 * tree.c (selftest::test_location_wrappers): Add more test
12437 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12439 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12440 (selftest::test_bit_in_range): Likewise.
12442 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12445 * doc/sourcebuild.texi (vect_float): Say that the selector
12446 only describes the situation when -funsafe-math-optimizations is on.
12447 (vect_float_strict): Document.
12449 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12451 PR tree-optimization/83965
12452 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12453 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12454 instead of checking only for a reduction.
12455 (vect_recog_widen_sum_pattern): Likewise.
12457 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12459 * predict.c (probably_never_executed): Only use precise profile info.
12460 (compute_function_frequency): Skip after inlining hack since we now
12461 have quality checking.
12463 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12465 * profile-count.h (profile_probability::very_unlikely,
12466 profile_probability::unlikely, profile_probability::even): Set
12467 precision to guessed.
12469 2018-01-23 Richard Biener <rguenther@suse.de>
12471 PR tree-optimization/83963
12472 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12473 Properly terminate dominator walk when crossing the exit edge not
12474 when visiting its source block.
12476 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12479 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12480 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12482 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12484 PR tree-optimization/83957
12485 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12486 semicolon after for body surrounded by braces.
12488 PR tree-optimization/83081
12489 * profile-count.h (profile_probability::split): New method.
12490 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12491 Use profile_probability::split.
12492 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12493 when splitting a single conditional jump into 2.
12495 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12497 PR tree-optimization/69452
12498 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12501 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12503 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12504 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12505 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12507 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12509 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12511 * config/rl78/rl78.md (movdi): New define_expand.
12512 * config/rl78/rl78.c (rl78_split_movdi): New function.
12514 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12517 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12519 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12520 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12521 128-bit to produce an UNSPEC move to get the double word with the
12522 signbit and then a shift directly to do signbit.
12523 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12524 implementation with a new version that just does either a direct
12525 move or a regular move. Move memory interface to separate insns.
12526 Move insns so they are next to the expander.
12527 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12528 with signbit move. Split big and little endian case.
12529 (signbit<mode>2_dm_mem_le): Likewise.
12530 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12531 (signbit<mode>2_dm2): Likewise.
12533 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12535 * config/rl78/rl78.md (anddi3): New define_expand.
12537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12539 * config/rl78/rl78.md (umindi3): New define_expand.
12541 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12543 * config/rl78/rl78.md (smindi3): New define_expand.
12545 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12547 * config/rl78/rl78.md (smaxdi3): New define_expand.
12549 2018-01-22 Carl Love <cel@us.ibm.com>
12551 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12552 LVX_V1TI): Add macro expansion.
12553 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12554 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12555 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12556 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12557 Change check to determine if the instruction is a byte reversing
12558 entry. Fix typo in comment.
12559 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12560 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12561 Add def_builtin calls for new builtins.
12562 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12563 Add define_insn expansion.
12565 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12567 * config/rl78/rl78.md (umaxdi3): New define_expand.
12569 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12571 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12572 for non-QImode registers.
12574 2018-01-22 Richard Biener <rguenther@suse.de>
12576 PR tree-optimization/83963
12577 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12578 including the loop exit block.
12579 (scop_detection::merge_sese): Likewise.
12580 (scop_detection::add_scop): Do it here instead.
12582 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12584 * doc/sourcebuild.texi (arm_softfloat): Document.
12586 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12589 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12590 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12591 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12593 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12594 David Edelsohn <dje.gcc@gmail.com>
12597 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12598 Change "crset eq" to "crset 2".
12599 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12600 (*call_indirect_aix<mode>_nospec): Likewise.
12601 (*call_value_indirect_aix<mode>_nospec): Likewise.
12602 (*call_indirect_elfv2<mode>_nospec): Likewise.
12603 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12604 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12605 change assembly output from . to $.
12606 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12607 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12608 (*tablejump<mode>_internal1_nospec): Likewise.
12610 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12613 * config/sh/sh_optimize_sett_clrt.cc:
12614 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12616 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12618 PR tree-optimization/83940
12619 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12620 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12621 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12622 use it to pass back the definition type.
12623 (vect_check_store_rhs): Likewise rhs_dt_out.
12624 (vect_build_gather_load_calls): Add a mask_dt argument and use
12625 it instead of a call to vect_is_simple_use.
12626 (vectorizable_store): Update calls to vect_check_load_store_mask
12627 and vect_check_store_rhs. Use the dt returned by the latter instead
12628 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12629 instead of calls to vect_is_simple_use. Pass the scalar rather
12630 than the vector operand to vect_is_simple_use when handling
12631 second and subsequent copies of an rhs value.
12632 (vectorizable_load): Update calls to vect_check_load_store_mask
12633 and vect_build_gather_load_calls. Use the cached mask_dt and
12634 gs_info.offset_dt instead of calls to vect_is_simple_use.
12636 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12638 PR middle-end/83945
12639 * tree-emutls.c: Include gimplify.h.
12640 (lower_emutls_2): New function.
12641 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12642 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12643 it before further processing.
12646 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12647 UINTVAL (trueop1) instead of INTVAL (op1).
12649 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12653 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12654 INCOMING_FRAME_SP_OFFSET if not defined.
12655 (scan_trace): Add ENTRY argument. If true and
12656 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12657 emit a note to adjust the CFA offset.
12658 (create_cfi_notes): Adjust scan_trace callers.
12659 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12660 INCOMING_FRAME_SP_OFFSET in the CIE.
12661 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12662 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12664 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12665 * doc/tm.texi: Regenerated.
12667 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12669 PR rtl-optimization/83147
12670 * lra-constraints.c (remove_inheritance_pseudos): Use
12671 lra_substitute_pseudo_within_insn.
12673 2018-01-19 Tom de Vries <tom@codesourcery.com>
12674 Cesar Philippidis <cesar@codesourcery.com>
12677 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12679 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12682 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12683 spaces for function labels.
12685 2018-01-19 Martin Liska <mliska@suse.cz>
12687 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12688 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12689 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12690 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12691 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12692 (PRED_CONST_RETURN): Change from 69 to 65.
12693 (PRED_NULL_RETURN): Change from 91 to 71.
12694 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12695 (PRED_LOOP_GUARD): Change from 66 to 73.
12697 2018-01-19 Martin Liska <mliska@suse.cz>
12699 * predict.c (predict_insn_def): Add new assert.
12700 (struct branch_predictor): Change type to signed integer.
12701 (test_prediction_value_range): Amend test to cover
12702 PROB_UNINITIALIZED.
12703 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12704 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12705 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12706 (PRED_LOOP_IV_COMPARE): Likewise.
12707 * predict.h (PROB_UNINITIALIZED): Define new constant.
12709 2018-01-19 Martin Liska <mliska@suse.cz>
12711 * predict.c (dump_prediction): Add new format for
12712 analyze_brprob.py script which is enabled with -details
12714 * profile-count.h (precise_p): New function.
12716 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12718 PR tree-optimization/83922
12719 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12720 there are no statements that need masking.
12721 (vect_active_double_reduction_p): New function.
12722 (vect_analyze_loop_operations): Use it when handling phis that
12723 are not in the loop header.
12725 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12727 PR tree-optimization/83914
12728 * tree-vect-loop.c (vectorizable_induction): Don't convert
12729 init_expr or apply the peeling adjustment for inductions
12730 that are nested within the vectorized loop.
12732 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12734 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12737 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12741 * function.h (gimplify_parameters): Add gimple_seq * argument.
12742 * function.c: Include gimple.h and options.h.
12743 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12744 for the added local temporaries if needed.
12745 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12746 if there are any parameter cleanups, wrap whole body into a
12747 try/finally with the cleanups.
12749 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12752 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12753 Use GET_MODE_CLASS for scalar floating point.
12755 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12759 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12760 Fix call of call_cgraph_insertion_hooks.
12762 2018-01-18 Martin Sebor <msebor@redhat.com>
12764 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12766 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12769 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12772 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12775 * common.opt: (-ffile-prefix-map): New option.
12776 * opts.c (common_handle_option): Defer it.
12777 * opts-global.c (handle_common_deferred_options): Handle it.
12778 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12779 * file-prefix-map.h: New file.
12780 (remap_debug_filename, add_debug_prefix_map): ...here.
12781 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12782 * final.c (debug_prefix_map, add_debug_prefix_map
12783 remap_debug_filename): Move to...
12784 * file-prefix-map.c: New file.
12785 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12786 generalize, get rid of alloca(), use strrchr() instead of strchr().
12787 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12788 Implement in terms of add_prefix_map().
12789 (remap_macro_filename, remap_debug_filename): Implement in term of
12791 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12792 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12793 * dbxout.c: Include file-prefix-map.h.
12794 * varasm.c: Likewise.
12795 * vmsdbgout.c: Likewise.
12796 * xcoffout.c: Likewise.
12797 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12798 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12799 * doc/invoke.texi (-ffile-prefix-map): Document.
12800 (-fdebug-prefix-map): Update description.
12802 2018-01-18 Martin Liska <mliska@suse.cz>
12804 * config/i386/i386.c (indirect_thunk_name): Document that also
12806 (output_indirect_thunk): Document why both instructions
12807 (pause and lfence) are generated.
12809 2018-01-18 Richard Biener <rguenther@suse.de>
12811 PR tree-optimization/83887
12812 * graphite-scop-detection.c
12813 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12814 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12815 (scop_detection::merge_sese): Re-implement with a flood-fill
12816 algorithm that properly finds a SESE region if it exists.
12818 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12821 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12822 pointer_diff optimizations use view_convert instead of convert.
12824 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12826 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12827 Generate different code for -mno-speculate-indirect-jumps.
12828 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12829 (*call_indirect_aix<mode>): Disable for
12830 -mno-speculate-indirect-jumps.
12831 (*call_indirect_aix<mode>_nospec): New define_insn.
12832 (*call_value_indirect_aix<mode>): Disable for
12833 -mno-speculate-indirect-jumps.
12834 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12835 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12836 -mno-speculate-indirect-jumps.
12837 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12839 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12841 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12842 long double type, set the flags for noting the default long double
12843 type, even if we don't pass or return a long double type.
12845 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12848 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12851 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12853 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12854 support for merge[hl].
12855 (fold_mergehl_helper): New helper function.
12856 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12857 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12858 (altivec_vmrglw_direct): Add xxmrglw insn.
12860 2018-01-17 Andrew Waterman <andrew@sifive.com>
12862 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12863 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12865 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12868 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12869 call the lto_location_cache before reading the
12870 DECL_SOURCE_LOCATION of the types.
12872 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12873 Richard Sandiford <richard.sandiford@linaro.org>
12875 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12876 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12877 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12878 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12879 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12881 * config/aarch64/constraints.md (aarch64_movti_operand):
12883 * config/aarch64/predicates.md (Uti): Add new constraint.
12885 2018-01-17 Carl Love <cel@us.ibm.com>
12887 * config/rs6000/vsx.md (define_expand xl_len_r,
12888 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12889 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12891 (define_expand, define_insn): Move the shift left from the
12892 define_insn to the define_expand for lxvl and stxvl instructions.
12893 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12894 and XL_LEN_R definitions to PURE.
12896 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12898 * config/i386/i386.c (indirect_thunk_name): Declare regno
12899 as unsigned int. Compare regno with INVALID_REGNUM.
12900 (output_indirect_thunk): Ditto.
12901 (output_indirect_thunk_function): Ditto.
12902 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12903 in the call to output_indirect_thunk_function.
12905 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12907 PR middle-end/83884
12908 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12909 rather than the size of inner_type to determine the stack slot size
12910 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12912 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12915 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12918 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12920 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12921 endian Linux systems to optionally enable multilibs for selecting
12922 the long double type if the user configured an explicit type.
12923 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12924 have no long double multilibs if not defined.
12925 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12926 warn if the user used -mabi={ieee,ibm}longdouble and we built
12927 multilibs for long double.
12928 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12929 appropriate multilib option.
12930 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12932 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12933 for building long double multilibs.
12934 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12936 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12938 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12941 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12943 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12946 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12949 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12952 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12954 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12955 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12957 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12959 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12960 different rtl trees depending on TARGET_64BIT.
12961 (rs6000_gen_lvx): Likewise.
12963 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12965 * config/visium/visium.md (nop): Tweak comment.
12966 (hazard_nop): Likewise.
12968 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12970 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12971 -mspeculate-indirect-jumps.
12972 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12973 for -mno-speculate-indirect-jumps.
12974 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12975 (*call_value_indirect_elfv2<mode>): Disable for
12976 -mno-speculate-indirect-jumps.
12977 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12978 (indirect_jump): Emit different RTL for
12979 -mno-speculate-indirect-jumps.
12980 (*indirect_jump<mode>): Disable for
12981 -mno-speculate-indirect-jumps.
12982 (*indirect_jump<mode>_nospec): New define_insn.
12983 (tablejump): Emit different RTL for
12984 -mno-speculate-indirect-jumps.
12985 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12986 (tablejumpsi_nospec): New define_expand.
12987 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12988 (tablejumpdi_nospec): New define_expand.
12989 (*tablejump<mode>_internal1): Disable for
12990 -mno-speculate-indirect-jumps.
12991 (*tablejump<mode>_internal1_nospec): New define_insn.
12992 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12995 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12997 * caller-save.c (insert_save): Drop unnecessary parameter. All
13000 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13001 Richard Biener <rguenth@suse.de>
13004 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
13005 return early, inline manually is_gimple_sizepos. Make sure if we
13006 call gimplify_expr we don't end up with a gimple constant.
13007 * tree.c (variably_modified_type_p): Don't return true for
13008 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
13009 * gimplify.h (is_gimple_sizepos): Remove.
13011 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13013 PR tree-optimization/83857
13014 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
13015 vectorizable_live_operation for pure SLP statements.
13016 (vectorizable_live_operation): Handle PHIs.
13018 2018-01-16 Richard Biener <rguenther@suse.de>
13020 PR tree-optimization/83867
13021 * tree-vect-stmts.c (vect_transform_stmt): Precompute
13022 nested_in_vect_loop_p since the scalar stmt may get invalidated.
13024 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13027 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
13028 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
13029 If off is not INTEGER_CST, issue a may not be aligned warning
13030 rather than isn't aligned. Use isn%'t rather than isn't.
13031 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
13033 <case MULT_EXPR>: Improve the case when bottom and one of the
13034 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
13035 operand, in that case check if the other operand is multiple of
13036 bottom divided by the INTEGER_CST operand.
13038 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13041 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
13042 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
13043 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
13044 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
13045 * config/pa/pa.c (pa_function_arg_advance): Likewise.
13046 (pa_function_arg, pa_arg_partial_bytes): Likewise.
13047 (pa_function_arg_size): New function.
13049 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13051 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
13052 in a separate statement.
13054 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
13056 PR tree-optimization/83847
13057 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
13058 group gathers and scatters.
13060 2018-01-16 Jakub Jelinek <jakub@redhat.com>
13062 PR rtl-optimization/86620
13063 * params.def (max-sched-ready-insns): Bump minimum value to 1.
13065 PR rtl-optimization/83213
13066 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
13067 to last if both are JUMP_INSNs.
13069 PR tree-optimization/83843
13070 * gimple-ssa-store-merging.c
13071 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
13072 store_immediate_info for bswap/nop orig_stores.
13074 2018-01-15 Andrew Waterman <andrew@sifive.com>
13076 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
13078 <UDIV>: Increase cost if !TARGET_DIV.
13080 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
13082 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
13083 (define_attr "cr_logical_3op"): New.
13084 (cceq_ior_compare): Adjust.
13085 (cceq_ior_compare_complement): Adjust.
13086 (*cceq_rev_compare): Adjust.
13087 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
13088 (is_cracked_insn): Adjust.
13089 (insn_must_be_first_in_group): Adjust.
13090 * config/rs6000/40x.md: Adjust.
13091 * config/rs6000/440.md: Adjust.
13092 * config/rs6000/476.md: Adjust.
13093 * config/rs6000/601.md: Adjust.
13094 * config/rs6000/603.md: Adjust.
13095 * config/rs6000/6xx.md: Adjust.
13096 * config/rs6000/7450.md: Adjust.
13097 * config/rs6000/7xx.md: Adjust.
13098 * config/rs6000/8540.md: Adjust.
13099 * config/rs6000/cell.md: Adjust.
13100 * config/rs6000/e300c2c3.md: Adjust.
13101 * config/rs6000/e500mc.md: Adjust.
13102 * config/rs6000/e500mc64.md: Adjust.
13103 * config/rs6000/e5500.md: Adjust.
13104 * config/rs6000/e6500.md: Adjust.
13105 * config/rs6000/mpc.md: Adjust.
13106 * config/rs6000/power4.md: Adjust.
13107 * config/rs6000/power5.md: Adjust.
13108 * config/rs6000/power6.md: Adjust.
13109 * config/rs6000/power7.md: Adjust.
13110 * config/rs6000/power8.md: Adjust.
13111 * config/rs6000/power9.md: Adjust.
13112 * config/rs6000/rs64.md: Adjust.
13113 * config/rs6000/titan.md: Adjust.
13115 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13117 * config/i386/predicates.md (indirect_branch_operand): Rewrite
13118 ix86_indirect_branch_register logic.
13120 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13122 * config/i386/constraints.md (Bs): Update
13123 ix86_indirect_branch_register check. Don't check
13124 ix86_indirect_branch_register with GOT_memory_operand.
13126 * config/i386/predicates.md (GOT_memory_operand): Don't check
13127 ix86_indirect_branch_register here.
13128 (GOT32_symbol_operand): Likewise.
13130 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13132 * config/i386/predicates.md (constant_call_address_operand):
13133 Rewrite ix86_indirect_branch_register logic.
13134 (sibcall_insn_operand): Likewise.
13136 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13138 * config/i386/constraints.md (Bs): Replace
13139 ix86_indirect_branch_thunk_register with
13140 ix86_indirect_branch_register.
13142 * config/i386/i386.md (indirect_jump): Likewise.
13143 (tablejump): Likewise.
13144 (*sibcall_memory): Likewise.
13145 (*sibcall_value_memory): Likewise.
13146 Peepholes of indirect call and jump via memory: Likewise.
13147 * config/i386/i386.opt: Likewise.
13148 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13149 (GOT_memory_operand): Likewise.
13150 (call_insn_operand): Likewise.
13151 (sibcall_insn_operand): Likewise.
13152 (GOT32_symbol_operand): Likewise.
13154 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13156 PR middle-end/83837
13157 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13158 type rather than type addr's type points to.
13159 (expand_omp_atomic_mutex): Likewise.
13160 (expand_omp_atomic): Likewise.
13162 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13165 * config/i386/i386.c (output_indirect_thunk_function): Use
13166 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13167 for __x86_return_thunk.
13169 2018-01-15 Richard Biener <rguenther@suse.de>
13171 PR middle-end/83850
13172 * expmed.c (extract_bit_field_1): Fix typo.
13174 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13177 * config/arm/iterators.md (VF): New mode iterator.
13178 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13179 Remove integer-related logic from pattern.
13180 (neon_vabd<mode>_3): Likewise.
13182 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13184 PR middle-end/82694
13185 * common.opt (fstrict-overflow): No longer an alias.
13186 (fwrapv-pointer): New option.
13187 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13188 also for pointer types based on flag_wrapv_pointer.
13189 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13190 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13191 opts->x_flag_wrapv got set.
13192 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13193 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13194 POINTER_TYPE_OVERFLOW_UNDEFINED.
13195 * match.pd: Likewise in address comparison pattern.
13196 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13198 2018-01-15 Richard Biener <rguenther@suse.de>
13201 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13202 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13203 Reset type names to their identifier if their TYPE_DECL doesn't
13204 have linkage (and thus is used for ODR and devirt).
13205 (save_debug_info_for_decl): Remove.
13206 (save_debug_info_for_type): Likewise.
13207 (add_tree_to_fld_list): Adjust.
13208 * tree-pretty-print.c (dump_generic_node): Make dumping of
13209 type names more robust.
13211 2018-01-15 Richard Biener <rguenther@suse.de>
13213 * BASE-VER: Bump to 8.0.1.
13215 2018-01-14 Martin Sebor <msebor@redhat.com>
13218 * builtins.c (check_access): Avoid warning when the no-warning bit
13221 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13223 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13224 * ira-color (allocno_hard_regs_compare): Likewise.
13226 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13229 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13230 Use .pushsection/.popsection.
13232 2018-01-14 Martin Sebor <msebor@redhat.com>
13235 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13237 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13239 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13240 entry from extra_headers.
13241 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13242 extra_headers, make the list bitwise identical to the i?86-*-* one.
13244 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13246 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13247 -mcmodel=large with -mindirect-branch=thunk,
13248 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13249 -mfunction-return=thunk-extern.
13250 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13251 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13252 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13254 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13256 * config/i386/i386.c (print_reg): Print the name of the full
13257 integer register without '%'.
13258 (ix86_print_operand): Handle 'V'.
13259 * doc/extend.texi: Document 'V' modifier.
13261 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13263 * config/i386/constraints.md (Bs): Disallow memory operand for
13264 -mindirect-branch-register.
13266 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13267 (GOT_memory_operand): Likewise.
13268 (call_insn_operand): Likewise.
13269 (sibcall_insn_operand): Likewise.
13270 (GOT32_symbol_operand): Likewise.
13271 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13272 for -mindirect-branch-register.
13273 (tablejump): Likewise.
13274 (*sibcall_memory): Likewise.
13275 (*sibcall_value_memory): Likewise.
13276 Disallow peepholes of indirect call and jump via memory for
13277 -mindirect-branch-register.
13278 (*call_pop): Replace m with Bw.
13279 (*call_value_pop): Likewise.
13280 (*sibcall_pop_memory): Replace m with Bs.
13281 * config/i386/i386.opt (mindirect-branch-register): New option.
13282 * doc/invoke.texi: Document -mindirect-branch-register option.
13284 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13286 * config/i386/i386-protos.h (ix86_output_function_return): New.
13287 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13288 set function_return_type.
13289 (indirect_thunk_name): Add ret_p to indicate thunk for function
13291 (output_indirect_thunk_function): Pass false to
13292 indirect_thunk_name.
13293 (ix86_output_indirect_branch_via_reg): Likewise.
13294 (ix86_output_indirect_branch_via_push): Likewise.
13295 (output_indirect_thunk_function): Create alias for function
13296 return thunk if regno < 0.
13297 (ix86_output_function_return): New function.
13298 (ix86_handle_fndecl_attribute): Handle function_return.
13299 (ix86_attribute_table): Add function_return.
13300 * config/i386/i386.h (machine_function): Add
13301 function_return_type.
13302 * config/i386/i386.md (simple_return_internal): Use
13303 ix86_output_function_return.
13304 (simple_return_internal_long): Likewise.
13305 * config/i386/i386.opt (mfunction-return=): New option.
13306 (indirect_branch): Mention -mfunction-return=.
13307 * doc/extend.texi: Document function_return function attribute.
13308 * doc/invoke.texi: Document -mfunction-return= option.
13310 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13312 * config/i386/i386-opts.h (indirect_branch): New.
13313 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13314 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13315 with local indirect jump when converting indirect call and jump.
13316 (ix86_set_indirect_branch_type): New.
13317 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13318 (indirectlabelno): New.
13319 (indirect_thunk_needed): Likewise.
13320 (indirect_thunk_bnd_needed): Likewise.
13321 (indirect_thunks_used): Likewise.
13322 (indirect_thunks_bnd_used): Likewise.
13323 (INDIRECT_LABEL): Likewise.
13324 (indirect_thunk_name): Likewise.
13325 (output_indirect_thunk): Likewise.
13326 (output_indirect_thunk_function): Likewise.
13327 (ix86_output_indirect_branch_via_reg): Likewise.
13328 (ix86_output_indirect_branch_via_push): Likewise.
13329 (ix86_output_indirect_branch): Likewise.
13330 (ix86_output_indirect_jmp): Likewise.
13331 (ix86_code_end): Call output_indirect_thunk_function if needed.
13332 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13334 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13335 (ix86_attribute_table): Add indirect_branch.
13336 * config/i386/i386.h (machine_function): Add indirect_branch_type
13337 and has_local_indirect_jump.
13338 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13340 (tablejump): Likewise.
13341 (*indirect_jump): Use ix86_output_indirect_jmp.
13342 (*tablejump_1): Likewise.
13343 (simple_return_indirect_internal): Likewise.
13344 * config/i386/i386.opt (mindirect-branch=): New option.
13345 (indirect_branch): New.
13348 (thunk-inline): Likewise.
13349 (thunk-extern): Likewise.
13350 * doc/extend.texi: Document indirect_branch function attribute.
13351 * doc/invoke.texi: Document -mindirect-branch= option.
13353 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13356 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13358 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13360 * ipa-inline.c (want_inline_small_function_p): Return false if
13361 inlining has already failed with CIF_FINAL_ERROR.
13362 (update_caller_keys): Call want_inline_small_function_p before
13364 (update_callee_keys): Likewise.
13366 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13368 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13370 (rs6000_quadword_masked_address_p): Likewise.
13371 (quad_aligned_load_p): Likewise.
13372 (quad_aligned_store_p): Likewise.
13373 (const_load_sequence_p): Add comment to describe the outer-most loop.
13374 (mimic_memory_attributes_and_flags): New function.
13375 (rs6000_gen_stvx): Likewise.
13376 (replace_swapped_aligned_store): Likewise.
13377 (rs6000_gen_lvx): Likewise.
13378 (replace_swapped_aligned_load): Likewise.
13379 (replace_swapped_load_constant): Capitalize argument name in
13380 comment describing this function.
13381 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13382 and stores that access quad-word aligned addresses and replace
13383 with stvx or lvx instructions when appropriate.
13384 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13385 New function prototype.
13386 (rs6000_quadword_masked_address_p): Likewise.
13387 (rs6000_gen_lvx): Likewise.
13388 (rs6000_gen_stvx): Likewise.
13389 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13390 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13391 when memory address is aligned.
13392 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13393 this split to select lvx instruction when memory address is aligned.
13394 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13395 instruction when memory address is aligned.
13396 (*vsx_le_perm_load_v16qi): Likewise.
13397 (four unnamed splitters): Modify to select the stvx instruction
13398 when memory is aligned.
13400 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13402 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13403 which appears in the queue multiple times.
13405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13406 Alan Hayward <alan.hayward@arm.com>
13407 David Sherwood <david.sherwood@arm.com>
13409 * tree-vectorizer.h (vec_lower_bound): New structure.
13410 (_loop_vec_info): Add check_nonzero and lower_bounds.
13411 (LOOP_VINFO_CHECK_NONZERO): New macro.
13412 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13413 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13414 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13415 fields. Make seg_len the distance travelled, not including the
13417 (dr_direction_indicator): Declare.
13418 (dr_zero_step_indicator): Likewise.
13419 (dr_known_forward_stride_p): Likewise.
13420 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13422 (runtime_alias_check_p): Allow runtime alias checks with
13424 (operator ==): Compare access_size and align.
13425 (prune_runtime_alias_test_list): Rework for new distinction between
13426 the access_size and seg_len.
13427 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13429 (get_segment_min_max): New function.
13430 (create_intersect_range_checks): Use it.
13431 (dr_step_indicator): New function.
13432 (dr_direction_indicator): Likewise.
13433 (dr_zero_step_indicator): Likewise.
13434 (dr_known_forward_stride_p): Likewise.
13435 * tree-loop-distribution.c (data_ref_segment_size): Return
13436 DR_STEP * (niters - 1).
13437 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13439 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13440 (vect_preserves_scalar_order_p): New function, split out from...
13441 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13442 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13443 (vect_vfa_access_size): New function.
13444 (vect_vfa_align): Likewise.
13445 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13446 (dump_lower_bound): New function.
13447 (vect_check_lower_bound): Likewise.
13448 (vect_small_gap_p): Likewise.
13449 (vectorizable_with_step_bound_p): Likewise.
13450 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13451 depencies if the vectorization factor is 1. Convert the checks
13452 for nonzero steps into checks on the bounds of DR_STEP. Try using
13453 a bunds check for variable steps if the minimum required step is
13454 relatively small. Update calls to the dr_with_seg_len
13455 constructor and to vect_compile_time_alias.
13456 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13458 (vect_loop_versioning): Call it.
13459 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13461 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13464 Alan Hayward <alan.hayward@arm.com>
13465 David Sherwood <david.sherwood@arm.com>
13467 * doc/sourcebuild.texi (vect_scatter_store): Document.
13468 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13470 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13472 * genopinit.c (main): Add supports_vec_scatter_store and
13473 supports_vec_scatter_store_cached to target_optabs.
13474 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13475 IFN_MASK_SCATTER_STORE.
13476 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13478 * internal-fn.h (internal_store_fn_p): Declare.
13479 (internal_fn_stored_value_index): Likewise.
13480 * internal-fn.c (scatter_store_direct): New macro.
13481 (expand_scatter_store_optab_fn): New function.
13482 (direct_scatter_store_optab_supported_p): New macro.
13483 (internal_store_fn_p): New function.
13484 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13485 IFN_MASK_SCATTER_STORE.
13486 (internal_fn_mask_index): Likewise.
13487 (internal_fn_stored_value_index): New function.
13488 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13489 for scatter stores.
13490 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13491 * optabs-query.c (supports_vec_scatter_store_p): New function.
13492 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13493 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13494 true for scatter stores.
13495 (vect_gather_scatter_fn_p): Handle scatter stores too.
13496 (vect_check_gather_scatter): Consider using scatter stores if
13497 supports_vec_scatter_store_p.
13498 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13499 scatter stores too.
13500 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13501 internal_fn_stored_value_index.
13502 (check_load_store_masking): Handle scatter stores too.
13503 (vect_get_store_rhs): Make public.
13504 (vectorizable_call): Use internal_store_fn_p.
13505 (vectorizable_store): Handle scatter store internal functions.
13506 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13507 when deciding whether the end of the group has been reached.
13508 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13509 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13510 (mask_scatter_store<mode>): New insns.
13512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13513 Alan Hayward <alan.hayward@arm.com>
13514 David Sherwood <david.sherwood@arm.com>
13516 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13517 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13518 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13520 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13521 Use vect_truncate_gather_scatter_offset if we can't treat the
13522 operation as a normal gather load or scatter store.
13523 (get_group_load_store_type): Take the gather_scatter_info
13524 as argument. Try using a gather load or scatter store for
13525 single-element groups.
13526 (get_load_store_type): Update calls to get_group_load_store_type
13527 and vect_use_strided_gather_scatters_p.
13529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13530 Alan Hayward <alan.hayward@arm.com>
13531 David Sherwood <david.sherwood@arm.com>
13533 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13534 optional tree argument.
13535 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13537 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13538 but continue to use the current value as a fallback.
13539 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13540 to compare the updates.
13541 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13542 (get_load_store_type): Use it when handling a strided access.
13543 (vect_get_strided_load_store_ops): New function.
13544 (vect_get_data_ptr_increment): Likewise.
13545 (vectorizable_load): Handle strided gather loads. Always pass
13546 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13549 Alan Hayward <alan.hayward@arm.com>
13550 David Sherwood <david.sherwood@arm.com>
13552 * doc/md.texi (gather_load@var{m}): Document.
13553 (mask_gather_load@var{m}): Likewise.
13554 * genopinit.c (main): Add supports_vec_gather_load and
13555 supports_vec_gather_load_cached to target_optabs.
13556 * optabs-tree.c (init_tree_optimization_optabs): Use
13557 ggc_cleared_alloc to allocate target_optabs.
13558 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13559 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13561 * internal-fn.h (internal_load_fn_p): Declare.
13562 (internal_gather_scatter_fn_p): Likewise.
13563 (internal_fn_mask_index): Likewise.
13564 (internal_gather_scatter_fn_supported_p): Likewise.
13565 * internal-fn.c (gather_load_direct): New macro.
13566 (expand_gather_load_optab_fn): New function.
13567 (direct_gather_load_optab_supported_p): New macro.
13568 (direct_internal_fn_optab): New function.
13569 (internal_load_fn_p): Likewise.
13570 (internal_gather_scatter_fn_p): Likewise.
13571 (internal_fn_mask_index): Likewise.
13572 (internal_gather_scatter_fn_supported_p): Likewise.
13573 * optabs-query.c (supports_at_least_one_mode_p): New function.
13574 (supports_vec_gather_load_p): Likewise.
13575 * optabs-query.h (supports_vec_gather_load_p): Declare.
13576 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13577 and memory_type field.
13578 (NUM_PATTERNS): Bump to 15.
13579 * tree-vect-data-refs.c: Include internal-fn.h.
13580 (vect_gather_scatter_fn_p): New function.
13581 (vect_describe_gather_scatter_call): Likewise.
13582 (vect_check_gather_scatter): Try using internal functions for
13583 gather loads. Recognize existing calls to a gather load function.
13584 (vect_analyze_data_refs): Consider using gather loads if
13585 supports_vec_gather_load_p.
13586 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13587 (vect_get_gather_scatter_offset_type): Likewise.
13588 (vect_convert_mask_for_vectype): Likewise.
13589 (vect_add_conversion_to_patterm): Likewise.
13590 (vect_try_gather_scatter_pattern): Likewise.
13591 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13592 (vect_vect_recog_func_ptrs): Add it.
13593 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13594 internal_fn_mask_index and internal_gather_scatter_fn_p.
13595 (check_load_store_masking): Take the gather_scatter_info as an
13596 argument and handle gather loads.
13597 (vect_get_gather_scatter_ops): New function.
13598 (vectorizable_call): Check internal_load_fn_p.
13599 (vectorizable_load): Likewise. Handle gather load internal
13601 (vectorizable_store): Update call to check_load_store_masking.
13602 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13603 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13604 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13605 (aarch64_gather_scale_operand_d): New predicates.
13606 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13607 (mask_gather_load<mode>): New insns.
13609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13610 Alan Hayward <alan.hayward@arm.com>
13611 David Sherwood <david.sherwood@arm.com>
13613 * optabs.def (fold_left_plus_optab): New optab.
13614 * doc/md.texi (fold_left_plus_@var{m}): Document.
13615 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13616 * internal-fn.c (fold_left_direct): Define.
13617 (expand_fold_left_optab_fn): Likewise.
13618 (direct_fold_left_optab_supported_p): Likewise.
13619 * fold-const-call.c (fold_const_fold_left): New function.
13620 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13621 * tree-parloops.c (valid_reduction_p): New function.
13622 (gather_scalar_reductions): Use it.
13623 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13624 (vect_finish_replace_stmt): Declare.
13625 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13626 (needs_fold_left_reduction_p): New function, split out from...
13627 (vect_is_simple_reduction): ...here. Accept reductions that
13628 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13629 (vect_force_simple_reduction): Also store the reduction type in
13630 the assignment's STMT_VINFO_REDUC_TYPE.
13631 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13632 (merge_with_identity): New function.
13633 (vect_expand_fold_left): Likewise.
13634 (vectorize_fold_left_reduction): Likewise.
13635 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13636 scalar phi in place for it. Check for target support and reject
13637 cases that would reassociate the operation. Defer the transform
13638 phase to vectorize_fold_left_reduction.
13639 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13640 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13641 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13645 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13646 call to ifc_temp_var.
13648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13649 Alan Hayward <alan.hayward@arm.com>
13650 David Sherwood <david.sherwood@arm.com>
13652 * target.def (legitimize_address_displacement): Take the original
13653 offset as a poly_int.
13654 * targhooks.h (default_legitimize_address_displacement): Update
13656 * targhooks.c (default_legitimize_address_displacement): Likewise.
13657 * doc/tm.texi: Regenerate.
13658 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13659 as an argument, moving assert of ad->disp == ad->disp_term to...
13660 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13661 Try calling targetm.legitimize_address_displacement before expanding
13662 the address rather than afterwards, and adjust for the new interface.
13663 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13664 Match the new hook interface. Handle SVE addresses.
13665 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13666 new hook interface.
13668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13670 * Makefile.in (OBJS): Add early-remat.o.
13671 * target.def (select_early_remat_modes): New hook.
13672 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13673 * doc/tm.texi: Regenerate.
13674 * targhooks.h (default_select_early_remat_modes): Declare.
13675 * targhooks.c (default_select_early_remat_modes): New function.
13676 * timevar.def (TV_EARLY_REMAT): New timevar.
13677 * passes.def (pass_early_remat): New pass.
13678 * tree-pass.h (make_pass_early_remat): Declare.
13679 * early-remat.c: New file.
13680 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13682 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13685 Alan Hayward <alan.hayward@arm.com>
13686 David Sherwood <david.sherwood@arm.com>
13688 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13689 vfm1 with a bound_epilog parameter.
13690 (vect_do_peeling): Update calls accordingly, and move the prologue
13691 call earlier in the function. Treat the base bound_epilog as 0 for
13692 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13693 this base when peeling for gaps.
13694 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13695 with fully-masked loops.
13696 (vect_estimate_min_profitable_iters): Handle the single peeled
13697 iteration in that case.
13699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13700 Alan Hayward <alan.hayward@arm.com>
13701 David Sherwood <david.sherwood@arm.com>
13703 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13704 single-element interleaving even if the size is not a power of 2.
13705 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13706 accesses for single-element interleaving if the group size is
13709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13710 Alan Hayward <alan.hayward@arm.com>
13711 David Sherwood <david.sherwood@arm.com>
13713 * doc/md.texi (fold_extract_last_@var{m}): Document.
13714 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13715 * optabs.def (fold_extract_last_optab): New optab.
13716 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13717 * internal-fn.c (fold_extract_direct): New macro.
13718 (expand_fold_extract_optab_fn): Likewise.
13719 (direct_fold_extract_optab_supported_p): Likewise.
13720 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13721 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13722 EXTRACT_LAST_REDUCTION.
13723 (get_initial_def_for_reduction): Do not create an initial vector
13724 for EXTRACT_LAST_REDUCTION reductions.
13725 (vectorizable_reduction): Leave the scalar phi in place for
13726 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13727 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13728 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13729 transform phase to vectorizable_condition.
13730 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13732 (vect_finish_stmt_generation): ...here.
13733 (vect_finish_replace_stmt): New function.
13734 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13735 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13737 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13740 Alan Hayward <alan.hayward@arm.com>
13741 David Sherwood <david.sherwood@arm.com>
13743 * doc/md.texi (extract_last_@var{m}): Document.
13744 * optabs.def (extract_last_optab): New optab.
13745 * internal-fn.def (EXTRACT_LAST): New internal function.
13746 * internal-fn.c (cond_unary_direct): New macro.
13747 (expand_cond_unary_optab_fn): Likewise.
13748 (direct_cond_unary_optab_supported_p): Likewise.
13749 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13750 loops using EXTRACT_LAST.
13751 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13752 (extract_last_<mode>): ...this optab.
13753 (vec_extract<mode><Vel>): Update accordingly.
13755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13756 Alan Hayward <alan.hayward@arm.com>
13757 David Sherwood <david.sherwood@arm.com>
13759 * target.def (empty_mask_is_expensive): New hook.
13760 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13761 * doc/tm.texi: Regenerate.
13762 * targhooks.h (default_empty_mask_is_expensive): Declare.
13763 * targhooks.c (default_empty_mask_is_expensive): New function.
13764 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13765 if the target says that empty masks are expensive.
13766 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13768 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13771 Alan Hayward <alan.hayward@arm.com>
13772 David Sherwood <david.sherwood@arm.com>
13774 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13775 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13776 (vect_use_loop_mask_for_alignment_p): New function.
13777 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13778 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13779 niters_skip argument. Make sure that the first niters_skip elements
13780 of the first iteration are inactive.
13781 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13782 Update call to vect_set_loop_masks_directly.
13783 (get_misalign_in_elems): New function, split out from...
13784 (vect_gen_prolog_loop_niters): ...here.
13785 (vect_update_init_of_dr): Take a code argument that specifies whether
13786 the adjustment should be added or subtracted.
13787 (vect_update_init_of_drs): Likewise.
13788 (vect_prepare_for_masked_peels): New function.
13789 (vect_do_peeling): Skip prologue peeling if we're using a mask
13790 instead. Update call to vect_update_inits_of_drs.
13791 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13793 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13794 alignment. Do not include the number of peeled iterations in
13795 the minimum threshold in that case.
13796 (vectorizable_induction): Adjust the start value down by
13797 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13798 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13799 Take the number of skipped iterations into account when calculating
13801 * tree-vect-stmts.c (vect_gen_while_not): New function.
13803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13804 Alan Hayward <alan.hayward@arm.com>
13805 David Sherwood <david.sherwood@arm.com>
13807 * doc/sourcebuild.texi (vect_fully_masked): Document.
13808 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13809 default value to 0.
13810 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13812 (vect_analyze_loop_2): ...here. Don't check the vectorization
13813 factor against the number of loop iterations if the loop is
13816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13817 Alan Hayward <alan.hayward@arm.com>
13818 David Sherwood <david.sherwood@arm.com>
13820 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13821 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13822 (dump_groups): Update accordingly.
13823 (iv_use::mem_type): New member variable.
13824 (address_p): New function.
13825 (record_use): Add a mem_type argument and initialize the new
13827 (record_group_use): Add a mem_type argument. Use address_p.
13828 Remove obsolete null checks of base_object. Update call to record_use.
13829 (find_interesting_uses_op): Update call to record_group_use.
13830 (find_interesting_uses_cond): Likewise.
13831 (find_interesting_uses_address): Likewise.
13832 (get_mem_type_for_internal_fn): New function.
13833 (find_address_like_use): Likewise.
13834 (find_interesting_uses_stmt): Try find_address_like_use before
13835 calling find_interesting_uses_op.
13836 (addr_offset_valid_p): Use the iv mem_type field as the type
13837 of the addressed memory.
13838 (add_autoinc_candidates): Likewise.
13839 (get_address_cost): Likewise.
13840 (split_small_address_groups_p): Use address_p.
13841 (split_address_groups): Likewise.
13842 (add_iv_candidate_for_use): Likewise.
13843 (autoinc_possible_for_pair): Likewise.
13844 (rewrite_groups): Likewise.
13845 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13846 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13847 (get_alias_ptr_type_for_ptr_address): New function.
13848 (rewrite_use_address): Rewrite address uses in calls that were
13849 identified by find_address_like_use.
13851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13852 Alan Hayward <alan.hayward@arm.com>
13853 David Sherwood <david.sherwood@arm.com>
13855 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13857 * gimple-expr.h (is_gimple_addressable: Likewise.
13858 * gimple-expr.c (is_gimple_address): Likewise.
13859 * internal-fn.c (expand_call_mem_ref): New function.
13860 (expand_mask_load_optab_fn): Use it.
13861 (expand_mask_store_optab_fn): Likewise.
13863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13864 Alan Hayward <alan.hayward@arm.com>
13865 David Sherwood <david.sherwood@arm.com>
13867 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13868 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13869 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13870 (cond_umax@var{mode}): Document.
13871 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13872 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13873 (cond_umin_optab, cond_umax_optab): New optabs.
13874 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13875 (COND_IOR, COND_XOR): New internal functions.
13876 * internal-fn.h (get_conditional_internal_fn): Declare.
13877 * internal-fn.c (cond_binary_direct): New macro.
13878 (expand_cond_binary_optab_fn): Likewise.
13879 (direct_cond_binary_optab_supported_p): Likewise.
13880 (get_conditional_internal_fn): New function.
13881 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13882 Cope with reduction statements that are vectorized as calls rather
13884 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13885 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13886 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13887 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13888 (UNSPEC_COND_EOR): New unspecs.
13889 (optab): Add mappings for them.
13890 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13891 (sve_int_op, sve_fp_op): New int attributes.
13893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13894 Alan Hayward <alan.hayward@arm.com>
13895 David Sherwood <david.sherwood@arm.com>
13897 * optabs.def (while_ult_optab): New optab.
13898 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13899 * internal-fn.def (WHILE_ULT): New internal function.
13900 * internal-fn.h (direct_internal_fn_supported_p): New override
13901 that takes two types as argument.
13902 * internal-fn.c (while_direct): New macro.
13903 (expand_while_optab_fn): New function.
13904 (convert_optab_supported_p): Likewise.
13905 (direct_while_optab_supported_p): New macro.
13906 * wide-int.h (wi::udiv_ceil): New function.
13907 * tree-vectorizer.h (rgroup_masks): New structure.
13908 (vec_loop_masks): New typedef.
13909 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13910 and fully_masked_p.
13911 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13912 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13913 (vect_max_vf): New function.
13914 (slpeel_make_loop_iterate_ntimes): Delete.
13915 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13916 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13917 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13918 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13919 internal-fn.h, stor-layout.h and optabs-query.h.
13920 (vect_set_loop_mask): New function.
13921 (add_preheader_seq): Likewise.
13922 (add_header_seq): Likewise.
13923 (interleave_supported_p): Likewise.
13924 (vect_maybe_permute_loop_masks): Likewise.
13925 (vect_set_loop_masks_directly): Likewise.
13926 (vect_set_loop_condition_masked): Likewise.
13927 (vect_set_loop_condition_unmasked): New function, split out from
13928 slpeel_make_loop_iterate_ntimes.
13929 (slpeel_make_loop_iterate_ntimes): Rename to..
13930 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13931 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13932 (vect_do_peeling): Update call accordingly.
13933 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13935 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13936 mask_compare_type, can_fully_mask_p and fully_masked_p.
13937 (release_vec_loop_masks): New function.
13938 (_loop_vec_info): Use it to free the loop masks.
13939 (can_produce_all_loop_masks_p): New function.
13940 (vect_get_max_nscalars_per_iter): Likewise.
13941 (vect_verify_full_masking): Likewise.
13942 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13943 retries, and free the mask rgroups before retrying. Check loop-wide
13944 reasons for disallowing fully-masked loops. Make the final decision
13945 about whether use a fully-masked loop or not.
13946 (vect_estimate_min_profitable_iters): Do not assume that peeling
13947 for the number of iterations will be needed for fully-masked loops.
13948 (vectorizable_reduction): Disable fully-masked loops.
13949 (vectorizable_live_operation): Likewise.
13950 (vect_halve_mask_nunits): New function.
13951 (vect_double_mask_nunits): Likewise.
13952 (vect_record_loop_mask): Likewise.
13953 (vect_get_loop_mask): Likewise.
13954 (vect_transform_loop): Handle the case in which the final loop
13955 iteration might handle a partial vector. Call vect_set_loop_condition
13956 instead of slpeel_make_loop_iterate_ntimes.
13957 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13958 (check_load_store_masking): New function.
13959 (prepare_load_store_mask): Likewise.
13960 (vectorizable_store): Handle fully-masked loops.
13961 (vectorizable_load): Likewise.
13962 (supportable_widening_operation): Use vect_halve_mask_nunits for
13964 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13965 (vect_gen_while): New function.
13966 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13967 (aarch64_uqdec<mode>): New insn.
13969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13970 Alan Hayward <alan.hayward@arm.com>
13971 David Sherwood <david.sherwood@arm.com>
13973 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13974 (reduc_xor_scal_optab): New optabs.
13975 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13976 (reduc_xor_scal_@var{m}): Document.
13977 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13978 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13979 internal functions.
13980 * fold-const-call.c (fold_const_call): Handle them.
13981 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13982 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13983 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13984 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13985 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13986 (UNSPEC_XORV): New unspecs.
13987 (optab): Add entries for them.
13988 (BITWISEV): New int iterator.
13989 (bit_reduc_op): New int attributes.
13991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13992 Alan Hayward <alan.hayward@arm.com>
13993 David Sherwood <david.sherwood@arm.com>
13995 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13996 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13997 * optabs.def (vec_shl_insert_optab): New optab.
13998 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13999 (duplicate_and_interleave): Likewise.
14000 * tree-vect-loop.c: Include internal-fn.h.
14001 (neutral_op_for_slp_reduction): New function, split out from
14002 get_initial_defs_for_reduction.
14003 (get_initial_def_for_reduction): Handle option 2 for variable-length
14004 vectors by loading the neutral value into a vector and then shifting
14005 the initial value into element 0.
14006 (get_initial_defs_for_reduction): Replace the code argument with
14007 the neutral value calculated by neutral_op_for_slp_reduction.
14008 Use gimple_build_vector for constant-length vectors.
14009 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
14010 but the first group_size elements have a neutral value.
14011 Use duplicate_and_interleave otherwise.
14012 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
14013 Update call to get_initial_defs_for_reduction. Handle SLP
14014 reductions for variable-length vectors by creating one vector
14015 result for each scalar result, with the elements associated
14016 with other scalar results stubbed out with the neutral value.
14017 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
14018 Require IFN_VEC_SHL_INSERT for double reductions on
14019 variable-length vectors, or SLP reductions that have
14020 a neutral value. Require can_duplicate_and_interleave_p
14021 support for variable-length unchained SLP reductions if there
14022 is no neutral value, such as for MIN/MAX reductions. Also require
14023 the number of vector elements to be a multiple of the number of
14024 SLP statements when doing variable-length unchained SLP reductions.
14025 Update call to vect_create_epilog_for_reduction.
14026 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
14027 and remove initial values.
14028 (duplicate_and_interleave): Make public.
14029 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
14030 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
14032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14033 Alan Hayward <alan.hayward@arm.com>
14034 David Sherwood <david.sherwood@arm.com>
14036 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
14037 (can_duplicate_and_interleave_p): New function.
14038 (vect_get_and_check_slp_defs): Take the vector of statements
14039 rather than just the current one. Remove excess parentheses.
14040 Restriction rejectinon of vect_constant_def and vect_external_def
14041 for variable-length vectors to boolean types, or types for which
14042 can_duplicate_and_interleave_p is false.
14043 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
14044 (duplicate_and_interleave): New function.
14045 (vect_get_constant_vectors): Use gimple_build_vector for
14046 constant-length vectors and suitable variable-length constant
14047 vectors. Use duplicate_and_interleave for other variable-length
14048 vectors. Don't defer the update when inserting new statements.
14050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14051 Alan Hayward <alan.hayward@arm.com>
14052 David Sherwood <david.sherwood@arm.com>
14054 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
14055 min_profitable_iters doesn't go negative.
14057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14058 Alan Hayward <alan.hayward@arm.com>
14059 David Sherwood <david.sherwood@arm.com>
14061 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
14062 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
14063 * optabs.def (vec_mask_load_lanes_optab): New optab.
14064 (vec_mask_store_lanes_optab): Likewise.
14065 * internal-fn.def (MASK_LOAD_LANES): New internal function.
14066 (MASK_STORE_LANES): Likewise.
14067 * internal-fn.c (mask_load_lanes_direct): New macro.
14068 (mask_store_lanes_direct): Likewise.
14069 (expand_mask_load_optab_fn): Handle masked operations.
14070 (expand_mask_load_lanes_optab_fn): New macro.
14071 (expand_mask_store_optab_fn): Handle masked operations.
14072 (expand_mask_store_lanes_optab_fn): New macro.
14073 (direct_mask_load_lanes_optab_supported_p): Likewise.
14074 (direct_mask_store_lanes_optab_supported_p): Likewise.
14075 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
14077 (vect_load_lanes_supported): Likewise.
14078 * tree-vect-data-refs.c (strip_conversion): New function.
14079 (can_group_stmts_p): Likewise.
14080 (vect_analyze_data_ref_accesses): Use it instead of checking
14081 for a pair of assignments.
14082 (vect_store_lanes_supported): Take a masked_p parameter.
14083 (vect_load_lanes_supported): Likewise.
14084 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
14085 vect_store_lanes_supported and vect_load_lanes_supported.
14086 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14087 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
14088 parameter. Don't allow gaps for masked accesses.
14089 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
14090 and vect_load_lanes_supported.
14091 (get_load_store_type): Take a masked_p parameter and update
14092 call to get_group_load_store_type.
14093 (vectorizable_store): Update call to get_load_store_type.
14094 Handle IFN_MASK_STORE_LANES.
14095 (vectorizable_load): Update call to get_load_store_type.
14096 Handle IFN_MASK_LOAD_LANES.
14098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14099 Alan Hayward <alan.hayward@arm.com>
14100 David Sherwood <david.sherwood@arm.com>
14102 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
14104 * config/aarch64/aarch64-protos.h
14105 (aarch64_sve_struct_memory_operand_p): Declare.
14106 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
14107 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
14108 (VPRED, vpred): Handle SVE structure modes.
14109 * config/aarch64/constraints.md (Utx): New constraint.
14110 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
14111 (aarch64_sve_struct_nonimmediate_operand): New predicates.
14112 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
14113 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
14114 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
14115 structure modes. Split into pieces after RA.
14116 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
14117 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
14119 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
14120 SVE structure modes.
14121 (aarch64_classify_address): Likewise.
14122 (sizetochar): Move earlier in file.
14123 (aarch64_print_operand): Handle SVE register lists.
14124 (aarch64_array_mode): New function.
14125 (aarch64_sve_struct_memory_operand_p): Likewise.
14126 (TARGET_ARRAY_MODE): Redefine.
14128 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14129 Alan Hayward <alan.hayward@arm.com>
14130 David Sherwood <david.sherwood@arm.com>
14132 * target.def (array_mode): New target hook.
14133 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14134 * doc/tm.texi: Regenerate.
14135 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14136 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14137 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14138 targetm.array_mode.
14139 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14143 Alan Hayward <alan.hayward@arm.com>
14144 David Sherwood <david.sherwood@arm.com>
14146 * fold-const.c (fold_binary_loc): Check the argument types
14147 rather than the result type when testing for a vector operation.
14149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14151 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14152 * doc/tm.texi: Regenerate.
14154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14155 Alan Hayward <alan.hayward@arm.com>
14156 David Sherwood <david.sherwood@arm.com>
14158 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14159 (sve): Document new AArch64 extension.
14160 * doc/md.texi (w): Extend the description of the AArch64
14161 constraint to include SVE vectors.
14162 (Upl, Upa): Document new AArch64 predicate constraints.
14163 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14165 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14166 (msve-vector-bits=): New option.
14167 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14168 SVE when these are disabled.
14169 (sve): New extension.
14170 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14171 modes. Adjust their number of units based on aarch64_sve_vg.
14172 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14173 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14174 aarch64_addr_query_type.
14175 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14176 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14177 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14178 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14179 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14180 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14181 (aarch64_simd_imm_zero_p): Delete.
14182 (aarch64_check_zero_based_sve_index_immediate): Declare.
14183 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14184 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14185 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14186 (aarch64_sve_float_mul_immediate_p): Likewise.
14187 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14188 rather than an rtx.
14189 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14190 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14191 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14192 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14193 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14194 (aarch64_regmode_natural_size): Likewise.
14195 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14196 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14198 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14199 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14200 for VG and the SVE predicate registers.
14201 (V_ALIASES): Add a "z"-prefixed alias.
14202 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14203 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14204 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14205 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14206 (REG_CLASS_NAMES): Add entries for them.
14207 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14208 and the predicate registers.
14209 (aarch64_sve_vg): Declare.
14210 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14211 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14212 (REGMODE_NATURAL_SIZE): Define.
14213 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14215 * config/aarch64/aarch64.c: Include cfgrtl.h.
14216 (simd_immediate_info): Add a constructor for series vectors,
14217 and an associated step field.
14218 (aarch64_sve_vg): New variable.
14219 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14220 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14221 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14222 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14223 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14224 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14225 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14226 (aarch64_get_mask_mode): New functions.
14227 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14228 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14229 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14230 predicate modes and predicate registers. Explicitly restrict
14231 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14232 to store a vector mode if it is recognized by
14233 aarch64_classify_vector_mode.
14234 (aarch64_regmode_natural_size): New function.
14235 (aarch64_hard_regno_caller_save_mode): Return the original mode
14237 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14238 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14239 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14240 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14242 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14243 does not overlap dest if the function is frame-related. Handle
14245 (aarch64_split_add_offset): New function.
14246 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14247 them aarch64_add_offset.
14248 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14249 and update call to aarch64_sub_sp.
14250 (aarch64_add_cfa_expression): New function.
14251 (aarch64_expand_prologue): Pass extra temporary registers to the
14252 functions above. Handle the case in which we need to emit new
14253 DW_CFA_expressions for registers that were originally saved
14254 relative to the stack pointer, but now have to be expressed
14255 relative to the frame pointer.
14256 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14258 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14259 IP0 and IP1 values for SVE frames.
14260 (aarch64_expand_vec_series): New function.
14261 (aarch64_expand_sve_widened_duplicate): Likewise.
14262 (aarch64_expand_sve_const_vector): Likewise.
14263 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14264 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14265 into the register, rather than emitting a SET directly.
14266 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14267 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14268 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14269 (offset_9bit_signed_scaled_p): New functions.
14270 (aarch64_replicate_bitmask_imm): New function.
14271 (aarch64_bitmask_imm): Use it.
14272 (aarch64_cannot_force_const_mem): Reject expressions involving
14273 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14274 (aarch64_classify_index): Handle SVE indices, by requiring
14275 a plain register index with a scale that matches the element size.
14276 (aarch64_classify_address): Handle SVE addresses. Assert that
14277 the mode of the address is VOIDmode or an integer mode.
14278 Update call to aarch64_classify_symbol.
14279 (aarch64_classify_symbolic_expression): Update call to
14280 aarch64_classify_symbol.
14281 (aarch64_const_vec_all_in_range_p): New function.
14282 (aarch64_print_vector_float_operand): Likewise.
14283 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14284 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14285 and the FP immediates 1.0 and 0.5.
14286 (aarch64_print_address_internal): Handle SVE addresses.
14287 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14288 (aarch64_regno_regclass): Handle predicate registers.
14289 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14291 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14292 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14293 (aarch64_convert_sve_vector_bits): New function.
14294 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14295 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14296 rather than an rtx.
14297 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14298 Handle SVE vector and predicate modes. Accept VL-based constants
14299 that need only one temporary register, and VL offsets that require
14300 no temporary registers.
14301 (aarch64_conditional_register_usage): Mark the predicate registers
14302 as fixed if SVE isn't available.
14303 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14304 Return true for SVE vector and predicate modes.
14305 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14306 rather than an unsigned int. Handle SVE modes.
14307 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14309 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14311 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14312 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14313 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14314 (aarch64_sve_float_mul_immediate_p): New functions.
14315 (aarch64_sve_valid_immediate): New function.
14316 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14317 Explicitly reject structure modes. Check for INDEX constants.
14318 Handle PTRUE and PFALSE constants.
14319 (aarch64_check_zero_based_sve_index_immediate): New function.
14320 (aarch64_simd_imm_zero_p): Delete.
14321 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14322 vector modes. Accept constants in the range of CNT[BHWD].
14323 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14324 ask for an Advanced SIMD mode.
14325 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14326 (aarch64_simd_vector_alignment): Handle SVE predicates.
14327 (aarch64_vectorize_preferred_vector_alignment): New function.
14328 (aarch64_simd_vector_alignment_reachable): Use it instead of
14330 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14331 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14333 (MAX_VECT_LEN): Delete.
14334 (expand_vec_perm_d): Add a vec_flags field.
14335 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14336 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14337 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14339 (aarch64_evpc_rev): Rename to...
14340 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14341 (aarch64_evpc_rev_global): New function.
14342 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14343 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14345 (aarch64_evpc_sve_tbl): New function.
14346 (aarch64_expand_vec_perm_const_1): Update after rename of
14347 aarch64_evpc_rev. Handle SVE permutes too, trying
14348 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14349 than aarch64_evpc_tbl.
14350 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14351 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14352 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14353 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14354 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14355 (aarch64_expand_sve_vcond): New functions.
14356 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14357 of aarch64_vector_mode_p.
14358 (aarch64_dwarf_poly_indeterminate_value): New function.
14359 (aarch64_compute_pressure_classes): Likewise.
14360 (aarch64_can_change_mode_class): Likewise.
14361 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14362 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14363 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14364 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14365 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14366 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14367 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14368 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14370 (Dn, Dl, Dr): Accept const as well as const_vector.
14371 (Dz): Likewise. Compare against CONST0_RTX.
14372 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14373 of "vector" where appropriate.
14374 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14375 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14376 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14377 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14378 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14379 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14380 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14381 (v_int_equiv): Extend to SVE modes.
14382 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14384 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14385 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14386 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14387 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14388 (SVE_COND_FP_CMP): New int iterators.
14389 (perm_hilo): Handle the new unpack unspecs.
14390 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14392 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14393 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14394 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14395 (aarch64_equality_operator, aarch64_constant_vector_operand)
14396 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14397 (aarch64_sve_nonimmediate_operand): Likewise.
14398 (aarch64_sve_general_operand): Likewise.
14399 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14400 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14401 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14402 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14403 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14404 (aarch64_sve_float_arith_immediate): Likewise.
14405 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14406 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14407 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14408 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14409 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14410 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14411 (aarch64_sve_float_arith_operand): Likewise.
14412 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14413 (aarch64_sve_float_mul_operand): Likewise.
14414 (aarch64_sve_vec_perm_operand): Likewise.
14415 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14416 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14417 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14418 as well as const_vector.
14419 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14420 in file. Use CONST0_RTX and CONSTM1_RTX.
14421 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14422 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14423 Use aarch64_simd_imm_zero.
14424 * config/aarch64/aarch64-sve.md: New file.
14425 * config/aarch64/aarch64.md: Include it.
14426 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14427 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14428 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14429 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14430 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14431 (sve): New attribute.
14432 (enabled): Disable instructions with the sve attribute unless
14434 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14435 aarch64_expand_mov_immediate.
14436 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14437 CNT[BHSD] immediates.
14438 (movti): Split CONST_POLY_INT moves into two halves.
14439 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14440 Split additions that need a temporary here if the destination
14441 is the stack pointer.
14442 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14443 (*add<mode>3_poly_1): New instruction.
14444 (set_clobber_cc): New expander.
14446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14448 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14449 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14450 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14451 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14452 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14453 Change innermode from fixed_mode_size to machine_mode.
14454 (simplify_subreg): Update call accordingly. Handle a constant-sized
14455 subreg of a variable-length CONST_VECTOR.
14457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14458 Alan Hayward <alan.hayward@arm.com>
14459 David Sherwood <david.sherwood@arm.com>
14461 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14462 (add_offset_to_base): New function, split out from...
14463 (create_mem_ref): ...here. When handling a scale other than 1,
14464 check first whether the address is valid without the offset.
14465 Add it into the base if so, leaving the index and scale as-is.
14467 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14470 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14471 fold_for_warn before checking if arg2 is INTEGER_CST.
14473 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14475 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14476 (store_multiple_operation): Delete.
14477 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14478 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14479 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14480 guarded by TARGET_STRING.
14481 (rs6000_output_load_multiple): Delete.
14482 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14483 OPTION_MASK_STRING / TARGET_STRING handling.
14484 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14485 (const rs6000_opt_masks) <"string">: Change mask to 0.
14486 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14487 (MASK_STRING): Delete.
14488 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14490 (load_multiple): Delete.
14497 (store_multiple): Delete.
14504 (movmemsi_8reg): Delete.
14505 (corresponding unnamed define_insn): Delete.
14506 (movmemsi_6reg): Delete.
14507 (corresponding unnamed define_insn): Delete.
14508 (movmemsi_4reg): Delete.
14509 (corresponding unnamed define_insn): Delete.
14510 (movmemsi_2reg): Delete.
14511 (corresponding unnamed define_insn): Delete.
14512 (movmemsi_1reg): Delete.
14513 (corresponding unnamed define_insn): Delete.
14514 * config/rs6000/rs6000.opt (mno-string): New.
14515 (mstring): Replace by deprecation warning stub.
14516 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14518 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14520 * regrename.c (regrename_do_replace): If replacing the same
14521 reg multiple times, try to reuse last created gen_raw_REG.
14524 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14525 main to workaround a bug in GDB.
14527 2018-01-12 Tom de Vries <tom@codesourcery.com>
14530 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14532 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14534 PR rtl-optimization/80481
14535 * ira-color.c (get_cap_member): New function.
14536 (allocnos_conflict_by_live_ranges_p): Use it.
14537 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14538 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14540 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14543 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14544 (*saddl_se_1): Ditto.
14545 (*ssubsi_1): Ditto.
14546 (*ssubl_se_1): Ditto.
14548 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14550 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14551 rather than wi::to_widest for DR_INITs.
14552 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14553 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14554 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14556 (vect_analyze_group_access_1): Note that here.
14558 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14560 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14561 polynomial type sizes.
14563 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14565 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14566 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14567 (gimple_add_tmp_var): Likewise.
14569 2018-01-12 Martin Liska <mliska@suse.cz>
14571 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14572 (gimple_alloc_sizes): Likewise.
14573 (dump_gimple_statistics): Use PRIu64 in printf format.
14574 * gimple.h: Change uint64_t to int.
14576 2018-01-12 Martin Liska <mliska@suse.cz>
14578 * tree-core.h: Use uint64_t instead of int.
14579 * tree.c (tree_node_counts): Likewise.
14580 (tree_node_sizes): Likewise.
14581 (dump_tree_statistics): Use PRIu64 in printf format.
14583 2018-01-12 Martin Liska <mliska@suse.cz>
14585 * Makefile.in: As qsort_chk is implemented in vec.c, add
14586 vec.o to linkage of gencfn-macros.
14587 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14588 passing the info to record_node_allocation_statistics.
14589 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14591 * ggc-common.c (struct ggc_usage): Add operator== and use
14592 it in operator< and compare function.
14593 * mem-stats.h (struct mem_usage): Likewise.
14594 * vec.c (struct vec_usage): Remove operator< and compare
14595 function. Can be simply inherited.
14597 2018-01-12 Martin Jambor <mjambor@suse.cz>
14600 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14601 * tree-ssa-math-opts.c: Include domwalk.h.
14602 (convert_mult_to_fma_1): New function.
14603 (fma_transformation_info): New type.
14604 (fma_deferring_state): Likewise.
14605 (cancel_fma_deferring): New function.
14606 (result_of_phi): Likewise.
14607 (last_fma_candidate_feeds_initial_phi): Likewise.
14608 (convert_mult_to_fma): Added deferring logic, split actual
14609 transformation to convert_mult_to_fma_1.
14610 (math_opts_dom_walker): New type.
14611 (math_opts_dom_walker::after_dom_children): New method, body moved
14612 here from pass_optimize_widening_mul::execute, added deferring logic
14614 (pass_optimize_widening_mul::execute): Moved most of code to
14615 math_opts_dom_walker::after_dom_children.
14616 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14617 * config/i386/i386.c (ix86_option_override_internal): Added
14618 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14620 2018-01-12 Richard Biener <rguenther@suse.de>
14623 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14624 inline instance vars.
14626 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14629 * config/rx/rx.c (rx_is_restricted_memory_address):
14630 Handle SUBREG case.
14632 2018-01-12 Richard Biener <rguenther@suse.de>
14634 PR tree-optimization/80846
14635 * target.def (split_reduction): New target hook.
14636 * targhooks.c (default_split_reduction): New function.
14637 * targhooks.h (default_split_reduction): Declare.
14638 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14639 target requests first reduce vectors by combining low and high
14641 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14642 (get_vectype_for_scalar_type_and_size): Export.
14643 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14644 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14645 * doc/tm.texi: Regenerate.
14646 * config/i386/i386.c (ix86_split_reduction): Implement
14647 TARGET_VECTORIZE_SPLIT_REDUCTION.
14649 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14652 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14653 in PIC mode except for TARGET_VXWORKS_RTP.
14654 * config/sparc/sparc.c: Include cfgrtl.h.
14655 (TARGET_INIT_PIC_REG): Define.
14656 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14657 (sparc_pic_register_p): New predicate.
14658 (sparc_legitimate_address_p): Use it.
14659 (sparc_legitimize_pic_address): Likewise.
14660 (sparc_delegitimize_address): Likewise.
14661 (sparc_mode_dependent_address_p): Likewise.
14662 (gen_load_pcrel_sym): Remove 4th parameter.
14663 (load_got_register): Adjust call to above. Remove obsolete stuff.
14664 (sparc_expand_prologue): Do not call load_got_register here.
14665 (sparc_flat_expand_prologue): Likewise.
14666 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14667 (sparc_use_pseudo_pic_reg): New function.
14668 (sparc_init_pic_reg): Likewise.
14669 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14670 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14672 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14674 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14675 Add item for branch_cost.
14677 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14679 PR rtl-optimization/83565
14680 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14681 not extend the result to a larger mode for rotate operations.
14682 (num_sign_bit_copies1): Likewise.
14684 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14687 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14689 Use values-Xc.o for -pedantic.
14690 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14692 2018-01-12 Martin Liska <mliska@suse.cz>
14695 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14697 (possible_polymorphic_call_targets): Use it.
14698 (ipa_devirt): Likewise.
14700 2018-01-12 Martin Liska <mliska@suse.cz>
14702 * profile-count.h (enum profile_quality): Use 0 as invalid
14703 enum value of profile_quality.
14705 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14707 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14708 -mext-string options.
14710 2018-01-12 Richard Biener <rguenther@suse.de>
14712 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14713 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14714 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14716 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14718 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14720 * configure.ac (--with-long-double-format): Add support for the
14721 configuration option to change the default long double format on
14723 * config.gcc (powerpc*-linux*-*): Likewise.
14724 * configure: Regenerate.
14725 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14726 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14727 used without modification.
14729 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14731 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14732 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14733 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14734 MISC_BUILTIN_SPEC_BARRIER.
14735 (rs6000_init_builtins): Likewise.
14736 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14738 (speculation_barrier): New define_insn.
14739 * doc/extend.texi: Document __builtin_speculation_barrier.
14741 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14744 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14745 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14746 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14748 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14749 integral modes instead of "ss" and "sd".
14750 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14751 vectors with 32-bit and 64-bit elements.
14752 (vecdupssescalarmodesuffix): New mode attribute.
14753 (vec_dup<mode>): Use it.
14755 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14758 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14759 frame if argument is passed on stack.
14761 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14764 * ree.c (combine_reaching_defs): Optimize also
14765 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14766 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14768 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14770 PR middle-end/83189
14771 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14773 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14775 PR middle-end/83718
14776 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14777 after they are computed.
14779 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14781 PR tree-optimization/83695
14782 * gimple-loop-linterchange.cc
14783 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14784 reset cached scev information after interchange.
14785 (pass_linterchange::execute): Remove call to scev_reset_htab.
14787 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14789 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14790 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14791 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14792 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14793 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14794 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14795 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14796 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14797 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14798 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14799 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14800 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14801 (V_lane_reg): Likewise.
14802 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14804 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14805 (vfmal_lane_low<mode>_intrinsic,
14806 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14807 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14808 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14809 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14810 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14811 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14813 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14815 * config/arm/arm-cpus.in (fp16fml): New feature.
14816 (ALL_SIMD): Add fp16fml.
14817 (armv8.2-a): Add fp16fml as an option.
14818 (armv8.3-a): Likewise.
14819 (armv8.4-a): Add fp16fml as part of fp16.
14820 * config/arm/arm.h (TARGET_FP16FML): Define.
14821 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14823 * config/arm/arm-modes.def (V2HF): Define.
14824 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14825 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14826 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14827 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14828 vfmsl_low, vfmsl_high): New set of builtins.
14829 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14830 (vfml_op): New code attribute.
14831 (VFMLHALVES): New int iterator.
14832 (VFML, VFMLSEL): New mode attributes.
14833 (V_reg): Define mapping for V2HF.
14834 (V_hi, V_lo): New mode attributes.
14835 (VF_constraint): Likewise.
14836 (vfml_half, vfml_half_selector): New int attributes.
14837 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14839 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14840 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14842 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14843 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14844 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14845 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14847 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14848 Document new effective target and option set.
14850 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14852 * config/arm/arm-cpus.in (armv8_4): New feature.
14853 (ARMv8_4a): New fgroup.
14854 (armv8.4-a): New arch.
14855 * config/arm/arm-tables.opt: Regenerate.
14856 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14857 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14858 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14859 Add matching rules for -march=armv8.4-a and extensions.
14860 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14862 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14865 * config/rx/rx.md (BW): New mode attribute.
14866 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14868 2018-01-11 Richard Biener <rguenther@suse.de>
14870 PR tree-optimization/83435
14871 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14872 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14873 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14875 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14876 Alan Hayward <alan.hayward@arm.com>
14877 David Sherwood <david.sherwood@arm.com>
14879 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14881 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14882 (aarch64_print_address_internal): Use it to check for a zero offset.
14884 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14885 Alan Hayward <alan.hayward@arm.com>
14886 David Sherwood <david.sherwood@arm.com>
14888 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14889 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14890 Return a poly_int64 rather than a HOST_WIDE_INT.
14891 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14892 rather than a HOST_WIDE_INT.
14893 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14894 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14895 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14896 final_offset from HOST_WIDE_INT to poly_int64.
14897 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14898 to_constant when getting the number of units in an Advanced SIMD
14900 (aarch64_builtin_vectorized_function): Check for a constant number
14902 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14904 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14905 attribute instead of GET_MODE_NUNITS.
14906 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14907 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14908 GET_MODE_SIZE for fixed-size registers.
14909 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14910 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14911 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14912 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14913 (aarch64_print_operand, aarch64_print_address_internal)
14914 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14915 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14916 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14917 Handle polynomial GET_MODE_SIZE.
14918 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14919 wider than SImode without modification.
14920 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14921 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14922 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14923 passing and returning SVE modes.
14924 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14925 rather than GEN_INT.
14926 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14927 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14928 (aarch64_allocate_and_probe_stack_space): Likewise.
14929 (aarch64_layout_frame): Cope with polynomial offsets.
14930 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14931 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14932 polynomial offsets.
14933 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14934 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14935 poly_int64 rather than a HOST_WIDE_INT.
14936 (aarch64_get_separate_components, aarch64_process_components)
14937 (aarch64_expand_prologue, aarch64_expand_epilogue)
14938 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14939 (aarch64_anchor_offset): New function, split out from...
14940 (aarch64_legitimize_address): ...here.
14941 (aarch64_builtin_vectorization_cost): Handle polynomial
14942 TYPE_VECTOR_SUBPARTS.
14943 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14945 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14946 number of elements from the PARALLEL rather than the mode.
14947 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14948 rather than GET_MODE_BITSIZE.
14949 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14950 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14951 (aarch64_expand_vec_perm_const_1): Handle polynomial
14952 d->perm.length () and d->perm elements.
14953 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14954 Apply to_constant to d->perm elements.
14955 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14956 polynomial CONST_VECTOR_NUNITS.
14957 (aarch64_move_pointer): Take amount as a poly_int64 rather
14959 (aarch64_progress_pointer): Avoid temporary variable.
14960 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14961 the mode attribute instead of GET_MODE.
14963 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14964 Alan Hayward <alan.hayward@arm.com>
14965 David Sherwood <david.sherwood@arm.com>
14967 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14968 x exists before using it.
14969 (aarch64_add_constant_internal): Rename to...
14970 (aarch64_add_offset_1): ...this. Replace regnum with separate
14971 src and dest rtxes. Handle the case in which they're different,
14972 including when the offset is zero. Replace scratchreg with an rtx.
14973 Use 2 additions if there is no spare register into which we can
14974 move a 16-bit constant.
14975 (aarch64_add_constant): Delete.
14976 (aarch64_add_offset): Replace reg with separate src and dest
14977 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14978 Use aarch64_add_offset_1.
14979 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14980 an rtx rather than an int. Take the delta as a poly_int64
14981 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14982 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14983 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14984 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14985 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14986 and aarch64_add_sp.
14987 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14988 aarch64_add_constant.
14990 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14992 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14993 Use scalar_float_mode.
14995 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14997 * config/aarch64/aarch64-simd.md
14998 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14999 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
15000 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
15001 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
15002 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
15003 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
15004 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
15005 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
15006 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
15007 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
15009 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15012 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
15013 targ_options->x_arm_arch_string is non NULL.
15015 2018-01-11 Tamar Christina <tamar.christina@arm.com>
15017 * config/aarch64/aarch64.h
15018 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
15020 2018-01-11 Sudakshina Das <sudi.das@arm.com>
15023 * expmed.c (emit_store_flag_force): Swap if const op0
15024 and change VOIDmode to mode of op0.
15026 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
15028 PR rtl-optimization/83761
15029 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
15030 than bytes to mode_for_size.
15032 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15034 PR middle-end/83189
15035 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
15036 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
15039 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15041 PR middle-end/83575
15042 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
15043 when in layout mode.
15044 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
15045 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
15048 2018-01-10 Michael Collison <michael.collison@arm.com>
15050 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
15051 * config/aarch64/aarch64-option-extension.def: Add
15052 AARCH64_OPT_EXTENSION of 'fp16fml'.
15053 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15054 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
15055 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
15056 * config/aarch64/constraints.md (Ui7): New constraint.
15057 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
15058 (VFMLA_SEL_W): Ditto.
15061 (VFMLA16_LOW): New int iterator.
15062 (VFMLA16_HIGH): Ditto.
15063 (UNSPEC_FMLAL): New unspec.
15064 (UNSPEC_FMLSL): Ditto.
15065 (UNSPEC_FMLAL2): Ditto.
15066 (UNSPEC_FMLSL2): Ditto.
15067 (f16mac): New code attribute.
15068 * config/aarch64/aarch64-simd-builtins.def
15069 (aarch64_fmlal_lowv2sf): Ditto.
15070 (aarch64_fmlsl_lowv2sf): Ditto.
15071 (aarch64_fmlalq_lowv4sf): Ditto.
15072 (aarch64_fmlslq_lowv4sf): Ditto.
15073 (aarch64_fmlal_highv2sf): Ditto.
15074 (aarch64_fmlsl_highv2sf): Ditto.
15075 (aarch64_fmlalq_highv4sf): Ditto.
15076 (aarch64_fmlslq_highv4sf): Ditto.
15077 (aarch64_fmlal_lane_lowv2sf): Ditto.
15078 (aarch64_fmlsl_lane_lowv2sf): Ditto.
15079 (aarch64_fmlal_laneq_lowv2sf): Ditto.
15080 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
15081 (aarch64_fmlalq_lane_lowv4sf): Ditto.
15082 (aarch64_fmlsl_lane_lowv4sf): Ditto.
15083 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
15084 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
15085 (aarch64_fmlal_lane_highv2sf): Ditto.
15086 (aarch64_fmlsl_lane_highv2sf): Ditto.
15087 (aarch64_fmlal_laneq_highv2sf): Ditto.
15088 (aarch64_fmlsl_laneq_highv2sf): Ditto.
15089 (aarch64_fmlalq_lane_highv4sf): Ditto.
15090 (aarch64_fmlsl_lane_highv4sf): Ditto.
15091 (aarch64_fmlalq_laneq_highv4sf): Ditto.
15092 (aarch64_fmlsl_laneq_highv4sf): Ditto.
15093 * config/aarch64/aarch64-simd.md:
15094 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
15095 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15096 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
15097 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
15098 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
15099 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
15100 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
15101 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
15102 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
15103 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
15104 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
15105 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
15106 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
15107 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
15108 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
15109 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
15110 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
15111 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
15112 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
15113 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
15114 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
15115 (vfmlsl_low_u32): Ditto.
15116 (vfmlalq_low_u32): Ditto.
15117 (vfmlslq_low_u32): Ditto.
15118 (vfmlal_high_u32): Ditto.
15119 (vfmlsl_high_u32): Ditto.
15120 (vfmlalq_high_u32): Ditto.
15121 (vfmlslq_high_u32): Ditto.
15122 (vfmlal_lane_low_u32): Ditto.
15123 (vfmlsl_lane_low_u32): Ditto.
15124 (vfmlal_laneq_low_u32): Ditto.
15125 (vfmlsl_laneq_low_u32): Ditto.
15126 (vfmlalq_lane_low_u32): Ditto.
15127 (vfmlslq_lane_low_u32): Ditto.
15128 (vfmlalq_laneq_low_u32): Ditto.
15129 (vfmlslq_laneq_low_u32): Ditto.
15130 (vfmlal_lane_high_u32): Ditto.
15131 (vfmlsl_lane_high_u32): Ditto.
15132 (vfmlal_laneq_high_u32): Ditto.
15133 (vfmlsl_laneq_high_u32): Ditto.
15134 (vfmlalq_lane_high_u32): Ditto.
15135 (vfmlslq_lane_high_u32): Ditto.
15136 (vfmlalq_laneq_high_u32): Ditto.
15137 (vfmlslq_laneq_high_u32): Ditto.
15138 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15139 (AARCH64_FL_FOR_ARCH8_4): New.
15140 (AARCH64_ISA_F16FML): New ISA flag.
15141 (TARGET_F16FML): New feature flag for fp16fml.
15142 (doc/invoke.texi): Document new fp16fml option.
15144 2018-01-10 Michael Collison <michael.collison@arm.com>
15146 * config/aarch64/aarch64-builtins.c:
15147 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15148 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15149 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15150 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15151 (AARCH64_ISA_SHA3): New ISA flag.
15152 (TARGET_SHA3): New feature flag for sha3.
15153 * config/aarch64/iterators.md (sha512_op): New int attribute.
15154 (CRYPTO_SHA512): New int iterator.
15155 (UNSPEC_SHA512H): New unspec.
15156 (UNSPEC_SHA512H2): Ditto.
15157 (UNSPEC_SHA512SU0): Ditto.
15158 (UNSPEC_SHA512SU1): Ditto.
15159 * config/aarch64/aarch64-simd-builtins.def
15160 (aarch64_crypto_sha512hqv2di): New builtin.
15161 (aarch64_crypto_sha512h2qv2di): Ditto.
15162 (aarch64_crypto_sha512su0qv2di): Ditto.
15163 (aarch64_crypto_sha512su1qv2di): Ditto.
15164 (aarch64_eor3qv8hi): Ditto.
15165 (aarch64_rax1qv2di): Ditto.
15166 (aarch64_xarqv2di): Ditto.
15167 (aarch64_bcaxqv8hi): Ditto.
15168 * config/aarch64/aarch64-simd.md:
15169 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15170 (aarch64_crypto_sha512su0qv2di): Ditto.
15171 (aarch64_crypto_sha512su1qv2di): Ditto.
15172 (aarch64_eor3qv8hi): Ditto.
15173 (aarch64_rax1qv2di): Ditto.
15174 (aarch64_xarqv2di): Ditto.
15175 (aarch64_bcaxqv8hi): Ditto.
15176 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15177 (vsha512h2q_u64): Ditto.
15178 (vsha512su0q_u64): Ditto.
15179 (vsha512su1q_u64): Ditto.
15180 (veor3q_u16): Ditto.
15181 (vrax1q_u64): Ditto.
15182 (vxarq_u64): Ditto.
15183 (vbcaxq_u16): Ditto.
15184 * config/arm/types.md (crypto_sha512): New type attribute.
15185 (crypto_sha3): Ditto.
15186 (doc/invoke.texi): Document new sha3 option.
15188 2018-01-10 Michael Collison <michael.collison@arm.com>
15190 * config/aarch64/aarch64-builtins.c:
15191 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15192 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15193 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15194 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15195 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15196 (AARCH64_ISA_SM4): New ISA flag.
15197 (TARGET_SM4): New feature flag for sm4.
15198 * config/aarch64/aarch64-simd-builtins.def
15199 (aarch64_sm3ss1qv4si): Ditto.
15200 (aarch64_sm3tt1aq4si): Ditto.
15201 (aarch64_sm3tt1bq4si): Ditto.
15202 (aarch64_sm3tt2aq4si): Ditto.
15203 (aarch64_sm3tt2bq4si): Ditto.
15204 (aarch64_sm3partw1qv4si): Ditto.
15205 (aarch64_sm3partw2qv4si): Ditto.
15206 (aarch64_sm4eqv4si): Ditto.
15207 (aarch64_sm4ekeyqv4si): Ditto.
15208 * config/aarch64/aarch64-simd.md:
15209 (aarch64_sm3ss1qv4si): Ditto.
15210 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15211 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15212 (aarch64_sm4eqv4si): Ditto.
15213 (aarch64_sm4ekeyqv4si): Ditto.
15214 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15215 (sm3part_op): Ditto.
15216 (CRYPTO_SM3TT): Ditto.
15217 (CRYPTO_SM3PART): Ditto.
15218 (UNSPEC_SM3SS1): New unspec.
15219 (UNSPEC_SM3TT1A): Ditto.
15220 (UNSPEC_SM3TT1B): Ditto.
15221 (UNSPEC_SM3TT2A): Ditto.
15222 (UNSPEC_SM3TT2B): Ditto.
15223 (UNSPEC_SM3PARTW1): Ditto.
15224 (UNSPEC_SM3PARTW2): Ditto.
15225 (UNSPEC_SM4E): Ditto.
15226 (UNSPEC_SM4EKEY): Ditto.
15227 * config/aarch64/constraints.md (Ui2): New constraint.
15228 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15229 * config/arm/types.md (crypto_sm3): New type attribute.
15230 (crypto_sm4): Ditto.
15231 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15232 (vsm3tt1aq_u32): Ditto.
15233 (vsm3tt1bq_u32): Ditto.
15234 (vsm3tt2aq_u32): Ditto.
15235 (vsm3tt2bq_u32): Ditto.
15236 (vsm3partw1q_u32): Ditto.
15237 (vsm3partw2q_u32): Ditto.
15238 (vsm4eq_u32): Ditto.
15239 (vsm4ekeyq_u32): Ditto.
15240 (doc/invoke.texi): Document new sm4 option.
15242 2018-01-10 Michael Collison <michael.collison@arm.com>
15244 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15245 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15246 (AARCH64_FL_FOR_ARCH8_4): New.
15247 (AARCH64_FL_V8_4): New flag.
15248 (doc/invoke.texi): Document new armv8.4-a option.
15250 2018-01-10 Michael Collison <michael.collison@arm.com>
15252 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15253 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15254 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15255 * config/aarch64/aarch64-option-extension.def: Add
15256 AARCH64_OPT_EXTENSION of 'sha2'.
15257 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15258 (crypto): Disable sha2 and aes if crypto disabled.
15259 (crypto): Enable aes and sha2 if enabled.
15260 (simd): Disable sha2 and aes if simd disabled.
15261 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15263 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15264 (TARGET_SHA2): New feature flag for sha2.
15265 (TARGET_AES): New feature flag for aes.
15266 * config/aarch64/aarch64-simd.md:
15267 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15268 conditional on TARGET_AES.
15269 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15270 (aarch64_crypto_sha1hsi): Make pattern conditional
15272 (aarch64_crypto_sha1hv4si): Ditto.
15273 (aarch64_be_crypto_sha1hv4si): Ditto.
15274 (aarch64_crypto_sha1su1v4si): Ditto.
15275 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15276 (aarch64_crypto_sha1su0v4si): Ditto.
15277 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15278 (aarch64_crypto_sha256su0v4si): Ditto.
15279 (aarch64_crypto_sha256su1v4si): Ditto.
15280 (doc/invoke.texi): Document new aes and sha2 options.
15282 2018-01-10 Martin Sebor <msebor@redhat.com>
15284 PR tree-optimization/83781
15285 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15288 2018-01-11 Martin Sebor <msebor@gmail.com>
15289 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15291 PR tree-optimization/83501
15292 PR tree-optimization/81703
15294 * tree-ssa-strlen.c (get_string_cst): Rename...
15295 (get_string_len): ...to this. Handle global constants.
15296 (handle_char_store): Adjust.
15298 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15299 Jim Wilson <jimw@sifive.com>
15301 * config/riscv/riscv-protos.h (riscv_output_return): New.
15302 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15303 (riscv_attribute_table, riscv_output_return),
15304 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15305 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15306 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15308 (riscv_expand_prologue): Add early return for naked function.
15309 (riscv_expand_epilogue): Likewise.
15310 (riscv_function_ok_for_sibcall): Return false for naked function.
15311 (riscv_set_current_function): New.
15312 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15313 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15314 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15315 * doc/extend.texi (RISC-V Function Attributes): New.
15317 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15319 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15320 check for 128-bit long double before checking TCmode.
15321 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15322 128-bit long doubles before checking TFmode or TCmode.
15323 (FLOAT128_IBM_P): Likewise.
15325 2018-01-10 Martin Sebor <msebor@redhat.com>
15327 PR tree-optimization/83671
15328 * builtins.c (c_strlen): Unconditionally return zero for the empty
15330 Use -Warray-bounds for warnings.
15331 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15332 for non-constant array indices with COMPONENT_REF, arrays of
15333 arrays, and pointers to arrays.
15334 (gimple_fold_builtin_strlen): Determine and set length range for
15335 non-constant character arrays.
15337 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15339 PR middle-end/81897
15340 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15343 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15345 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15347 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15350 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15351 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15352 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15353 indexed_or_indirect_operand predicate.
15354 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15355 (*vsx_le_perm_load_v8hi): Likewise.
15356 (*vsx_le_perm_load_v16qi): Likewise.
15357 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15358 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15359 (*vsx_le_perm_store_v8hi): Likewise.
15360 (*vsx_le_perm_store_v16qi): Likewise.
15361 (eight unnamed splitters): Likewise.
15363 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15365 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15366 * config/rs6000/emmintrin.h: Likewise.
15367 * config/rs6000/mmintrin.h: Likewise.
15368 * config/rs6000/xmmintrin.h: Likewise.
15370 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15373 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15375 * tree.c (tree_nop_conversion): Return true for location wrapper
15377 (maybe_wrap_with_location): New function.
15378 (selftest::check_strip_nops): New function.
15379 (selftest::test_location_wrappers): New function.
15380 (selftest::tree_c_tests): Call it.
15381 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15382 (maybe_wrap_with_location): New decl.
15383 (EXPR_LOCATION_WRAPPER_P): New macro.
15384 (location_wrapper_p): New inline function.
15385 (tree_strip_any_location_wrapper): New inline function.
15387 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15390 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15391 stack_realign_offset for the largest alignment of stack slot
15393 (ix86_find_max_used_stack_alignment): New function.
15394 (ix86_finalize_stack_frame_flags): Use it. Set
15395 max_used_stack_alignment if we don't realign stack.
15396 * config/i386/i386.h (machine_function): Add
15397 max_used_stack_alignment.
15399 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15401 * config/arm/arm.opt (-mbranch-cost): New option.
15402 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15405 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15408 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15409 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15411 2018-01-10 Richard Biener <rguenther@suse.de>
15414 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15415 early out so it also covers the case where we have a non-NULL
15418 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15420 PR tree-optimization/83753
15421 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15422 for non-strided grouped accesses if the number of elements is 1.
15424 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15427 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15428 * i386.h (TARGET_USE_GATHER): Define.
15429 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15431 2018-01-10 Martin Liska <mliska@suse.cz>
15434 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15435 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15437 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15438 CLEANUP_NO_PARTITIONING is not set.
15440 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15442 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15443 for vectors, as a partial revert of r254296.
15444 * rtl.h (const_vec_p): Delete.
15445 (const_vec_duplicate_p): Don't test for vector CONSTs.
15446 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15447 * expmed.c (make_tree): Likewise.
15450 * common.md (E, F): Use CONSTANT_P instead of checking for
15452 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15453 checking for CONST_VECTOR.
15455 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15457 PR middle-end/83575
15458 * predict.c (force_edge_cold): Handle in more sane way edges
15459 with no prediction.
15461 2018-01-09 Carl Love <cel@us.ibm.com>
15463 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15465 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15466 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15467 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15468 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15469 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15470 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15471 * config/rs6000/rs6000-protos.h: Add extern defition for
15472 rs6000_generate_float2_double_code.
15473 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15475 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15476 (float2_v2df): Add define_expand.
15478 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15481 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15482 op_mode in the force_to_mode call.
15484 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15486 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15487 instead of checking each element individually.
15488 (aarch64_evpc_uzp): Likewise.
15489 (aarch64_evpc_zip): Likewise.
15490 (aarch64_evpc_ext): Likewise.
15491 (aarch64_evpc_rev): Likewise.
15492 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15493 instead of checking each element individually. Return true without
15495 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15496 whether all selected elements come from the same input, instead of
15497 checking each element individually. Remove calls to gen_rtx_REG,
15498 start_sequence and end_sequence and instead assert that no rtl is
15501 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15503 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15504 order of HIGH and CONST checks.
15506 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15508 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15509 if the destination isn't an SSA_NAME.
15511 2018-01-09 Richard Biener <rguenther@suse.de>
15513 PR tree-optimization/83668
15514 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15516 (canonicalize_loop_form): ... here, renamed from ...
15517 (canonicalize_loop_closed_ssa_form): ... this and amended to
15518 swap successor edges for loop exit blocks to make us use
15519 the RPO order we need for initial schedule generation.
15521 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15523 PR tree-optimization/64811
15524 * match.pd: When optimizing comparisons with Inf, avoid
15525 introducing or losing exceptions from comparisons with NaN.
15527 2018-01-09 Martin Liska <mliska@suse.cz>
15530 * asan.c (shadow_mem_size): Add gcc_assert.
15532 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15534 Don't save registers in main().
15537 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15538 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15539 * config/avr/avr.c (avr_set_current_function): Don't error if
15540 naked, OS_task or OS_main are specified at the same time.
15541 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15543 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15545 * common/config/avr/avr-common.c (avr_option_optimization_table):
15546 Switch on -mmain-is-OS_task for optimizing compilations.
15548 2018-01-09 Richard Biener <rguenther@suse.de>
15550 PR tree-optimization/83572
15551 * graphite.c: Include cfganal.h.
15552 (graphite_transform_loops): Connect infinite loops to exit
15553 and remove fake edges at the end.
15555 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15557 * ipa-inline.c (edge_badness): Revert accidental checkin.
15559 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15562 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15563 symbols; not inline clones.
15565 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15568 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15569 hard registers. Formatting fixes.
15571 PR preprocessor/83722
15572 * gcc.c (try_generate_repro): Pass
15573 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15574 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15577 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15578 Kito Cheng <kito.cheng@gmail.com>
15580 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15581 (riscv_leaf_function_p): Delete.
15582 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15584 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15586 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15588 (do_ifelse): New function.
15589 (do_isel): New function.
15590 (do_sub3): New function.
15591 (do_add3): New function.
15592 (do_load_mask_compare): New function.
15593 (do_overlap_load_compare): New function.
15594 (expand_compare_loop): New function.
15595 (expand_block_compare): Call expand_compare_loop() when appropriate.
15596 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15597 option description.
15598 (-mblock-compare-inline-loop-limit): New option.
15600 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15603 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15604 Reverse order of second and third operands in first alternative.
15605 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15606 of first and second elements in UNSPEC_VPERMR vector.
15607 (altivec_expand_vec_perm_le): Likewise.
15609 2018-01-08 Jeff Law <law@redhat.com>
15611 PR rtl-optimizatin/81308
15612 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15613 (process_switch): If group_case_labels makes a change, then set
15615 (pass_convert_switch::execute): If a switch is converted, then
15616 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15618 PR rtl-optimization/81308
15619 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15622 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15624 PR target/83663 - Revert r255946
15625 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15626 generation for cases where splatting a value is not useful.
15627 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15628 across a vec_duplicate and a paradoxical subreg forming a vector
15629 mode to a vec_concat.
15631 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15633 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15634 -march=armv8.3-a variants.
15635 * config/arm/t-multilib: Likewise.
15636 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15638 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15640 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15642 (cceq_ior_compare_complement): Give it a name so I can use it, and
15643 change boolean_or_operator predicate to boolean_operator so it can
15644 be used to generate a crand.
15645 (eqne): New code iterator.
15646 (bd/bd_neg): New code_attrs.
15647 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15648 a single define_insn.
15649 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15650 decrement (bdnzt/bdnzf/bdzt/bdzf).
15651 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15652 with the new names of the branch decrement patterns, and added the
15653 names of the branch decrement conditional patterns.
15655 2018-01-08 Richard Biener <rguenther@suse.de>
15657 PR tree-optimization/83563
15658 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15661 2018-01-08 Richard Biener <rguenther@suse.de>
15663 PR middle-end/83713
15664 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15666 2018-01-08 Richard Biener <rguenther@suse.de>
15668 PR tree-optimization/83685
15669 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15670 references to abnormals.
15672 2018-01-08 Richard Biener <rguenther@suse.de>
15675 * dwarf2out.c (output_indirect_strings): Handle empty
15676 skeleton_debug_str_hash.
15677 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15679 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15681 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15682 (emit_store_direct): Likewise.
15683 (arc_trampoline_adjust_address): Likewise.
15684 (arc_asm_trampoline_template): New function.
15685 (arc_initialize_trampoline): Use asm_trampoline_template.
15686 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15687 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15688 * config/arc/arc.md (flush_icache): Delete pattern.
15690 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15692 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15693 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15696 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15699 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15700 by not USED_FOR_TARGET.
15701 (make_pass_resolve_sw_modes): Likewise.
15703 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15705 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15708 2018-01-08 Richard Biener <rguenther@suse.de>
15710 PR middle-end/83580
15711 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15713 2018-01-08 Richard Biener <rguenther@suse.de>
15715 PR middle-end/83517
15716 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15718 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15720 PR middle-end/81897
15721 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15722 basic blocks with a small number of successors.
15723 (convert_control_dep_chain_into_preds): Improve handling of
15725 (dump_predicates): Split apart into...
15726 (dump_pred_chain): ...here...
15727 (dump_pred_info): ...and here.
15728 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15729 (can_chain_union_be_invalidated_p): Improve check for invalidation
15731 (uninit_uses_cannot_happen): Avoid unnecessary if
15732 convert_control_dep_chain_into_preds yielded nothing.
15734 2018-01-06 Martin Sebor <msebor@redhat.com>
15736 PR tree-optimization/83640
15737 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15738 subtracting negative offset from size.
15739 (builtin_access::overlap): Adjust offset bounds of the access to fall
15740 within the size of the object if possible.
15742 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15744 PR rtl-optimization/83699
15745 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15746 extract_bit_field_as_subreg to cases in which the extracted
15747 value is also a vector.
15749 * lra-constraints.c (process_alt_operands): Test for the equivalence
15750 substitutions when detecting a possible reload cycle.
15752 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15755 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15756 by default if flag_selective_schedling{,2}. Formatting fixes.
15758 PR rtl-optimization/83682
15759 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15760 if it has non-VECTOR_MODE element mode.
15761 (vec_duplicate_p): Likewise.
15763 PR middle-end/83694
15764 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15765 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15767 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15770 * config/i386/i386-builtin.def
15771 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15772 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15773 Require also OPTION_MASK_ISA_AVX512F in addition to
15774 OPTION_MASK_ISA_GFNI.
15775 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15776 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15777 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15778 to OPTION_MASK_ISA_GFNI.
15779 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15780 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15781 OPTION_MASK_ISA_AVX512BW.
15782 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15783 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15784 addition to OPTION_MASK_ISA_GFNI.
15785 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15786 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15787 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15788 to OPTION_MASK_ISA_GFNI.
15789 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15790 a requirement for all ISAs rather than any of them with a few
15792 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15794 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15795 bitmasks to be enabled with 3 exceptions, instead of requiring any
15796 enabled ISA with lots of exceptions.
15797 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15798 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15799 Change avx512bw in isa attribute to avx512f.
15800 * config/i386/sgxintrin.h: Add license boilerplate.
15801 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15802 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15803 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15804 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15806 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15807 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15808 temporarily sse2 rather than sse if not enabled already.
15811 * config/i386/sse.md (VI248_VLBW): Rename to ...
15812 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15813 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15814 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15815 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15816 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15817 mode iterator instead of VI248_VLBW.
15819 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15821 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15822 (record_modified): Skip clobbers; add debug output.
15823 (param_change_prob): Use sreal frequencies.
15825 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15827 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15828 punt for user-aligned variables.
15830 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15832 * tree-chrec.c (chrec_contains_symbols): Return true for
15835 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15838 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15839 of (x|y) == x for BICS pattern.
15841 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15843 PR tree-optimization/83605
15844 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15845 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15848 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15850 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15851 * config/epiphany/rtems.h: New file.
15853 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15854 Uros Bizjak <ubizjak@gmail.com>
15857 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15858 QIreg_operand instead of register_operand predicate.
15859 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15860 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15861 comments instead of -fmitigate[-_]rop.
15863 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15866 * cgraphunit.c (symbol_table::compile): Switch to text_section
15867 before calling assembly_start debug hook.
15868 * run-rtl-passes.c (run_rtl_passes): Likewise.
15871 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15873 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15874 range_int_cst_p rather than !symbolic_range_p before calling
15875 extract_range_from_multiplicative_op_1.
15877 2018-01-04 Jeff Law <law@redhat.com>
15879 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15880 redundant test in assertion.
15882 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15884 * doc/rtl.texi: Document machine_mode wrapper classes.
15886 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15888 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15889 using tree_to_uhwi.
15891 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15893 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15894 the VEC_PERM_EXPR fold to fail.
15896 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15899 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15900 to switched_sections.
15902 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15905 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15906 test for d.testing.
15908 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15911 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15912 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15914 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15917 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15918 is BLKmode and bitpos not zero or mode change is needed.
15920 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15923 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15926 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15929 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15930 instead of MULT rtx. Update all corresponding splitters.
15931 (*saddl_se): Ditto.
15932 (*ssub<modesuffix>): Ditto.
15933 (*ssubl_se): Ditto.
15934 (*cmp_sadd_di): Update split patterns.
15935 (*cmp_sadd_si): Ditto.
15936 (*cmp_sadd_sidi): Ditto.
15937 (*cmp_ssub_di): Ditto.
15938 (*cmp_ssub_si): Ditto.
15939 (*cmp_ssub_sidi): Ditto.
15940 * config/alpha/predicates.md (const23_operand): New predicate.
15941 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15942 Look for ASHIFT, not MULT inner operand.
15943 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15945 2018-01-04 Martin Liska <mliska@suse.cz>
15947 PR gcov-profile/83669
15948 * gcov.c (output_intermediate_file): Add version to intermediate
15950 * doc/gcov.texi: Document new field 'version' in intermediate
15951 file format. Fix location of '-k' option of gcov command.
15953 2018-01-04 Martin Liska <mliska@suse.cz>
15956 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15958 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15960 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15962 2018-01-03 Martin Sebor <msebor@redhat.com>
15964 PR tree-optimization/83655
15965 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15966 checking calls with invalid arguments.
15968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15970 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15971 (vectorizable_mask_load_store): Delete.
15972 (vectorizable_call): Return false for masked loads and stores.
15973 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15974 instead of gimple_assign_rhs1.
15975 (vectorizable_load): Handle IFN_MASK_LOAD.
15976 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15980 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15982 (vectorizable_mask_load_store): ...here.
15983 (vectorizable_load): ...and here.
15985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15987 * tree-vect-stmts.c (vect_build_all_ones_mask)
15988 (vect_build_zero_merge_argument): New functions, split out from...
15989 (vectorizable_load): ...here.
15991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15993 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15995 (vectorizable_mask_load_store): ...here.
15996 (vectorizable_store): ...and here.
15998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16000 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
16002 (vectorizable_mask_load_store): ...here.
16004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16006 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
16007 (vect_model_store_cost): Take a vec_load_store_type instead of a
16009 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
16010 (vect_model_store_cost): Take a vec_load_store_type instead of a
16012 (vectorizable_mask_load_store): Update accordingly.
16013 (vectorizable_store): Likewise.
16014 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
16016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16018 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
16019 IFN_MASK_LOAD calls here rather than...
16020 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
16022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16023 Alan Hayward <alan.hayward@arm.com>
16024 David Sherwood <david.sherwood@arm.com>
16026 * expmed.c (extract_bit_field_1): For vector extracts,
16027 fall back to extract_bit_field_as_subreg if vec_extract
16030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16031 Alan Hayward <alan.hayward@arm.com>
16032 David Sherwood <david.sherwood@arm.com>
16034 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
16035 they are variable or constant sized.
16036 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
16037 slots for constant-sized data.
16039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16040 Alan Hayward <alan.hayward@arm.com>
16041 David Sherwood <david.sherwood@arm.com>
16043 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
16044 handling COND_EXPRs with boolean comparisons, try to find a better
16045 basis for the mask type than the boolean itself.
16047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16049 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
16050 is calculated and how it can be overridden.
16051 * genmodes.c (max_bitsize_mode_any_mode): New variable.
16052 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
16054 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
16057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16058 Alan Hayward <alan.hayward@arm.com>
16059 David Sherwood <david.sherwood@arm.com>
16061 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
16062 Remove the mode argument.
16063 (aarch64_simd_valid_immediate): Remove the mode and inverse
16065 * config/aarch64/iterators.md (bitsize): New iterator.
16066 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
16067 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
16068 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
16069 aarch64_simd_valid_immediate.
16070 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
16071 (aarch64_reg_or_bic_imm): Likewise.
16072 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
16073 with an insn_type enum and msl with a modifier_type enum.
16074 Replace element_width with a scalar_mode. Change the shift
16075 to unsigned int. Add constructors for scalar_float_mode and
16076 scalar_int_mode elements.
16077 (aarch64_vect_float_const_representable_p): Delete.
16078 (aarch64_can_const_movi_rtx_p)
16079 (aarch64_simd_scalar_immediate_valid_for_move)
16080 (aarch64_simd_make_constant): Update call to
16081 aarch64_simd_valid_immediate.
16082 (aarch64_advsimd_valid_immediate_hs): New function.
16083 (aarch64_advsimd_valid_immediate): Likewise.
16084 (aarch64_simd_valid_immediate): Remove mode and inverse
16085 arguments. Rewrite to use the above. Use const_vec_duplicate_p
16086 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
16087 and aarch64_float_const_representable_p on the result.
16088 (aarch64_output_simd_mov_immediate): Remove mode argument.
16089 Update call to aarch64_simd_valid_immediate and use of
16090 simd_immediate_info.
16091 (aarch64_output_scalar_simd_mov_immediate): Update call
16094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16095 Alan Hayward <alan.hayward@arm.com>
16096 David Sherwood <david.sherwood@arm.com>
16098 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
16099 (mode_nunits): Likewise CONST_MODE_NUNITS.
16100 * machmode.def (ADJUST_NUNITS): Document.
16101 * genmodes.c (mode_data::need_nunits_adj): New field.
16102 (blank_mode): Update accordingly.
16103 (adj_nunits): New variable.
16104 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
16106 (emit_mode_size_inline): Set need_bytesize_adj for all modes
16107 listed in adj_nunits.
16108 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
16109 listed in adj_nunits. Don't emit case statements for such modes.
16110 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
16111 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
16112 nothing if adj_nunits is nonnull.
16113 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
16114 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
16115 (emit_mode_fbit): Update use of print_maybe_const_decl.
16116 (emit_move_size): Likewise. Treat the array as non-const
16118 (emit_mode_adjustments): Handle adj_nunits.
16120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16122 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
16123 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
16124 (VECTOR_MODES): Use it.
16125 (make_vector_modes): Take the prefix as an argument.
16127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16128 Alan Hayward <alan.hayward@arm.com>
16129 David Sherwood <david.sherwood@arm.com>
16131 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16132 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16133 for MODE_VECTOR_BOOL.
16134 * machmode.def (VECTOR_BOOL_MODE): Document.
16135 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16136 (make_vector_bool_mode): New function.
16137 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16139 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16140 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16142 * stor-layout.c (int_mode_for_mode): Likewise.
16143 * tree.c (build_vector_type_for_mode): Likewise.
16144 * varasm.c (output_constant_pool_2): Likewise.
16145 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16146 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16147 for MODE_VECTOR_BOOL.
16148 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16149 of mode class checks.
16150 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16151 instead of a list of mode class checks.
16152 (expand_vector_scalar_condition): Likewise.
16153 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16156 Alan Hayward <alan.hayward@arm.com>
16157 David Sherwood <david.sherwood@arm.com>
16159 * machmode.h (mode_size): Change from unsigned short to
16161 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16162 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16163 or if measurement_type is not polynomial.
16164 (fixed_size_mode::includes_p): Check for constant-sized modes.
16165 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16166 return a poly_uint16 rather than an unsigned short.
16167 (emit_mode_size): Change the type of mode_size from unsigned short
16168 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16169 (emit_mode_adjustments): Cope with polynomial vector sizes.
16170 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16172 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16174 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16175 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16176 * caller-save.c (setup_save_areas): Likewise.
16177 (replace_reg_with_saved_mem): Likewise.
16178 * calls.c (emit_library_call_value_1): Likewise.
16179 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16180 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16181 (gen_lowpart_for_combine): Likewise.
16182 * convert.c (convert_to_integer_1): Likewise.
16183 * cse.c (equiv_constant, cse_insn): Likewise.
16184 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16185 (cselib_subst_to_values): Likewise.
16186 * dce.c (word_dce_process_block): Likewise.
16187 * df-problems.c (df_word_lr_mark_ref): Likewise.
16188 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16189 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16190 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16191 (rtl_for_decl_location): Likewise.
16192 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16193 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16194 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16195 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16196 (expand_expr_real_1): Likewise.
16197 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16198 (pad_below): Likewise.
16199 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16200 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16201 * ira.c (get_subreg_tracking_sizes): Likewise.
16202 * ira-build.c (ira_create_allocno_objects): Likewise.
16203 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16204 (ira_sort_regnos_for_alter_reg): Likewise.
16205 * ira-costs.c (record_operand_costs): Likewise.
16206 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16207 (resolve_simple_move): Likewise.
16208 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16209 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16210 (lra_constraints): Likewise.
16211 (CONST_POOL_OK_P): Reject variable-sized modes.
16212 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16213 (add_pseudo_to_slot, lra_spill): Likewise.
16214 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16215 * optabs-query.c (get_best_extraction_insn): Likewise.
16216 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16217 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16218 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16219 * recog.c (offsettable_address_addr_space_p): Likewise.
16220 * regcprop.c (maybe_mode_change): Likewise.
16221 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16222 * regrename.c (build_def_use): Likewise.
16223 * regstat.c (dump_reg_info): Likewise.
16224 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16225 (find_reloads, find_reloads_subreg_address): Likewise.
16226 * reload1.c (eliminate_regs_1): Likewise.
16227 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16228 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16229 (simplify_binary_operation_1, simplify_subreg): Likewise.
16230 * targhooks.c (default_function_arg_padding): Likewise.
16231 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16232 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16233 (verify_gimple_assign_ternary): Likewise.
16234 * tree-inline.c (estimate_move_cost): Likewise.
16235 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16236 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16237 (get_address_cost_ainc): Likewise.
16238 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16239 (vect_supportable_dr_alignment): Likewise.
16240 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16241 (vectorizable_reduction): Likewise.
16242 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16243 (vectorizable_operation, vectorizable_load): Likewise.
16244 * tree.c (build_same_sized_truth_vector_type): Likewise.
16245 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16246 * var-tracking.c (emit_note_insn_var_location): Likewise.
16247 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16248 (ADDR_VEC_ALIGN): Likewise.
16250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16251 Alan Hayward <alan.hayward@arm.com>
16252 David Sherwood <david.sherwood@arm.com>
16254 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16256 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16257 or if measurement_type is polynomial.
16258 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16259 * combine.c (make_extraction): Likewise.
16260 * dse.c (find_shift_sequence): Likewise.
16261 * dwarf2out.c (mem_loc_descriptor): Likewise.
16262 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16263 (extract_bit_field, extract_low_bits): Likewise.
16264 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16265 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16266 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16267 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16268 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16269 * reload.c (find_reloads): Likewise.
16270 * reload1.c (alter_reg): Likewise.
16271 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16272 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16273 * tree-if-conv.c (predicate_mem_writes): Likewise.
16274 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16275 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16276 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16277 * valtrack.c (dead_debug_insert_temp): Likewise.
16278 * varasm.c (mergeable_constant_section): Likewise.
16279 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16282 Alan Hayward <alan.hayward@arm.com>
16283 David Sherwood <david.sherwood@arm.com>
16285 * expr.c (expand_assignment): Cope with polynomial mode sizes
16286 when assigning to a CONCAT.
16288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16289 Alan Hayward <alan.hayward@arm.com>
16290 David Sherwood <david.sherwood@arm.com>
16292 * machmode.h (mode_precision): Change from unsigned short to
16294 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16296 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16297 or if measurement_type is not polynomial.
16298 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16299 in which the mode is already known to be a scalar_int_mode.
16300 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16301 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16303 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16304 for GET_MODE_PRECISION.
16305 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16306 for GET_MODE_PRECISION.
16307 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16309 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16310 (expand_field_assignment, make_extraction): Likewise.
16311 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16312 (get_last_value): Likewise.
16313 * convert.c (convert_to_integer_1): Likewise.
16314 * cse.c (cse_insn): Likewise.
16315 * expr.c (expand_expr_real_1): Likewise.
16316 * lra-constraints.c (simplify_operand_subreg): Likewise.
16317 * optabs-query.c (can_atomic_load_p): Likewise.
16318 * optabs.c (expand_atomic_load): Likewise.
16319 (expand_atomic_store): Likewise.
16320 * ree.c (combine_reaching_defs): Likewise.
16321 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16322 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16323 * tree.h (type_has_mode_precision_p): Likewise.
16324 * ubsan.c (instrument_si_overflow): Likewise.
16326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16327 Alan Hayward <alan.hayward@arm.com>
16328 David Sherwood <david.sherwood@arm.com>
16330 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16331 polynomial numbers of units.
16332 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16333 (valid_vector_subparts_p): New function.
16334 (build_vector_type): Remove temporary shim and take the number
16335 of units as a poly_uint64 rather than an int.
16336 (build_opaque_vector_type): Take the number of units as a
16337 poly_uint64 rather than an int.
16338 * tree.c (build_vector_from_ctor): Handle polynomial
16339 TYPE_VECTOR_SUBPARTS.
16340 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16341 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16342 (build_vector_from_val): If the number of units is variable,
16343 use build_vec_duplicate_cst for constant operands and
16344 VEC_DUPLICATE_EXPR otherwise.
16345 (make_vector_type): Remove temporary is_constant ().
16346 (build_vector_type, build_opaque_vector_type): Take the number of
16347 units as a poly_uint64 rather than an int.
16348 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16350 * cfgexpand.c (expand_debug_expr): Likewise.
16351 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16352 (store_constructor, expand_expr_real_1): Likewise.
16353 (const_scalar_mask_from_tree): Likewise.
16354 * fold-const-call.c (fold_const_reduction): Likewise.
16355 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16356 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16357 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16358 (fold_relational_const): Likewise.
16359 (native_interpret_vector): Likewise. Change the size from an
16360 int to an unsigned int.
16361 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16362 TYPE_VECTOR_SUBPARTS.
16363 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16364 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16365 duplicating a non-constant operand into a variable-length vector.
16366 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16367 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16368 * ipa-icf.c (sem_variable::equals): Likewise.
16369 * match.pd: Likewise.
16370 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16371 * print-tree.c (print_node): Likewise.
16372 * stor-layout.c (layout_type): Likewise.
16373 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16374 * tree-cfg.c (verify_gimple_comparison): Likewise.
16375 (verify_gimple_assign_binary): Likewise.
16376 (verify_gimple_assign_ternary): Likewise.
16377 (verify_gimple_assign_single): Likewise.
16378 * tree-pretty-print.c (dump_generic_node): Likewise.
16379 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16380 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16381 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16382 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16383 (vect_shift_permute_load_chain): Likewise.
16384 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16385 (expand_vector_condition, optimize_vector_constructor): Likewise.
16386 (lower_vec_perm, get_compute_type): Likewise.
16387 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16388 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16389 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16390 (vect_recog_mask_conversion_pattern): Likewise.
16391 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16392 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16393 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16394 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16395 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16396 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16397 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16398 (supportable_widening_operation): Likewise.
16399 (supportable_narrowing_operation): Likewise.
16400 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16402 * varasm.c (output_constant): Likewise.
16404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16405 Alan Hayward <alan.hayward@arm.com>
16406 David Sherwood <david.sherwood@arm.com>
16408 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16409 so that both the length == 3 and length != 3 cases set up their
16410 own permute vectors. Add comments explaining why we know the
16411 number of elements is constant.
16412 (vect_permute_load_chain): Likewise.
16414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16415 Alan Hayward <alan.hayward@arm.com>
16416 David Sherwood <david.sherwood@arm.com>
16418 * machmode.h (mode_nunits): Change from unsigned char to
16420 (ONLY_FIXED_SIZE_MODES): New macro.
16421 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16422 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16423 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16425 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16426 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16427 or if measurement_type is not polynomial.
16428 * genmodes.c (ZERO_COEFFS): New macro.
16429 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16431 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16432 Use ZERO_COEFFS when emitting initializers.
16433 * data-streamer.h (bp_pack_poly_value): New function.
16434 (bp_unpack_poly_value): Likewise.
16435 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16436 for GET_MODE_NUNITS.
16437 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16438 for GET_MODE_NUNITS.
16439 * tree.c (make_vector_type): Remove temporary shim and make
16440 the real function take the number of units as a poly_uint64
16441 rather than an int.
16442 (build_vector_type_for_mode): Handle polynomial nunits.
16443 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16444 * emit-rtl.c (const_vec_series_p_1): Likewise.
16445 (gen_rtx_CONST_VECTOR): Likewise.
16446 * fold-const.c (test_vec_duplicate_folding): Likewise.
16447 * genrecog.c (validate_pattern): Likewise.
16448 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16449 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16450 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16451 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16452 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16453 * rtlanal.c (subreg_get_info): Likewise.
16454 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16455 (vect_grouped_load_supported): Likewise.
16456 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16457 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16458 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16459 (simplify_const_unary_operation, simplify_binary_operation_1)
16460 (simplify_const_binary_operation, simplify_ternary_operation)
16461 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16462 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16463 instead of CONST_VECTOR_NUNITS.
16464 * varasm.c (output_constant_pool_2): Likewise.
16465 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16466 explicit-encoded elements in the XVEC for variable-length vectors.
16468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16470 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16473 Alan Hayward <alan.hayward@arm.com>
16474 David Sherwood <david.sherwood@arm.com>
16476 * coretypes.h (fixed_size_mode): Declare.
16477 (fixed_size_mode_pod): New typedef.
16478 * builtins.h (target_builtins::x_apply_args_mode)
16479 (target_builtins::x_apply_result_mode): Change type to
16480 fixed_size_mode_pod.
16481 * builtins.c (apply_args_size, apply_result_size, result_vector)
16482 (expand_builtin_apply_args_1, expand_builtin_apply)
16483 (expand_builtin_return): Update accordingly.
16485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16487 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16488 * cselib.c (cselib_hash_rtx): Likewise.
16489 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16490 CONST_VECTOR encoding.
16492 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16493 Jeff Law <law@redhat.com>
16496 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16497 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16498 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16499 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16502 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16503 explicitly probe *sp in a noreturn function if there were any callee
16504 register saves or frame pointer is needed.
16506 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16509 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16510 BLKmode for ternary, binary or unary expressions.
16513 * var-tracking.c (delete_vta_debug_insn): New inline function.
16514 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16515 insns from get_insns () to NULL instead of each bb separately.
16516 Use delete_vta_debug_insn. No longer static.
16517 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16518 delete_vta_debug_insns callers.
16519 * rtl.h (delete_vta_debug_insns): Declare.
16520 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16521 instead of variable_tracking_main.
16523 2018-01-03 Martin Sebor <msebor@redhat.com>
16525 PR tree-optimization/83603
16526 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16527 arguments past the endof the argument list in functions declared
16528 without a prototype.
16529 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16530 Avoid checking when arguments are null.
16532 2018-01-03 Martin Sebor <msebor@redhat.com>
16535 * doc/extend.texi (attribute const): Fix a typo.
16536 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16537 issuing -Wsuggest-attribute for void functions.
16539 2018-01-03 Martin Sebor <msebor@redhat.com>
16541 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16542 offset_int::from instead of wide_int::to_shwi.
16543 (maybe_diag_overlap): Remove assertion.
16544 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16545 * gimple-ssa-sprintf.c (format_directive): Same.
16546 (parse_directive): Same.
16547 (sprintf_dom_walker::compute_format_length): Same.
16548 (try_substitute_return_value): Same.
16550 2018-01-03 Jeff Law <law@redhat.com>
16552 PR middle-end/83654
16553 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16554 non-constant residual for zero at runtime and avoid probing in
16555 that case. Reorganize code for trailing problem to mirror handling
16558 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16560 PR tree-optimization/83501
16561 * tree-ssa-strlen.c (get_string_cst): New.
16562 (handle_char_store): Call get_string_cst.
16564 2018-01-03 Martin Liska <mliska@suse.cz>
16566 PR tree-optimization/83593
16567 * tree-ssa-strlen.c: Include tree-cfg.h.
16568 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16569 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16570 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16572 (strlen_dom_walker::before_dom_children): Call
16573 gimple_purge_dead_eh_edges. Dump tranformation with details
16575 (strlen_dom_walker::before_dom_children): Update call by adding
16576 new argument cleanup_eh.
16577 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16579 2018-01-03 Martin Liska <mliska@suse.cz>
16582 * cif-code.def (VARIADIC_THUNK): New enum value.
16583 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16586 2018-01-03 Jan Beulich <jbeulich@suse.com>
16588 * sse.md (mov<mode>_internal): Tighten condition for when to use
16589 vmovdqu<ssescalarsize> for TI and OI modes.
16591 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16593 Update copyright years.
16595 2018-01-03 Martin Liska <mliska@suse.cz>
16598 * ipa-visibility.c (function_and_variable_visibility): Skip
16599 functions with noipa attribure.
16601 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16603 * gcc.c (process_command): Update copyright notice dates.
16604 * gcov-dump.c (print_version): Ditto.
16605 * gcov.c (print_version): Ditto.
16606 * gcov-tool.c (print_version): Ditto.
16607 * gengtype.c (create_file): Ditto.
16608 * doc/cpp.texi: Bump @copying's copyright year.
16609 * doc/cppinternals.texi: Ditto.
16610 * doc/gcc.texi: Ditto.
16611 * doc/gccint.texi: Ditto.
16612 * doc/gcov.texi: Ditto.
16613 * doc/install.texi: Ditto.
16614 * doc/invoke.texi: Ditto.
16616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16618 * vector-builder.h (vector_builder::m_full_nelts): Change from
16619 unsigned int to poly_uint64.
16620 (vector_builder::full_nelts): Update prototype accordingly.
16621 (vector_builder::new_vector): Likewise.
16622 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16623 (vector_builder::operator ==): Likewise.
16624 (vector_builder::finalize): Likewise.
16625 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16626 Take the number of elements as a poly_uint64 rather than an
16628 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16629 from unsigned int to poly_uint64.
16630 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16631 (vec_perm_indices::new_vector): Likewise.
16632 (vec_perm_indices::length): Likewise.
16633 (vec_perm_indices::nelts_per_input): Likewise.
16634 (vec_perm_indices::input_nelts): Likewise.
16635 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16636 number of elements per input as a poly_uint64 rather than an
16637 unsigned int. Use the original encoding for variable-length
16638 vectors, rather than clamping each individual element.
16639 For the second and subsequent elements in each pattern,
16640 clamp the step and base before clamping their sum.
16641 (vec_perm_indices::series_p): Handle polynomial element counts.
16642 (vec_perm_indices::all_in_range_p): Likewise.
16643 (vec_perm_indices_to_tree): Likewise.
16644 (vec_perm_indices_to_rtx): Likewise.
16645 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16646 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16647 (tree_vector_builder::new_binary_operation): Handle polynomial
16648 element counts. Return false if we need to know the number
16649 of elements at compile time.
16650 * fold-const.c (fold_vec_perm): Punt if the number of elements
16651 isn't known at compile time.
16653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16655 * vec-perm-indices.h (vec_perm_builder): Change element type
16656 from HOST_WIDE_INT to poly_int64.
16657 (vec_perm_indices::element_type): Update accordingly.
16658 (vec_perm_indices::clamp): Handle polynomial element_types.
16659 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16660 (vec_perm_indices::all_in_range_p): Likewise.
16661 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16663 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16664 polynomial vec_perm_indices element types.
16665 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16666 * fold-const.c (fold_vec_perm): Likewise.
16667 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16668 * tree-vect-generic.c (lower_vec_perm): Likewise.
16669 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16670 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16671 element type to HOST_WIDE_INT.
16673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16674 Alan Hayward <alan.hayward@arm.com>
16675 David Sherwood <david.sherwood@arm.com>
16677 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16678 rather than an int. Use plus_constant.
16679 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16680 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16683 Alan Hayward <alan.hayward@arm.com>
16684 David Sherwood <david.sherwood@arm.com>
16686 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16687 a HOST_WIDE_INT to a poly_int64.
16689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16690 Alan Hayward <alan.hayward@arm.com>
16691 David Sherwood <david.sherwood@arm.com>
16693 * calls.c (load_register_parameters): Cope with polynomial
16694 mode sizes. Require a constant size for BLKmode parameters
16695 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16696 forces a parameter to be padded at the lsb end in order to
16697 fill a complete number of words, require the parameter size
16698 to be ordered wrt UNITS_PER_WORD.
16700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16701 Alan Hayward <alan.hayward@arm.com>
16702 David Sherwood <david.sherwood@arm.com>
16704 * reload1.c (spill_stack_slot_width): Change element type
16705 from unsigned int to poly_uint64_pod.
16706 (alter_reg): Treat mode sizes as polynomial.
16708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16709 Alan Hayward <alan.hayward@arm.com>
16710 David Sherwood <david.sherwood@arm.com>
16712 * reload.c (complex_word_subreg_p): New function.
16713 (reload_inner_reg_of_subreg, push_reload): Use it.
16715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16716 Alan Hayward <alan.hayward@arm.com>
16717 David Sherwood <david.sherwood@arm.com>
16719 * lra-constraints.c (process_alt_operands): Reject matched
16720 operands whose sizes aren't ordered.
16721 (match_reload): Refer to this check here.
16723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16724 Alan Hayward <alan.hayward@arm.com>
16725 David Sherwood <david.sherwood@arm.com>
16727 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16728 that the mode size is in the set {1, 2, 4, 8, 16}.
16730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16731 Alan Hayward <alan.hayward@arm.com>
16732 David Sherwood <david.sherwood@arm.com>
16734 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16735 Use plus_constant instead of gen_rtx_PLUS.
16737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16738 Alan Hayward <alan.hayward@arm.com>
16739 David Sherwood <david.sherwood@arm.com>
16741 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16742 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16743 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16744 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16745 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16746 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16747 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16748 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16749 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16750 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16752 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16753 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16754 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16755 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16756 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16757 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16758 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16759 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16760 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16761 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16763 * expr.c (emit_move_resolve_push): Treat the input and result
16764 of PUSH_ROUNDING as a poly_int64.
16765 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16766 (emit_push_insn): Likewise.
16767 * lra-eliminations.c (mark_not_eliminable): Likewise.
16768 * recog.c (push_operand): Likewise.
16769 * reload1.c (elimination_effects): Likewise.
16770 * rtlanal.c (nonzero_bits1): Likewise.
16771 * calls.c (store_one_arg): Likewise. Require the padding to be
16772 known at compile time.
16774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16775 Alan Hayward <alan.hayward@arm.com>
16776 David Sherwood <david.sherwood@arm.com>
16778 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16779 Use plus_constant instead of gen_rtx_PLUS.
16781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16782 Alan Hayward <alan.hayward@arm.com>
16783 David Sherwood <david.sherwood@arm.com>
16785 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16786 rather than an int.
16788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16789 Alan Hayward <alan.hayward@arm.com>
16790 David Sherwood <david.sherwood@arm.com>
16792 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16793 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16794 via stack temporaries. Treat the mode size as polynomial too.
16796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16797 Alan Hayward <alan.hayward@arm.com>
16798 David Sherwood <david.sherwood@arm.com>
16800 * expr.c (expand_expr_real_2): When handling conversions involving
16801 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16802 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16803 as a poly_uint64 too.
16805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16806 Alan Hayward <alan.hayward@arm.com>
16807 David Sherwood <david.sherwood@arm.com>
16809 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16812 Alan Hayward <alan.hayward@arm.com>
16813 David Sherwood <david.sherwood@arm.com>
16815 * combine.c (can_change_dest_mode): Handle polynomial
16816 REGMODE_NATURAL_SIZE.
16817 * expmed.c (store_bit_field_1): Likewise.
16818 * expr.c (store_constructor): Likewise.
16819 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16820 and polynomial REGMODE_NATURAL_SIZE.
16821 (gen_lowpart_common): Likewise.
16822 * reginfo.c (record_subregs_of_mode): Likewise.
16823 * rtlanal.c (read_modify_subreg_p): Likewise.
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 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16830 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
16838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16839 Alan Hayward <alan.hayward@arm.com>
16840 David Sherwood <david.sherwood@arm.com>
16842 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16843 in a POINTER_PLUS_EXPR.
16845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16846 Alan Hayward <alan.hayward@arm.com>
16847 David Sherwood <david.sherwood@arm.com>
16849 * omp-simd-clone.c (simd_clone_subparts): New function.
16850 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16851 (ipa_simd_modify_function_body): Likewise.
16853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16854 Alan Hayward <alan.hayward@arm.com>
16855 David Sherwood <david.sherwood@arm.com>
16857 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16858 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16859 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16860 (expand_vector_condition, vector_element): Likewise.
16861 (subparts_gt): New function.
16862 (get_compute_type): Use subparts_gt.
16863 (count_type_subparts): Delete.
16864 (expand_vector_operations_1): Use subparts_gt instead of
16865 count_type_subparts.
16867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16868 Alan Hayward <alan.hayward@arm.com>
16869 David Sherwood <david.sherwood@arm.com>
16871 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16872 (vect_compile_time_alias): ...this new function. Do the calculation
16873 on poly_ints rather than trees.
16874 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16882 (vect_schedule_slp_instance): Likewise.
16884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16885 Alan Hayward <alan.hayward@arm.com>
16886 David Sherwood <david.sherwood@arm.com>
16888 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16889 constant and extern definitions for variable-length vectors.
16890 (vect_get_constant_vectors): Note that the number of units
16891 is known to be constant.
16893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16894 Alan Hayward <alan.hayward@arm.com>
16895 David Sherwood <david.sherwood@arm.com>
16897 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16898 of units as polynomial. Choose between WIDE and NARROW based
16901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16902 Alan Hayward <alan.hayward@arm.com>
16903 David Sherwood <david.sherwood@arm.com>
16905 * tree-vect-stmts.c (simd_clone_subparts): New function.
16906 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16909 Alan Hayward <alan.hayward@arm.com>
16910 David Sherwood <david.sherwood@arm.com>
16912 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16913 vectors as polynomial. Use build_index_vector for
16914 IFN_GOMP_SIMD_LANE.
16916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16917 Alan Hayward <alan.hayward@arm.com>
16918 David Sherwood <david.sherwood@arm.com>
16920 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16921 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16922 for variable-length vectors.
16923 (vectorizable_mask_load_store): Treat the number of units as
16924 polynomial, asserting that it is constant if the condition has
16925 already been enforced.
16926 (vectorizable_store, vectorizable_load): Likewise.
16928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16929 Alan Hayward <alan.hayward@arm.com>
16930 David Sherwood <david.sherwood@arm.com>
16932 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16933 of units as polynomial. Punt if we can't tell at compile time
16934 which vector contains the final result.
16936 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16937 Alan Hayward <alan.hayward@arm.com>
16938 David Sherwood <david.sherwood@arm.com>
16940 * tree-vect-loop.c (vectorizable_induction): Treat the number
16941 of units as polynomial. Punt on SLP inductions. Use an integer
16942 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16943 cast of such a series for variable-length floating-point
16946 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16947 Alan Hayward <alan.hayward@arm.com>
16948 David Sherwood <david.sherwood@arm.com>
16950 * tree.h (build_index_vector): Declare.
16951 * tree.c (build_index_vector): New function.
16952 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16953 of units as polynomial, forcibly converting it to a constant if
16954 vectorizable_reduction has already enforced the condition.
16955 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16956 to create a {1,2,3,...} vector.
16957 (vectorizable_reduction): Treat the number of units as polynomial.
16958 Choose vectype_in based on the largest scalar element size rather
16959 than the smallest number of units. Enforce the restrictions
16962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16963 Alan Hayward <alan.hayward@arm.com>
16964 David Sherwood <david.sherwood@arm.com>
16966 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16967 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16974 * target.def (autovectorize_vector_sizes): Return the vector sizes
16975 by pointer, using vector_sizes rather than a bitmask.
16976 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16977 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16978 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16980 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16981 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16982 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16983 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16984 * omp-general.c (omp_max_vf): Likewise.
16985 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16986 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16987 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16988 * tree-vect-slp.c (vect_slp_bb): Likewise.
16989 * doc/tm.texi: Regenerate.
16990 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16992 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16993 the vector size as a poly_uint64 rather than an unsigned int.
16994 (current_vector_size): Change from an unsigned int to a poly_uint64.
16995 (get_vectype_for_scalar_type): Update accordingly.
16996 * tree.h (build_truth_vector_type): Take the size and number of
16997 units as a poly_uint64 rather than an unsigned int.
16998 (build_vector_type): Add a temporary overload that takes
16999 the number of units as a poly_uint64 rather than an unsigned int.
17000 * tree.c (make_vector_type): Likewise.
17001 (build_truth_vector_type): Take the number of units as a poly_uint64
17002 rather than an unsigned int.
17004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17005 Alan Hayward <alan.hayward@arm.com>
17006 David Sherwood <david.sherwood@arm.com>
17008 * target.def (get_mask_mode): Take the number of units and length
17009 as poly_uint64s rather than unsigned ints.
17010 * targhooks.h (default_get_mask_mode): Update accordingly.
17011 * targhooks.c (default_get_mask_mode): Likewise.
17012 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17013 * doc/tm.texi: Regenerate.
17015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17016 Alan Hayward <alan.hayward@arm.com>
17017 David Sherwood <david.sherwood@arm.com>
17019 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
17020 * omp-general.c (omp_max_vf): Likewise.
17021 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
17022 (expand_omp_simd): Handle polynomial safelen.
17023 * omp-low.c (omplow_simd_context): Add a default constructor.
17024 (omplow_simd_context::max_vf): Change from int to poly_uint64.
17025 (lower_rec_simd_input_clauses): Update accordingly.
17026 (lower_rec_input_clauses): Likewise.
17028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17029 Alan Hayward <alan.hayward@arm.com>
17030 David Sherwood <david.sherwood@arm.com>
17032 * tree-vectorizer.h (vect_nunits_for_cost): New function.
17033 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
17034 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
17035 (vect_analyze_slp_cost): Likewise.
17036 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
17037 (vect_model_load_cost): Likewise.
17039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17040 Alan Hayward <alan.hayward@arm.com>
17041 David Sherwood <david.sherwood@arm.com>
17043 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
17044 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
17045 from an unsigned int * to a poly_uint64_pod *.
17046 (calculate_unrolling_factor): New function.
17047 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
17049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17050 Alan Hayward <alan.hayward@arm.com>
17051 David Sherwood <david.sherwood@arm.com>
17053 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
17054 from an unsigned int to a poly_uint64.
17055 (_loop_vec_info::slp_unrolling_factor): Likewise.
17056 (_loop_vec_info::vectorization_factor): Change from an int
17058 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
17059 (vect_get_num_vectors): New function.
17060 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
17061 (vect_get_num_copies): Use vect_get_num_vectors.
17062 (vect_analyze_data_ref_dependences): Change max_vf from an int *
17063 to an unsigned int *.
17064 (vect_analyze_data_refs): Change min_vf from an int * to a
17066 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17067 than an unsigned HOST_WIDE_INT.
17068 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
17069 (vect_analyze_data_ref_dependence): Change max_vf from an int *
17070 to an unsigned int *.
17071 (vect_analyze_data_ref_dependences): Likewise.
17072 (vect_compute_data_ref_alignment): Handle polynomial vf.
17073 (vect_enhance_data_refs_alignment): Likewise.
17074 (vect_prune_runtime_alias_test_list): Likewise.
17075 (vect_shift_permute_load_chain): Likewise.
17076 (vect_supportable_dr_alignment): Likewise.
17077 (dependence_distance_ge_vf): Take the vectorization factor as a
17078 poly_uint64 rather than an unsigned HOST_WIDE_INT.
17079 (vect_analyze_data_refs): Change min_vf from an int * to a
17081 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
17082 vfm1 as a poly_uint64 rather than an int. Make the same change
17083 for the returned bound_scalar.
17084 (vect_gen_vector_loop_niters): Handle polynomial vf.
17085 (vect_do_peeling): Likewise. Update call to
17086 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
17087 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
17089 * tree-vect-loop.c (vect_determine_vectorization_factor)
17090 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
17091 (vect_get_known_peeling_cost): Likewise.
17092 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
17093 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
17094 (vect_transform_loop): Likewise. Use the lowest possible VF when
17095 updating the upper bounds of the loop.
17096 (vect_min_worthwhile_factor): Make static. Return an unsigned int
17097 rather than an int.
17098 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
17099 polynomial unroll factors.
17100 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
17101 (vect_make_slp_decision): Likewise.
17102 (vect_supported_load_permutation_p): Likewise, and polynomial
17104 (vect_analyze_slp_cost): Handle polynomial vf.
17105 (vect_slp_analyze_node_operations): Likewise.
17106 (vect_slp_analyze_bb_1): Likewise.
17107 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
17108 than an unsigned HOST_WIDE_INT.
17109 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
17110 (vectorizable_load): Handle polynomial vf.
17111 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
17113 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
17115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17116 Alan Hayward <alan.hayward@arm.com>
17117 David Sherwood <david.sherwood@arm.com>
17119 * match.pd: Handle bit operations involving three constants
17120 and try to fold one pair.
17122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
17124 * tree-vect-loop-manip.c: Include gimple-fold.h.
17125 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17126 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17127 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17128 Add a path that uses a step of VF instead of 1, but disable it
17130 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17131 and niters_no_overflow parameters. Update calls to
17132 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17133 Create a new SSA name if the latter choses to use a ste other
17134 than zero, and return it via niters_vector_mult_vf_var.
17135 * tree-vect-loop.c (vect_transform_loop): Update calls to
17136 vect_do_peeling, vect_gen_vector_loop_niters and
17137 slpeel_make_loop_iterate_ntimes.
17138 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17139 (vect_gen_vector_loop_niters): Update declarations after above changes.
17141 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17143 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17144 128-bit round to integer instructions.
17145 (ceil<mode>2): Likewise.
17146 (btrunc<mode>2): Likewise.
17147 (round<mode>2): Likewise.
17149 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17151 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17152 unaligned VSX load/store on P8/P9.
17153 (expand_block_clear): Allow the use of unaligned VSX
17154 load/store on P8/P9.
17156 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17158 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17160 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17161 swap associated with both a load and a store.
17163 2018-01-02 Andrew Waterman <andrew@sifive.com>
17165 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17166 * config/riscv/riscv.md (clear_cache): Use it.
17168 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17170 * web.c: Remove out-of-date comment.
17172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17174 * expr.c (fixup_args_size_notes): Check that any existing
17175 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17176 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17177 (emit_single_push_insn): ...here.
17179 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17181 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17182 (const_vector_encoded_nelts): New function.
17183 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17184 (const_vector_int_elt, const_vector_elt): Declare.
17185 * emit-rtl.c (const_vector_int_elt_1): New function.
17186 (const_vector_elt): Likewise.
17187 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17188 of CONST_VECTOR_ELT.
17190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17192 * expr.c: Include rtx-vector-builder.h.
17193 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17194 directly on the tree encoding.
17195 (const_vector_from_tree): Likewise.
17196 * optabs.c: Include rtx-vector-builder.h.
17197 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17198 sequence of "u" values.
17199 * vec-perm-indices.c: Include rtx-vector-builder.h.
17200 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17201 directly on the vec_perm_indices encoding.
17203 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17205 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17206 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17207 * rtx-vector-builder.h: New file.
17208 * rtx-vector-builder.c: Likewise.
17209 * rtl.h (rtx_def::u2): Add a const_vector field.
17210 (CONST_VECTOR_NPATTERNS): New macro.
17211 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17212 (CONST_VECTOR_DUPLICATE_P): Likewise.
17213 (CONST_VECTOR_STEPPED_P): Likewise.
17214 (CONST_VECTOR_ENCODED_ELT): Likewise.
17215 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17216 (unwrap_const_vec_duplicate): Likewise.
17217 (const_vec_series_p): Check for a non-duplicated vector encoding.
17218 Say that the function only returns true for integer vectors.
17219 * emit-rtl.c: Include rtx-vector-builder.h.
17220 (gen_const_vec_duplicate_1): Delete.
17221 (gen_const_vector): Call gen_const_vec_duplicate instead of
17222 gen_const_vec_duplicate_1.
17223 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17224 (gen_const_vec_duplicate): Use rtx_vector_builder.
17225 (gen_const_vec_series): Likewise.
17226 (gen_rtx_CONST_VECTOR): Likewise.
17227 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17228 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17229 Build a new vector rather than modifying a CONST_VECTOR in-place.
17230 (handle_special_swappables): Update call accordingly.
17231 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17232 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17233 Build a new vector rather than modifying a CONST_VECTOR in-place.
17234 (handle_special_swappables): Update call accordingly.
17236 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17238 * simplify-rtx.c (simplify_const_binary_operation): Use
17239 CONST_VECTOR_ELT instead of XVECEXP.
17241 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17243 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17244 the selector elements to be different from the data elements
17245 if the selector is a VECTOR_CST.
17246 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17247 ssizetype for the selector.
17249 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17251 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17252 before testing each element individually.
17253 * tree-vect-generic.c (lower_vec_perm): Likewise.
17255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17257 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17258 * selftest-run-tests.c (selftest::run_tests): Call it.
17259 * vector-builder.h (vector_builder::operator ==): New function.
17260 (vector_builder::operator !=): Likewise.
17261 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17262 (vec_perm_indices::all_from_input_p): New function.
17263 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17264 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17265 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17266 instead of reading the VECTOR_CST directly. Detect whether both
17267 vector inputs are the same before constructing the vec_perm_indices,
17268 and update the number of inputs argument accordingly. Use the
17269 utility functions added above. Only construct sel2 if we need to.
17271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17273 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17274 the broadcast of the low byte.
17275 (expand_mult_highpart): Use an explicit encoding for the permutes.
17276 * optabs-query.c (can_mult_highpart_p): Likewise.
17277 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17278 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17279 (vectorizable_bswap): Likewise.
17280 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17281 explicit encoding for the power-of-2 permutes.
17282 (vect_permute_store_chain): Likewise.
17283 (vect_grouped_load_supported): Likewise.
17284 (vect_permute_load_chain): Likewise.
17286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17288 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17289 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17290 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17291 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17292 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17293 (vect_gen_perm_mask_any): Likewise.
17295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17297 * int-vector-builder.h: New file.
17298 * vec-perm-indices.h: Include int-vector-builder.h.
17299 (vec_perm_indices): Redefine as an int_vector_builder.
17300 (auto_vec_perm_indices): Delete.
17301 (vec_perm_builder): Redefine as a stand-alone class.
17302 (vec_perm_indices::vec_perm_indices): New function.
17303 (vec_perm_indices::clamp): Likewise.
17304 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17305 (vec_perm_indices::new_vector): New function.
17306 (vec_perm_indices::new_expanded_vector): Update for new
17307 vec_perm_indices class.
17308 (vec_perm_indices::rotate_inputs): New function.
17309 (vec_perm_indices::all_in_range_p): Operate directly on the
17310 encoded form, without computing elided elements.
17311 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17312 encoding. Update for new vec_perm_indices class.
17313 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17314 the given vec_perm_builder.
17315 (expand_vec_perm_var): Update vec_perm_builder constructor.
17316 (expand_mult_highpart): Use vec_perm_builder instead of
17317 auto_vec_perm_indices.
17318 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17319 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17320 or double series encoding as appropriate.
17321 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17322 vec_perm_indices instead of auto_vec_perm_indices.
17323 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17324 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17325 (vect_permute_store_chain): Likewise.
17326 (vect_grouped_load_supported): Likewise.
17327 (vect_permute_load_chain): Likewise.
17328 (vect_shift_permute_load_chain): Likewise.
17329 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17330 (vect_transform_slp_perm_load): Likewise.
17331 (vect_schedule_slp_instance): Likewise.
17332 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17333 (vectorizable_mask_load_store): Likewise.
17334 (vectorizable_bswap): Likewise.
17335 (vectorizable_store): Likewise.
17336 (vectorizable_load): Likewise.
17337 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17338 vec_perm_indices instead of auto_vec_perm_indices. Use
17339 tree_to_vec_perm_builder to read the vector from a tree.
17340 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17341 vec_perm_builder instead of a vec_perm_indices.
17342 (have_whole_vector_shift): Use vec_perm_builder and
17343 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17344 truncation to calc_vec_perm_mask_for_shift.
17345 (vect_create_epilog_for_reduction): Likewise.
17346 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17347 from auto_vec_perm_indices to vec_perm_indices.
17348 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17349 instead of changing individual elements.
17350 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17351 the vector in d.perm.
17352 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17353 from auto_vec_perm_indices to vec_perm_indices.
17354 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17355 instead of changing individual elements.
17356 (arm_vectorize_vec_perm_const): Use new_vector to install
17357 the vector in d.perm.
17358 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17359 Update vec_perm_builder constructor.
17360 (rs6000_expand_interleave): Likewise.
17361 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17362 (rs6000_expand_interleave): Likewise.
17364 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17366 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17367 to qimode could truncate the indices.
17368 * optabs.c (expand_vec_perm_var): Likewise.
17370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17372 * Makefile.in (OBJS): Add vec-perm-indices.o.
17373 * vec-perm-indices.h: New file.
17374 * vec-perm-indices.c: Likewise.
17375 * target.h (vec_perm_indices): Replace with a forward class
17377 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17378 * optabs.h: Include vec-perm-indices.h.
17379 (expand_vec_perm): Delete.
17380 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17381 (expand_vec_perm_const): Declare.
17382 * target.def (vec_perm_const_ok): Replace with...
17383 (vec_perm_const): ...this new hook.
17384 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17385 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17386 * doc/tm.texi: Regenerate.
17387 * optabs.def (vec_perm_const): Delete.
17388 * doc/md.texi (vec_perm_const): Likewise.
17389 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17390 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17391 expand_vec_perm for constant permutation vectors. Assert that
17392 the mode of variable permutation vectors is the integer equivalent
17393 of the mode that is being permuted.
17394 * optabs-query.h (selector_fits_mode_p): Declare.
17395 * optabs-query.c: Include vec-perm-indices.h.
17396 (selector_fits_mode_p): New function.
17397 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17398 is defined, instead of checking whether the vec_perm_const_optab
17399 exists. Use targetm.vectorize.vec_perm_const instead of
17400 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17401 fit in the vector mode before using a variable permute.
17402 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17403 vec_perm_indices instead of an rtx.
17404 (expand_vec_perm): Replace with...
17405 (expand_vec_perm_const): ...this new function. Take the selector
17406 as a vec_perm_indices rather than an rtx. Also take the mode of
17407 the selector. Update call to shift_amt_for_vec_perm_mask.
17408 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17409 Use vec_perm_indices::new_expanded_vector to expand the original
17410 selector into bytes. Check whether the indices fit in the vector
17411 mode before using a variable permute.
17412 (expand_vec_perm_var): Make global.
17413 (expand_mult_highpart): Use expand_vec_perm_const.
17414 * fold-const.c: Includes vec-perm-indices.h.
17415 * tree-ssa-forwprop.c: Likewise.
17416 * tree-vect-data-refs.c: Likewise.
17417 * tree-vect-generic.c: Likewise.
17418 * tree-vect-loop.c: Likewise.
17419 * tree-vect-slp.c: Likewise.
17420 * tree-vect-stmts.c: Likewise.
17421 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17423 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17424 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17425 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17426 (aarch64_vectorize_vec_perm_const): ...this new function.
17427 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17428 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17429 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17430 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17431 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17432 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17433 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17435 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17436 check for NEON modes.
17437 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17438 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17439 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17440 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17442 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17443 the old VEC_PERM_CONST conditions.
17444 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17445 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17446 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17447 (ia64_vectorize_vec_perm_const_ok): Merge into...
17448 (ia64_vectorize_vec_perm_const): ...this new function.
17449 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17450 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17451 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17452 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17453 * config/mips/mips.c (mips_expand_vec_perm_const)
17454 (mips_vectorize_vec_perm_const_ok): Merge into...
17455 (mips_vectorize_vec_perm_const): ...this new function.
17456 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17457 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17458 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17459 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17460 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17461 (rs6000_expand_vec_perm_const): Delete.
17462 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17464 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17465 (altivec_expand_vec_perm_const_le): Take each operand individually.
17466 Operate on constant selectors rather than rtxes.
17467 (altivec_expand_vec_perm_const): Likewise. Update call to
17468 altivec_expand_vec_perm_const_le.
17469 (rs6000_expand_vec_perm_const): Delete.
17470 (rs6000_vectorize_vec_perm_const_ok): Delete.
17471 (rs6000_vectorize_vec_perm_const): New function.
17472 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17473 an element count and rtx array.
17474 (rs6000_expand_extract_even): Update call accordingly.
17475 (rs6000_expand_interleave): Likewise.
17476 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17477 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17478 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17479 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17480 (rs6000_expand_vec_perm_const): Delete.
17481 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17482 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17483 (altivec_expand_vec_perm_const_le): Take each operand individually.
17484 Operate on constant selectors rather than rtxes.
17485 (altivec_expand_vec_perm_const): Likewise. Update call to
17486 altivec_expand_vec_perm_const_le.
17487 (rs6000_expand_vec_perm_const): Delete.
17488 (rs6000_vectorize_vec_perm_const_ok): Delete.
17489 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17490 reference to the SPE evmerge intructions.
17491 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17492 an element count and rtx array.
17493 (rs6000_expand_extract_even): Update call accordingly.
17494 (rs6000_expand_interleave): Likewise.
17495 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17496 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17498 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17500 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17502 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17503 vector mode and that that mode matches the mode of the data
17505 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17506 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17507 directly using expand_vec_perm_1 when forcing selectors into
17509 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17511 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17513 * optabs-query.h (can_vec_perm_p): Delete.
17514 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17515 * optabs-query.c (can_vec_perm_p): Split into...
17516 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17517 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17518 particular selector is valid.
17519 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17520 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17521 (vect_grouped_load_supported): Likewise.
17522 (vect_shift_permute_load_chain): Likewise.
17523 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17524 (vect_transform_slp_perm_load): Likewise.
17525 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17526 (vectorizable_bswap): Likewise.
17527 (vect_gen_perm_mask_checked): Likewise.
17528 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17529 implementations of variable permutation vectors into account
17530 when deciding which selector to use.
17531 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17532 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17533 with a false third argument.
17534 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17535 to test whether the constant selector is valid and can_vec_perm_var_p
17536 to test whether a variable selector is valid.
17538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17540 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17541 * optabs-query.c (can_vec_perm_p): Likewise.
17542 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17543 instead of vec_perm_indices.
17544 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17545 (vect_gen_perm_mask_checked): Likewise,
17546 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17547 (vect_gen_perm_mask_checked): Likewise,
17549 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17551 * optabs-query.h (qimode_for_vec_perm): Declare.
17552 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17553 (qimode_for_vec_perm): ...this new function.
17554 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17556 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17558 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17559 does not have a conditional at the top.
17561 2018-01-02 Richard Biener <rguenther@suse.de>
17563 * ipa-inline.c (big_speedup_p): Fix expression.
17565 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17568 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17571 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17575 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17576 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17577 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17578 cond_taken_branch_cost 3->4.
17580 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17582 PR tree-optimization/83581
17583 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17584 TODO_cleanup_cfg if any changes have been made.
17586 PR middle-end/83608
17587 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17588 convert_modes if target mode has the right side, but different mode
17591 PR middle-end/83609
17592 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17593 last argument when extracting from CONCAT. If either from_real or
17594 from_imag is NULL, use expansion through memory. If result is not
17595 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17596 the parts directly to inner mode, if even that fails, use expansion
17599 PR middle-end/83623
17600 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17601 check for bswap in mode rather than HImode and use that in expand_unop
17604 Copyright (C) 2018 Free Software Foundation, Inc.
17606 Copying and distribution of this file, with or without modification,
17607 are permitted in any medium without royalty provided the copyright
17608 notice and this notice are preserved.