1 2018-06-21 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/86232
4 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
5 max for constant niter.
7 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9 * config/aarch64/aarch64-simd.md
10 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
12 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
14 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15 Make opernads of the unspec commutative.
17 2018-06-21 Richard Biener <rguenther@suse.de>
19 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
20 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
21 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
22 (vect_analyze_data_ref_dependence): Re-order checks to deal with
24 (vect_record_base_alignments): Do not record base alignment
25 for gathers or scatters.
26 (vect_compute_data_ref_alignment): Drop return value that is always
27 true. Bail out early for gathers or scatters.
28 (vect_enhance_data_refs_alignment): Bail out early for gathers
30 (vect_find_same_alignment_drs): Likewise.
31 (vect_analyze_data_refs_alignment): Remove dead code.
32 (vect_slp_analyze_and_verify_node_alignment): Likewise.
33 (vect_analyze_data_refs): For possible gathers or scatters do
34 not create an alternate DR, just check their possible validity
35 and mark them. Adjust DECL_NONALIASED handling to not rely
37 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
38 update inits of gathers or scatters.
39 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
40 Also copy gather/scatter flag to pattern vinfo.
42 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
44 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
45 behavior of vec_packsu (vector unsigned long long, vector unsigned
46 long long) to match behavior of vec_packs with same signature.
48 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
49 Thomas Schwinge <thomas@codesourcery.com>
50 Cesar Philippidis <cesar@codesourcery.com>
52 * gimplify.c (gimplify_scan_omp_clauses): Add support for
53 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
54 (gimplify_adjust_omp_clauses): Likewise.
55 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
56 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
57 (gimplify_omp_target_update): Update handling of acc update and
59 * omp-low.c (install_var_field): Remove unused parameter
60 base_pointers_restrict.
61 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
62 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
64 (omp_target_base_pointers_restrict_p): Delete.
65 (scan_omp_target): Update call to scan_sharing_clauses.
66 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
68 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
69 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
70 (convert_local_omp_clauses): Likewise.
71 * tree-pretty-print.c (dump_omp_clause): Likewise.
72 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
74 (omp_clause_code_name): Likewise.
76 2018-06-20 Jakub Jelinek <jakub@redhat.com>
79 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
82 PR tree-optimization/86231
83 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
84 anti-range don't overwrite *vr0min before using it to compute *vr0max.
86 2018-06-20 Tom de Vries <tdevries@suse.de>
88 PR tree-optimization/86097
89 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
90 iv type if signedness of iv type is not the same as that of *nit.
92 2018-06-20 Jakub Jelinek <jakub@redhat.com>
94 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
95 EDGE_EH edges, verify they are all EDGE_EH.
97 2018-06-20 Maya Rashish <coypu@sdf.org>
99 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
101 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
103 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
104 * config/aarch64/aarch64.c (xgene1_tunings): Add
105 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
106 (aarch64_mode_valid_for_sched_fusion_p):
108 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
109 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
111 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
113 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
114 * config/aarch64/iterators.md (VQ2): New mode iterator.
116 2018-06-20 Martin Liska <mliska@suse.cz>
118 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
119 Change default ratio from 10 to 8.
121 2018-06-20 Martin Liska <mliska@suse.cz>
123 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
125 (bit_test_cluster::find_bit_tests): Likewise.
126 (switch_decision_tree::analyze_switch_statement): Find clusters.
127 * tree-switch-conversion.h (struct jump_table_cluster): Document
130 2018-06-20 Martin Liska <mliska@suse.cz>
132 * tree-switch-conversion.c (switch_conversion::collect):
133 Record m_uniq property.
134 (switch_conversion::expand): Bail out for special conditions.
135 (group_cluster::~group_cluster): New.
136 (group_cluster::group_cluster): Likewise.
137 (group_cluster::dump): Likewise.
138 (jump_table_cluster::emit): New.
139 (switch_decision_tree::fix_phi_operands_for_edges): New.
140 (struct case_node): Remove struct.
141 (jump_table_cluster::can_be_handled): New.
142 (case_values_threshold): Moved to header.
143 (reset_out_edges_aux): Likewise.
144 (jump_table_cluster::is_beneficial): New.
145 (bit_test_cluster::can_be_handled): Likewise.
146 (add_case_node): Remove.
147 (bit_test_cluster::is_beneficial): New.
148 (case_bit_test::cmp): New.
149 (bit_test_cluster::emit): New.
150 (expand_switch_as_decision_tree_p): Remove.
151 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
152 (fix_phi_operands_for_edge): Likewise.
153 (switch_decision_tree::analyze_switch_statement): New.
154 (compute_cases_per_edge): Move ...
155 (switch_decision_tree::compute_cases_per_edge): ... here.
156 (try_switch_expansion): Likewise.
157 (switch_decision_tree::try_switch_expansion): Likewise.
158 (record_phi_operand_mapping): Likewise.
159 (switch_decision_tree::record_phi_operand_mapping): Likewise.
160 (emit_case_decision_tree): Likewise.
161 (switch_decision_tree::emit): Likewise.
162 (balance_case_nodes): Likewise.
163 (switch_decision_tree::balance_case_nodes): Likewise.
164 (dump_case_nodes): Likewise.
165 (switch_decision_tree::dump_case_nodes): Likewise.
166 (emit_jump): Likewise.
167 (switch_decision_tree::emit_jump): Likewise.
168 (emit_cmp_and_jump_insns): Likewise.
169 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
170 (emit_case_nodes): Likewise.
171 (switch_decision_tree::emit_case_nodes): Likewise.
172 (conditional_probability): Remove.
173 * tree-switch-conversion.h (enum cluster_type): New.
175 (struct cluster): Likewise.
176 (cluster::cluster): Likewise.
177 (struct simple_cluster): Likewise.
178 (simple_cluster::simple_cluster): Likewise.
179 (struct group_cluster): Likewise.
180 (struct jump_table_cluster): Likewise.
181 (struct bit_test_cluster): Likewise.
182 (struct min_cluster_item): Likewise.
183 (struct case_tree_node): Likewise.
184 (case_tree_node::case_tree_node): Likewise.
185 (jump_table_cluster::case_values_threshold): Likewise.
186 (struct case_bit_test): Likewise.
187 (struct switch_decision_tree): Likewise.
188 (struct switch_conversion): Likewise.
189 (switch_decision_tree::reset_out_edges_aux): Likewise.
191 2018-06-20 Martin Liska <mliska@suse.cz>
193 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
194 (hoist_edge_and_branch_if_true): Likewise.
195 (expand_switch_using_bit_tests_p): Likewise.
196 (struct case_bit_test): Likewise.
197 (case_bit_test_cmp): Likewise.
198 (emit_case_bit_tests): Likewise.
199 (switch_conversion::switch_conversion): New class.
200 (struct switch_conv_info): Remove old struct.
201 (collect_switch_conv_info): More to ...
202 (switch_conversion::collect): ... this.
203 (check_range): Likewise.
204 (switch_conversion::check_range): Likewise.
205 (check_all_empty_except_final): Likewise.
206 (switch_conversion::check_all_empty_except_final): Likewise.
207 (check_final_bb): Likewise.
208 (switch_conversion::check_final_bb): Likewise.
209 (create_temp_arrays): Likewise.
210 (switch_conversion::create_temp_arrays): Likewise.
211 (free_temp_arrays): Likewise.
212 (gather_default_values): Likewise.
213 (switch_conversion::gather_default_values): Likewise.
214 (build_constructors): Likewise.
215 (switch_conversion::build_constructors): Likewise.
216 (constructor_contains_same_values_p): Likewise.
217 (switch_conversion::contains_same_values_p): Likewise.
218 (array_value_type): Likewise.
219 (switch_conversion::array_value_type): Likewise.
220 (build_one_array): Likewise.
221 (switch_conversion::build_one_array): Likewise.
222 (build_arrays): Likewise.
223 (switch_conversion::build_arrays): Likewise.
224 (gen_def_assigns): Likewise.
225 (switch_conversion::gen_def_assigns): Likewise.
226 (prune_bbs): Likewise.
227 (switch_conversion::prune_bbs): Likewise.
228 (fix_phi_nodes): Likewise.
229 (switch_conversion::fix_phi_nodes): Likewise.
230 (gen_inbound_check): Likewise.
231 (switch_conversion::gen_inbound_check): Likewise.
232 (process_switch): Use the newly created class.
233 (switch_conversion::expand): New.
234 (switch_conversion::~switch_conversion): New.
235 * tree-switch-conversion.h: New file.
237 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
239 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
240 tree-vect-patterns.c.
241 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
242 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
243 (vect_recog_sad_pattern): Likewise.
244 (vect_recog_widen_sum_pattern): Likewise.
245 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
246 (vect_recog_widen_shift_pattern): Remove the type_in argument.
247 (vect_recog_rotate_pattern): Likewise.
248 (vect_recog_mult_pattern): Likewise.
249 (vect_recog_vector_vector_shift_pattern): Likewise.
250 (vect_recog_divmod_pattern): Likewise.
251 (vect_recog_mixed_size_cond_pattern): Likewise.
252 (vect_recog_bool_pattern): Likewise.
253 (vect_recog_mask_conversion_pattern): Likewise.
254 (vect_try_gather_scatter_pattern): Likewise.
255 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
256 (vect_recog_over_widening_pattern): Likewise.
257 (vect_recog_gather_scatter_pattern): Likewise.
258 (vect_recog_func_ptr): Move from tree-vectorizer.h
259 (vect_vect_recog_func_ptrs): Move further down the file.
260 (vect_recog_func): Likewise. Remove the third argument.
261 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
262 (vect_pattern_recog_1): Expect the pattern function to do any
263 necessary target tests. Also expect it to provide a vector type.
264 Remove the type_in handling.
266 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
268 * tree-vect-patterns.c (vect_pattern_detected): New function.
269 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
270 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
271 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
272 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
273 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
274 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
275 (vect_recog_mask_conversion_pattern)
276 (vect_try_gather_scatter_pattern): Likewise.
278 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
280 * tree-vect-patterns.c (vect_get_internal_def): New function.
281 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
282 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
283 (search_type_for_mask_1): Use it.
285 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
287 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
288 redundant WIDEN_SUM_EXPR handling.
289 (vect_recog_sad_pattern): Likewise.
291 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
293 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
294 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
295 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
296 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
297 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
299 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
301 * tree-vect-stmts.c (vectorizable_call): Make sure that we
302 use the stmt_vec_info of the original bb statement for the
303 new zero assignment, even if the call is part of a pattern.
305 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
307 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
308 that the sequence is attached to the original statement rather
309 than the pattern statement.
310 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
311 PATTERN_DEF_SEQ from the original statement rather than
312 the main pattern statement.
313 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
314 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
315 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
317 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
319 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
320 definition statements before the early exit for statements that aren't
322 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
324 (vect_transform_loop): ...here. Process pattern definition
325 statements without first checking whether the main pattern
326 statement is live or relevant.
328 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
330 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
331 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
333 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
335 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
336 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
337 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
338 (expand_block_compare): Change select_block_compare_mode call.
339 (expand_strncmp_align_check): Use new functions, fix comment.
340 (emit_final_str_compare_gpr): New function.
341 (expand_strn_compare): Refactor and clean up code.
342 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
344 2018-06-19 Tony Reix <tony.reix@atos.com>
345 Damien Bergamini <damien.bergamini@atos.com>
346 David Edelsohn <dje.gcc@gmail.com>
348 * collect2.c (static_obj): New variable.
349 (static_libs): New variable.
350 (is_in_list): Uncomment declaration.
351 (main): Track AIX libraries linked statically.
352 (is_in_list): Uncomment definition.
353 (scan_prog_file): Don't add AIX shared libraries initializer
354 to constructor list if linking statically.
356 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
358 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
360 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
362 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
364 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
367 2018-06-19 Martin Liska <mliska@suse.cz>
369 * config/i386/i386.c (ix86_can_inline_p): Do not use
370 ipa_fn_summaries::get_create.
371 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
373 (devirtualization_time_bonus): Likewise.
374 (ipcp_propagate_stage): Likewise.
375 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
376 (edge_set_predicate): Likewise.
377 (evaluate_conditions_for_known_args): Likewise.
378 (evaluate_properties_for_edge): Likewise.
379 (ipa_call_summary::reset): Tranform to ...
380 (ipa_call_summary::~ipa_call_summary): ... this.
381 (ipa_fn_summary::reset): Transform to ...
382 (ipa_fn_summary::~ipa_fn_summary): ... this.
383 (ipa_fn_summary_t::remove): Rename to ...
384 (ipa_fn_summary_t::remove_callees): ... this.
385 (ipa_fn_summary_t::duplicate): Use placement new
386 instead of memory copy.
387 (ipa_call_summary_t::duplicate): Likewise.
388 (ipa_call_summary_t::remove): Remove.
389 (dump_ipa_call_summary): Change get_create to get.
390 (ipa_dump_fn_summary): Dump only when summary exists.
391 (analyze_function_body): Use symbol_summary::get instead
393 (compute_fn_summary): Likewise.
394 (estimate_edge_devirt_benefit): Likewise.
395 (estimate_edge_size_and_time): Likewise.
396 (inline_update_callee_summaries): Likewise.
397 (remap_edge_change_prob): Likewise.
398 (remap_edge_summaries): Likewise.
399 (ipa_merge_fn_summary_after_inlining): Likewise.
400 (write_ipa_call_summary): Likewise.
401 (ipa_fn_summary_write): Likewise.
402 (ipa_free_fn_summary): Likewise.
403 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
404 (struct ipa_call_summary): Likewise.
405 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
407 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
408 (estimate_size_after_inlining): Likewise.
409 (estimate_growth): Likewise.
410 (growth_likely_positive): Likewise.
411 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
412 (inline_call): Likewise.
413 * ipa-inline.c (caller_growth_limits): Likewise.
414 (can_inline_edge_p): Likewise.
415 (can_inline_edge_by_limits_p): Likewise.
416 (compute_uninlined_call_time): Likewise.
417 (compute_inlined_call_time): Likewise.
418 (want_inline_small_function_p): Likewise.
419 (edge_badness): Likewise.
420 (update_caller_keys): Likewise.
421 (update_callee_keys): Likewise.
422 (inline_small_functions): Likewise.
423 (inline_to_all_callers_1): Likewise.
424 (dump_overall_stats): Likewise.
425 (early_inline_small_functions): Likewise.
426 (early_inliner): Likewise.
427 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
428 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
429 * ipa-pure-const.c (malloc_candidate_p): Likewise.
430 * ipa-split.c (execute_split_functions): Likewise.
431 * symbol-summary.h: Likewise.
432 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
434 2018-06-19 Richard Biener <rguenther@suse.de>
436 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
437 (vectorize_loops): ... here. Fix dbgcnt handling.
438 (try_vectorize_loop): Wrap try_vectorize_loop_1.
440 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
443 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
444 ieee128 argument takes up only one (vector) register, not two (floating
447 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
449 * gimplify.c (gimplify_init_constructor): Really never clear for an
450 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
452 2018-06-19 Richard Biener <rguenther@suse.de>
454 PR tree-optimization/86179
455 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
456 after failed recognition.
458 2018-06-18 Martin Sebor <msebor@redhat.com>
461 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
462 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
463 Handle integer subtraction.
464 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
465 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
467 2018-06-18 David Malcolm <dmalcolm@redhat.com>
469 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
470 param from rtx to rtx_insn *.
471 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
473 (frv_ifcvt_modify_insn): Likwise.
474 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
475 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
476 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
477 as_a <rtx_insn *> cast to local "unprotected_region" once
478 it's been established that it's not NULL or pc_rtx.
479 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
480 param "sethi" from rtx to rtx_insn *.
481 (nds32_group_float_insns): Likewise for param "insn".
482 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
484 (vax_output_int_subtract): Likewise.
485 * config/vax/vax.c (vax_output_int_add): Likewise for param
487 (vax_output_int_subtract): Likewise.
488 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
489 (emit_pattern_after): Likewise for param "after".
490 (emit_insn_after): Likewise.
491 (emit_jump_insn_after): Likewise.
492 (emit_call_insn_after): Likewise.
493 (emit_debug_insn_after): Likewise.
494 (emit_pattern_before): Likewise for param "before".
495 (emit_insn_before): Likewise.
496 (emit_jump_insn_before): Likewise.
497 * final.c (get_insn_template): Likewise for param "insn", removing
499 * output.h (get_insn_template): Likewise for 2nd param.
500 * rtl.h (emit_insn_before): Likewise.
501 (emit_jump_insn_before): Likewise.
502 (emit_debug_insn_before_noloc): Likewise.
503 (emit_insn_after): Likewise.
504 (emit_jump_insn_after): Likewise.
505 (emit_call_insn_after): Likewise.
506 (emit_debug_insn_after): Likewise.
507 (set_insn_deleted): Likewise for param.
509 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
512 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
513 floating point modes, so that IFmode is numerically greater than
514 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
515 to declare the ordering. This prevents IFmode from being
516 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
517 machine. Include rs6000-modes.h to share the fractional values
518 between genmodes* and the rest of the compiler.
522 * config/rs6000/rs6000-modes.h: New file.
523 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
524 meaning of rs6000_long_double_size so that 126..128 selects an
525 appropriate 128-bit floating point type.
526 (rs6000_option_override_internal): Likewise.
527 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
528 (TARGET_LONG_DOUBLE_128): Change the meaning of
529 rs6000_long_double_size so that 126..128 selects an appropriate
530 128-bit floating point type.
531 (LONG_DOUBLE_TYPE_SIZE): Update comment.
532 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
533 source and destination to match the standard usage.
534 (truncifkf2): Likewise.
535 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
536 ISA 2.07 to use an explicit clobber, instead of passing in a
538 (copysign<mode>3_soft): Likewise.
540 2018-06-18 David Malcolm <dmalcolm@redhat.com>
542 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
543 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
544 (vect_slp_analyze_instance_dependence): Likewise.
545 (vect_enhance_data_refs_alignment): Likewise.
546 (vect_analyze_data_refs_alignment): Likewise.
547 (vect_slp_analyze_and_verify_instance_alignment
548 (vect_analyze_data_ref_accesses): Likewise.
549 (vect_prune_runtime_alias_test_list): Likewise.
550 (vect_analyze_data_refs): Likewise.
551 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
552 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
553 (vect_analyze_scalar_cycles_1): Likewise.
554 (vect_get_loop_niters): Likewise.
555 (vect_analyze_loop_form_1): Likewise.
556 (vect_update_vf_for_slp): Likewise.
557 (vect_analyze_loop_operations): Likewise.
558 (vect_analyze_loop): Likewise.
559 (vectorizable_induction): Likewise.
560 (vect_transform_loop): Likewise.
561 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
562 * tree-vect-slp.c (vect_analyze_slp): Likewise.
563 (vect_make_slp_decision): Likewise.
564 (vect_detect_hybrid_slp): Likewise.
565 (vect_slp_analyze_operations): Likewise.
566 (vect_slp_bb): Likewise.
567 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
568 (vectorizable_bswap): Likewise.
569 (vectorizable_call): Likewise.
570 (vectorizable_simd_clone_call): Likewise.
571 (vectorizable_conversion): Likewise.
572 (vectorizable_assignment): Likewise.
573 (vectorizable_shift): Likewise.
574 (vectorizable_operation): Likewise.
575 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
577 2018-06-18 Martin Sebor <msebor@redhat.com>
579 PR tree-optimization/81384
580 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
581 * builtins.c (expand_builtin_strnlen): New function.
582 (expand_builtin): Call it.
583 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
584 * builtins.def (BUILT_IN_STRNLEN): New.
585 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
586 Warn for bounds in excess of maximum object size.
587 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
588 single-value ranges. Handle strnlen.
589 (handle_builtin_strlen): Handle strnlen.
590 (strlen_check_and_optimize_stmt): Same.
591 * doc/extend.texi (Other Builtins): Document strnlen.
593 2018-06-18 Maya Rashish <coypu@sdf.org>
595 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
596 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
597 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
599 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
601 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
603 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
605 * tree.c (escaped_string::escape): Replace cast to char * by
606 const_cast<char *> (unescaped).
608 2018-06-18 Nick Clifton <nickc@redhat.com>
611 * tree.c (escaped_string): New class. Converts an unescaped
612 string into its escaped equivalent.
613 (warn_deprecated_use): Use the new class to convert the
614 deprecation message, if present.
615 (test_escaped_strings): New self test.
616 (test_c_tests): Add test_escaped_strings.
617 * doc/extend.texi (deprecated): Add a note that the
618 deprecation message is affected by the -fmessage-length
619 option, and that control characters will be escaped.
620 (#pragma GCC error): Document this pragma.
621 (#pragma GCC warning): Likewise.
622 * doc/invoke.texi (-fmessage-length): Document this option's
623 effect on the #warning and #error preprocessor directives and
624 the deprecated attribute.
626 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
628 * tree.c (decl_value_expr_lookup): Revert latest change.
629 (decl_value_expr_insert): Likewise.
631 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
633 * gimplify.c (nonlocal_vlas): Delete.
634 (nonlocal_vla_vars): Likewise.
635 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
637 (gimplify_body): Do not create and destroy nonlocal_vlas.
638 * tree-nested.c: Include diagnostic.h.
639 (use_pointer_in_frame): Tweak.
640 (lookup_field_for_decl): Add assertion and declare the transformation.
641 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
642 internal error when the reference is in a wrong context. Do not
643 create a debug decl by default.
644 (note_nonlocal_block_vlas): Delete.
645 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
646 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
647 create a debug decl by default.
648 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
649 call is in a wrong context.
650 (fixup_vla_decls): New function.
651 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
652 debug variables were created.
653 * tree.c (decl_value_expr_lookup): Add checking assertion.
654 (decl_value_expr_insert): Likewise.
656 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
659 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
660 * tree-scalar-evolution.c (interpret_expr): Likewise.
661 (expression_expensive_p): Likewise.
662 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
663 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
664 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
665 (ssa_defined_by_minus_one_stmt_p): New.
667 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
670 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
671 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
672 * dojump.c (do_jump): Likewise.
673 * expr.c (expand_expr_real_2): Check operand type's sign.
674 * fold-const.c (const_unop): Handle ABSU_EXPR.
675 (fold_abs_const): Likewise.
676 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
677 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
678 (strip_sign_op_1): Likesise.
679 * match.pd: Add new pattern to generate ABSU_EXPR.
680 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
681 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
682 * tree-eh.c (operation_could_trap_helper_p): Likewise.
683 * tree-inline.c (estimate_operator_cost): Likewise.
684 * tree-pretty-print.c (dump_generic_node): Likewise.
685 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
686 * tree.def (ABSU_EXPR): New.
688 2018-06-16 Jakub Jelinek <jakub@redhat.com>
691 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
692 documented as preserved for backward compatibility only.
693 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
695 PR rtl-optimization/86108
696 * bb-reorder.c (create_forwarder_block): Renamed to ...
697 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
698 jump from new landing pad to the second part.
699 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
702 2018-06-15 Jakub Jelinek <jakub@redhat.com>
705 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
706 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
707 Only call store_expr for halves if the mode is the same.
710 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
711 Fix up comment formatting.
713 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
715 * typed-splay-tree.h (typed_splay_tree::remove): New function.
716 (typed_splay_tree::closure,
717 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
718 (typed_splay_tree::typed_splay_tree,
719 typed_splay_tree::operator =): Declared private.
720 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
721 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
722 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
723 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
724 typed_splay_tree::splay_tree_splay,
725 typed_splay_tree::splay_tree_foreach_helper,
726 typed_splay_tree::splay_tree_insert,
727 typed_splay_tree::splay_tree_remove,
728 typed_splay_tree::splay_tree_lookup,
729 typed_splay_tree::splay_tree_predecessor,
730 typed_splay_tree::splay_tree_successor,
731 typed_splay_tree::splay_tree_min,
732 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
733 (typed_splay_tree::root, typed_splay_tree::comp,
734 typed_splay_tree::delete_key,
735 typed_splay_tree::delete_value): New data members.
736 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
737 typed_splay_tree::remove.
739 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
741 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
742 -mginv and -mno-ginv to the assembler.
743 * config/mips/mips.opt (-mcrc): New option.
745 * doc/invoke.text (-mcrc): Document.
748 2018-06-15 Nick Clifton <nickc@redhat.com>
751 * tree.c (escaped_string): New class. Converts an unescaped
752 string into its escaped equivalent.
753 (warn_deprecated_use): Use the new class to convert the
754 deprecation message, if present.
755 (test_escaped_strings): New self test.
756 (test_c_tests): Add test_escaped_strings.
757 * doc/extend.texi (deprecated): Add a note that the
758 deprecation message is affected by the -fmessage-length
759 option, and that control characters will be escaped.
760 (#pragma GCC error): Document this pragma.
761 (#pragma GCC warning): Likewise.
762 * doc/invoke.texi (-fmessage-length): Document this option's
763 effect on the #warning and #error preprocessor directives and
764 the deprecated attribute.
766 2018-06-15 Richard Biener <rguenther@suse.de>
768 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
769 here, also noting vector size used.
770 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
771 size used in MSG_OPTIMIZED_LOCATIONS dump.
772 (pass_slp_vectorize::execute): Adjust.
774 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
777 * config/arc/arc.c (arc_return_address_register): Fix
780 2018-06-15 Richard Biener <rguenther@suse.de>
783 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
784 leave useless conversion stripping to force_gimple_operand_gsi.
785 (gimplify_build2): Likewise.
786 (gimplify_build1): Likewise.
788 2018-06-15 Richard Biener <rguenther@suse.de>
791 * tree-cfg.c (move_stmt_op): unshare invariant addresses
792 before adjusting their block.
794 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
796 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
797 multilibs for *-*-rtems*.
798 * config/riscv/t-rtems: New file.
800 2018-06-14 Jakub Jelinek <jakub@redhat.com>
803 * match.pd ((A +- CST1) +- CST2): Punt if last resort
804 unsigned_type_for returns NULL.
807 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
808 subregs of multi-word pseudos unless the float mode has word size.
810 2018-06-14 Richard Biener <rguenther@suse.de>
813 * tree-vect-generic.c (build_word_mode_vector_type): Remove
814 duplicate and harmful type_hash_canon.
815 * tree.c (type_hash_canon): Assert we didn't find ourselves.
817 2018-06-14 Richard Biener <rguenther@suse.de>
820 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
823 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
825 * config/rtems.h (STDINT_LONG32): Define.
827 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
828 Prachi Godbole <prachi.godbole@imgtec.com>
830 * config/mips/mips-cpus.def: Define P6600.
831 * config/mips/mips-tables.opt: Regenerate.
832 * config/mips/mips.c (mips_ucbranch_type): New enum.
833 (mips_rtx_cost_data): Add support for P6600.
834 (mips_issue_rate): Likewise.
835 (mips_multipass_dfa_lookahead): Likewise.
836 (mips_avoid_hazard): Likewise.
837 (mips_reorg_process_insns): Likewise.
838 (mips_classify_branch_p6600): New function.
839 * config/mips/mips.h (TUNE_P6600): New define.
840 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
841 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
842 * config/mips/mips.md: Include p6600.md.
843 (processor): Add p6600.
844 * config/mips/p6600.md: New file.
845 * doc/invoke.texi: Add p6600 to supported architectures.
847 2018-06-13 Martin Sebor <msebor@redhat.com>
849 PR tree-optimization/86114
850 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
852 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
854 2018-06-13 Richard Biener <rguenther@suse.de>
856 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
857 Properly set vector type of the intermediate stmt.
858 * tree-vect-stmts.c (vectorizable_operation): The destination
859 var always has vectype_out type.
861 2018-06-13 Jeff Law <law@redhat.com>
863 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
864 integer 0 for argument to print_rtl_with_bb.
865 (rl78_reorg): Likewise.
867 2018-06-13 David Malcolm <dmalcolm@redhat.com>
869 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
870 from rtx to rtx_insn *.
871 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
873 (add_sched_insns_for_speculation): Likewise for local "target",
874 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
875 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
876 from rtx_insn ** to rtx_code_label **.
877 (reorg_emit_nops): Likewise.
878 (c6x_reorg): Likewise for local "call_labels".
879 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
881 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
882 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
883 the loops over LABEL_REFS.
884 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
886 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
887 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
888 (split_branches): Strengthen local "olabel" from rtx to
889 rtx_insn *, adding a safe_as_a cast.
890 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
892 (add_insn_after): Likewise for first two params.
893 (add_insn_before): Likewise.
894 (remove_insn): Likewise for param.
895 (emit_pattern_before_noloc): Likewise for second and third params.
896 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
897 (emit_call_insn_before_noloc): Likewise.
898 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
900 (emit_barrier_before): Likewise.
901 (emit_label_before): Strengthen "label" param from "rtx" to
902 "rtx_code_label *". Strengthen "before" param from "rtx" to
904 (emit_insn_after_1): Strengthen "after" param from "rtx" to
906 (emit_pattern_after_noloc): Likewise.
907 (emit_insn_after_noloc): Likewise.
908 (emit_jump_insn_after_noloc): Likewise.
909 (emit_call_insn_after_noloc): Likewise.
910 (emit_debug_insn_after_noloc): Likewise.
911 (emit_barrier_after): Likewise.
912 (emit_label_after): Likewise for both params.
913 (emit_pattern_after_setloc): Likewise for "after" param. Convert
914 "loc" param from "int" to "location_t".
915 (emit_insn_after_setloc): Likewise.
916 (emit_jump_insn_after_setloc): Likewise.
917 (emit_call_insn_after_setloc): Likewise.
918 (emit_debug_insn_after_setloc): Likewise.
919 (emit_pattern_before_setloc): Likewise for "before" param. Convert
920 "loc" param from "int" to "location_t".
921 (emit_pattern_before): Convert NULL_RTX to NULL.
922 (emit_insn_before_setloc): Convert "loc" param from "int" to
924 (emit_jump_insn_before_setloc): Likewise.
925 (emit_call_insn_before_setloc): Likewise.
926 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
927 rtx_insn *. Convert "loc" param from "int" to "location_t".
928 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
929 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
930 Convert 3rd param from "int" to "location_t".
931 (emit_barrier_before, emit_barrier_after, next_real_insn):
932 Strengthen param from rtx to rtx_insn *.
933 (emit_label_before): Strengthen 1st param from "rtx" to
934 "rtx_code_label *". Strengthen 2nd param from "rtx" to
936 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
937 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
938 Strengthen 2nd param from "rtx" to "rtx_insn *".
939 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
940 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
941 Likewise. Convert 3rd param from "int" to "location_t".
942 (emit_label_after): Strengthen 1st param from "rtx" to
944 (next_real_insn, remove_insn): Strengthen param from "rtx" to
946 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
947 from "rtx" to "rtx_insn *".
949 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
951 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
952 bodies streamed in with -Q.
953 * dumpfile.c (dump_files): Add lto-stream-out dump file.
954 * dumpfile.h (tree_dump_index): Add lto_stream_out.
955 * gimple-streamer-out.c: Include gimple-pretty-print.h
956 (output_bb): Dump stmts streamed.
957 * lto-section-out.c: Include print-tree.h
958 (lto_begin_section): Dump sections created.
959 (lto_output_decl_index): Dump decl encoded.
960 * lto-streamer-out.c: Include print-tree.h
961 (create_output_block): Dump output block created.
962 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
963 (output_function): Dump function output.
964 (output_constructor): Dump constructor streamed.
965 (write_global_stream): Output indexes encoded.
966 (produce_asm_for_decls): Dump streams encoded.
967 * lto-streamer.c (streamer_dump_file): New global var.
968 * lto-streamer.h (streamer_dump_file): Declare.
969 * passes.c (ipa_write_summaries): Initialize streamer dump.
970 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
973 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
976 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
977 offsets for register save directives. Emit a second batch of save
978 directives, if need be, when the function accesses prior frames.
980 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
982 * config/arc/fpu.md (fmasf4): Force operand to register.
985 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
987 * config/arc/arc-protos.h (arc_pad_return): Remove.
988 * config/arc/arc.c (machine_function): Remove force_short_suffix
990 (arc_print_operand): Adjust printing of '&'.
991 (arc_verify_short): Remove conditional printing of short suffix.
992 (arc_final_prescan_insn): Remove reference to size_reason.
993 (pad_return): New function.
994 (arc_reorg): Call pad_return.
995 (arc_pad_return): Remove.
996 (arc_init_machine_status): Remove reference to force_short_suffix.
997 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
998 (attr length): When attribute iscompact is true force to 2
999 regardless; in the case of maybe check if we want to force the
1000 instruction to have 4 bytes length.
1001 (nopv): Change it to generate 4 byte long nop as well.
1002 (blockage): New pattern.
1003 (simple_return): Remove call to arc_pad_return.
1004 (p_return_i): Likewise.
1006 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1008 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
1010 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1012 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
1015 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1017 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1018 for ARC700 and ARCv2.
1020 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1023 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1024 operands[2] instead of operands[1].
1027 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1029 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1030 case, check whether the outer register overlaps an unallocatable
1031 register, not just whether it fits the required class.
1033 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1035 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1037 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1038 (memrefs_conflict_p): Likewise.
1039 (init_alias_analysis): Likewise.
1040 * cfgexpand.c (expand_debug_expr): Likewise.
1041 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1042 * cse.c (fold_rtx): Likewise.
1043 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1044 * expr.c (emit_block_move_hints): Likewise.
1045 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1046 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1047 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1048 (emit_group_store): Likewise.
1049 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1050 to read the PRE/POST_MODIFY increment.
1051 * calls.c (store_one_arg): Use strip_offset.
1052 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1054 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1056 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1057 (simplify_binary_operation_1): Extend CONST_INT handling to
1059 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1060 than a HOST_WIDE_INT.
1061 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1063 (adjust_mems, add_stores): Update accodingly.
1064 (vt_canonicalize_addr): Track polynomial offsets.
1065 (emit_note_insn_var_location): Likewise.
1066 (vt_add_function_parameter): Likewise.
1067 (vt_initialize): Likewise.
1069 2018-06-12 Jeff Law <law@redhat.com>
1071 * config.gcc (alpha*-*-freebsd*): Remove.
1072 * config/alpha/freebsd.h: Remove.
1074 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1077 * spellcheck-tree.c (levenshtein_distance): Rename to...
1078 (get_edit_distance): ...this, and update for underlying renaming.
1079 * spellcheck-tree.h (levenshtein_distance): Rename to...
1080 (get_edit_distance): ...this.
1081 * spellcheck.c (levenshtein_distance): Rename to...
1082 (get_edit_distance): ...this. Convert from Levenshtein distance
1083 to Damerau-Levenshtein distance by supporting transpositions of
1084 adjacent characters. Rename "v1" to "v_next" and "v0" to
1086 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1087 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1088 update for underlying renaming.
1089 (selftest::levenshtein_distance_unit_test): Rename to...
1090 (selftest::test_get_edit_distance_unit): ...this, and update for
1091 underlying renaming.
1092 (selftest::test_find_closest_string): Add example from PR 69968
1093 where transposition helps
1094 (selftest::test_metric_conditions): Update for renaming.
1095 (selftest::test_metric_conditions): Likewise.
1096 (selftest::spellcheck_c_tests): Likewise.
1097 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1098 (get_edit_distance): ...this.
1099 (best_match::consider): Update for renaming.
1101 2018-06-12 Martin Sebor <msebor@redhat.com>
1103 PR tree-optimization/85259
1104 * builtins.c (compute_objsize): Handle constant offsets.
1105 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1106 true iff a warning has been issued.
1107 * gimple.h (gimple_nonartificial_location): New function.
1108 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1109 gimple_nonartificial_location and handle -Wno-system-headers.
1110 (handle_builtin_stxncpy): Same.
1112 2018-06-12 Martin Sebor <msebor@redhat.com>
1115 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1117 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1119 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1120 BUILTIN_VEC_XST entries for pointer to double and long long.
1122 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1125 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1127 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1130 2018-06-12 Martin Liska <mliska@suse.cz>
1132 * doc/options.texi: Document IntegerRange.
1134 2018-06-12 Martin Liska <mliska@suse.cz>
1136 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1137 * opt-functions.awk: Handle Deprecated flag.
1138 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1140 (read_cmdline_option): Report warning for a deprecated option.
1141 * opts.h (struct cl_option): Add new field cl_deprecated.
1142 (CL_ERR_DEPRECATED): New.
1144 2018-06-12 Martin Liska <mliska@suse.cz>
1146 * doc/options.texi: Document Deprecated option flag.
1148 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1150 * config/arc/arc-arch.h (arc_extras): New enum.
1151 (arc_cpu_t):Add field extra.
1152 (arc_cpu_types): Consider the extras.
1153 * config/arc/arc-cpus.def: Add extras info.
1154 * config/arc/arc-opts.h (processor_type): Consider extra field.
1155 * config/arc/arc.c (arc_override_options): Handle extra field.
1157 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1159 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1160 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1162 * config/arc/arc.c (arc_init): Update pic warning.
1163 (irq_range): Update irq range parsing warnings.
1164 (arc_override_options): Update various warning messages.
1165 (arc_handle_aux_attribute): Likewise.
1167 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1169 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1171 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1173 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1174 compiler messages directives.
1176 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1178 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1179 * config/mips/mips-tables.opt: Regenerate.
1180 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1182 * doc/invoke.texi: Document -march=i6500.
1184 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1186 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1187 (i6400_gpmul): Add cpu_unit.
1188 (i6400_gpdiv): Likewise.
1189 (i6400_msa_add_d): Update reservations.
1190 (i6400_msa_int_add) Likewise.
1191 (i6400_msa_short_logic3) Likewise.
1192 (i6400_msa_short_logic2) Likewise.
1193 (i6400_msa_short_logic) Likewise.
1194 (i6400_msa_move) Likewise.
1195 (i6400_msa_cmp) Likewise.
1196 (i6400_msa_short_float2) Likewise.
1197 (i6400_msa_div_d) Likewise.
1198 (i6400_msa_long_logic1) Likewise.
1199 (i6400_msa_long_logic2) Likewise.
1200 (i6400_msa_mult) Likewise.
1201 (i6400_msa_long_float2) Likewise.
1202 (i6400_msa_long_float4) Likewise.
1203 (i6400_msa_long_float5) Likewise.
1204 (i6400_msa_long_float8) Likewise.
1205 (i6400_fpu_fadd): Include frint type.
1206 (i6400_fpu_store): New define_insn_reservation.
1207 (i6400_fpu_load): Likewise.
1208 (i6400_fpu_move): Likewise.
1209 (i6400_fpu_fcmp): Likewise.
1210 (i6400_fpu_fmadd): Likewise.
1211 (i6400_int_mult): Include imul3nc type and update reservation.
1212 (i6400_int_div): Include idiv3 type and update reservation.
1213 (i6400_int_load): Update to check type not move_type.
1214 (i6400_int_store): Likewise.
1215 (i6400_int_prefetch): Set zero latency.
1217 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1219 * gcc.c: Document new %@{...} sequence.
1220 (LINK_COMMAND_SPEC): Use it for the -L switches.
1221 (cpp_unique_options): Use it for the -I switches.
1222 (at_file_argbuf): New global variable.
1223 (in_at_file): Likewise.
1224 (alloc_args): Create at_file_argbuf.
1225 (clear_args): Truncate at_file_argbuf.
1226 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1227 (open_at_file): New function.
1228 (close_at_file): Likewise.
1229 (create_at_file): Delete.
1230 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1233 (validate_switches_from_spec): Deal with %@{...} sequence.
1234 (validate_switches): Likewise.
1235 (driver::finalize): Call clear_args.
1237 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1239 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1241 2018-06-11 Martin Sebor <msebor@redhat.com>
1243 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1244 (Wno-class-memaccess): Add @opindex.
1245 (Wno-templates, Wno-multiple-inheritance): Same.
1246 (Wno-virtual-inheritance, Wno-namespaces): Same.
1247 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1248 (Wno-format-overflow, Wno-format-truncation): Same.
1249 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1250 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1251 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1252 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1253 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1254 (Wno-misspelled-isr): Same.
1256 2018-06-11 Martin Sebor <msebor@redhat.com>
1258 * PR tree-optimization/86083
1259 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1261 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1263 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1265 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1268 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1269 on the correct operand.
1270 (*movdi_internal64): Ditto.
1272 2018-06-11 Martin Liska <mliska@suse.cz>
1274 PR tree-optimization/86089
1275 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1277 2018-06-11 Julia Koval <julia.koval@intel.com>
1279 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1280 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1281 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1283 2018-06-11 Olivier Hainque <hainque@adacore.com>
1285 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1286 for Ada with strict dwarf2.
1288 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1291 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1294 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1296 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1298 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1300 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1303 2018-06-08 Martin Liska <mliska@suse.cz>
1305 * tree-cfg.h (debug_function): Fix argument type to match
1308 2018-06-08 Martin Liska <mliska@suse.cz>
1310 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1311 Remove usage of MPX-related (and removed) fields.
1312 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1314 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1316 * cfg.c (debug): Use TDF_NONE rather than 0.
1317 * cfghooks.c (debug): Likewise.
1318 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1319 (struct dump_option_value_info): Convert to...
1320 (struct kv_pair): ...this template type.
1321 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1323 (optinfo_verbosity_options): Likewise.
1324 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1326 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1327 than int for "optgroup_flags" param.
1328 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1330 (dump_dec): Likewise.
1331 (dump_finish): Use TDF_NONE rather than 0.
1332 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1333 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1334 than 0. Update for change to option_ptr.
1335 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1336 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1337 0. Update for changes to optinfo_verbosity_options and
1339 (opt_info_switch_p): Convert optgroup_flags from int to
1341 (dump_basic_block): Use dump_flags_t rather than int
1342 for "dump_kind" param.
1343 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1344 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1345 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1346 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1347 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1348 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1349 TDF_NONE): Convert from macros to...
1350 (enum dump_flag): ...this new enum.
1351 (dump_flags_t): Update to use enum.
1352 (operator|, operator&, operator~, operator|=, operator&=):
1353 Implement for dump_flags_t.
1354 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1355 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1356 Convert from macros to...
1357 (enum optgroup_flag): ...this new enum.
1358 (optgroup_flags_t): New typedef.
1359 (operator|, operator|=): Implement for optgroup_flags_t.
1360 (struct dump_file_info): Convert field "alt_flags" to
1361 dump_flags_t. Convert field "optgroup_flags" to
1363 (dump_basic_block): Use dump_flags_t rather than int for param.
1364 (dump_generic_expr_loc): Likewise.
1365 (dump_dec): Likewise.
1366 (dump_register): Convert param "optgroup_flags" to
1368 (opt_info_enable_passes): Likewise.
1369 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1371 * gimple-pretty-print.c (debug): Likewise.
1372 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1373 (merged_store_group::apply_stores): Likewise.
1374 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1375 * gimple.c (verify_gimple_pp): Likewise.
1376 * graphite-poly.c (print_pbb_body): Likewise.
1377 * passes.c (pass_manager::register_one_dump_file): Convert
1378 local "optgroup_flags" to optgroup_flags_t.
1379 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1381 (debug_body): Likewise.
1382 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1383 to optgroup_flags_t.
1384 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1386 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1387 (convert_mult_to_fma): Likewise.
1388 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1389 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1390 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1391 "dump_kind" to dump_flags_t.
1393 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1395 * config/rs6000/rs6000.c (min, max): Delete.
1397 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1399 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1400 -mabi=spe and -mabi=no-spe.
1402 2018-06-08 Martin Liska <mliska@suse.cz>
1404 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1405 where we expect an existing summary.
1407 2018-06-08 Martin Liska <mliska@suse.cz>
1409 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1410 * ipa-inline.h (estimate_edge_growth): Likewise.
1412 2018-06-08 Martin Liska <mliska@suse.cz>
1414 * cgraph.c (function_version_hasher::hash): Use
1415 cgraph_node::get_uid ().
1416 (function_version_hasher::equal):
1417 * cgraph.h (cgraph_node::get_uid): New method.
1418 * ipa-inline.c (update_caller_keys): Use
1419 cgraph_node::get_uid ().
1420 (update_callee_keys): Likewise.
1421 * ipa-utils.c (searchc): Likewise.
1422 (ipa_reduced_postorder): Likewise.
1423 * lto-cgraph.c (input_node): Likewise.
1424 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1425 * symbol-summary.h (symtab_insertion): Likewise.
1426 (symtab_removal): Likewise.
1427 (symtab_duplication): Likewise.
1428 * tree-pretty-print.c (dump_function_header): Likewise.
1429 * tree-sra.c (convert_callers_for_node): Likewise.
1431 2018-06-08 Martin Liska <mliska@suse.cz>
1433 * cgraph.c (symbol_table::create_edge): Always assign a new
1435 (symbol_table::free_edge): Do not recycle numbers.
1436 * cgraph.h (cgraph_edge::get): New method.
1437 * symbol-summary.h (symtab_removal): Use it.
1438 (symtab_duplication): Likewise.
1439 (call_summary::hashable_uid): Remove.
1441 2018-06-08 Martin Liska <mliska@suse.cz>
1443 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1444 (initialize_growth_caches): Remove.
1445 (free_growth_caches): Likewise.
1446 (do_estimate_edge_time): Use edge_growth_cache.
1447 (do_estimate_edge_size): Likewise.
1448 (do_estimate_edge_hints): Likewise.
1449 * ipa-inline.c (reset_edge_caches): Likewise.
1450 (recursive_inlining): Likewise.
1451 (inline_small_functions): Likewise.
1452 * ipa-inline.h (initialize_growth_caches): Remove.
1453 (estimate_edge_size): Likewise.
1454 (estimate_edge_time): Likewise.
1455 (estimate_edge_hints): Likewise.
1456 (reset_edge_growth_cache): Likewise.
1457 * symbol-summary.h (call_summary::remove): New method.
1459 2018-06-08 Martin Liska <mliska@suse.cz>
1461 * ipa-cp.c (class edge_clone_summary): New summary.
1462 (grow_edge_clone_vectors): Remove.
1463 (ipcp_edge_duplication_hook): Remove.
1464 (class edge_clone_summary_t): New call_summary class.
1465 (ipcp_edge_removal_hook): Remove.
1466 (edge_clone_summary_t::duplicate): New function.
1467 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1468 (create_specialized_node): Likewise.
1469 (ipcp_driver): Initialize edge_clone_summaries and do not
1472 2018-06-08 Martin Liska <mliska@suse.cz>
1474 * symbol-summary.h (get): New function.
1475 (call_summary::m_initialize_when_cloning): New class member.
1477 2018-06-08 Martin Liska <mliska@suse.cz>
1479 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1480 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1481 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1482 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1483 (remove_cgraph_node_from_order): Use the removed_nodes set.
1484 (do_per_function_toporder): Likwise.
1485 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1486 instead of summary_uid.
1487 (symtab_removal): Likewise.
1488 (symtab_duplication): Likewise.
1490 2018-06-08 Martin Liska <mliska@suse.cz>
1492 * ipa-cp.c (ipcp_store_bits_results): Use
1493 ipcp_transformation_sum.
1494 (ipcp_store_vr_results): Likewise.
1495 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1497 (ipcp_transformation_initialize): ... this.
1498 (ipa_set_node_agg_value_chain):
1499 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1500 (write_ipcp_transformation_info): Likewise.
1501 (read_ipcp_transformation_info): Likewise.
1502 (ipcp_update_bits): Likewise.
1503 (ipcp_update_vr): Likewise.
1504 (ipcp_transform_function): Likewise.
1505 * ipa-prop.h: Rename ipcp_transformation_summary to
1506 ipcp_transformation.
1507 (class ipcp_transformation_t): New function summary.
1508 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1509 (ipa_get_agg_replacements_for_node): Likewise.
1511 2018-06-08 Martin Liska <mliska@suse.cz>
1513 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1515 (class funct_state_summary_t): New function_summary class.
1516 (has_function_state): Remove.
1517 (get_function_state): Likewise.
1518 (set_function_state): Likewise.
1519 (add_new_function): Likewise.
1520 (funct_state_summary_t::insert): New function.
1521 (duplicate_node_data): Remove.
1522 (remove_node_data): Remove.
1523 (funct_state_summary_t::duplicate): New function.
1524 (register_hooks): Create new funct_state_summaries.
1525 (pure_const_generate_summary): Use it.
1526 (pure_const_write_summary): Likewise.
1527 (pure_const_read_summary): Likewise.
1528 (propagate_pure_const): Likewise.
1529 (propagate_nothrow): Likewise.
1530 (dump_malloc_lattice): Likewise.
1531 (propagate_malloc): Likewise.
1532 (execute): Do not register hooks, just remove summary
1534 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1537 2018-06-08 Martin Liska <mliska@suse.cz>
1539 * ipa-reference.c (remove_node_data): Remove.
1540 (duplicate_node_data): Likewise.
1541 (class ipa_ref_var_info_summary_t): New class.
1542 (class ipa_ref_opt_summary_t): Likewise.
1543 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1544 (get_reference_optimization_summary): Use
1545 ipa_ref_opt_sum_summaries.
1546 (set_reference_vars_info): Remove.
1547 (set_reference_optimization_summary): Likewise.
1548 (ipa_init): Create summaries.
1549 (init_function_info): Use function summary.
1550 (ipa_ref_opt_summary_t::duplicate): New function.
1551 (ipa_ref_opt_summary_t::remove): New function.
1552 (get_read_write_all_from_node): Fix GNU coding style.
1553 (propagate): Use function summary.
1554 (write_node_summary_p): Fix GNU coding style.
1555 (stream_out_bitmap): Likewise.
1556 (ipa_reference_read_optimization_summary): Use function summary.
1557 (ipa_reference_c_finalize): Do not release hooks.
1559 2018-06-08 Martin Liska <mliska@suse.cz>
1561 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1562 (analyze_function_body): Extract multiple calls of get_create.
1563 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1564 * ipa-inline.c (recursive_inlining): Use ::get method.
1565 * ipa-inline.h (estimate_edge_growth): Likewise.
1567 2018-06-08 Martin Liska <mliska@suse.cz>
1569 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1571 (hsa_function_summary::hsa_function_summary): Use the new enum
1573 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1574 * hsa-gen.c (hsa_get_host_function): Likewise.
1575 (get_brig_function_name): Likewise.
1576 * ipa-hsa.c (process_hsa_functions): Likewise.
1577 (ipa_hsa_write_summary): Likewise.
1578 * symbol-summary.h (symtab_duplication): Use ::get function/
1579 (get): New function.
1581 2018-06-08 Martin Liska <mliska@suse.cz>
1583 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1585 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1586 (hsa_register_kernel): Likewise.
1587 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1588 * hsa-gen.c (hsa_get_host_function): Likewise.
1589 (get_brig_function_name): Likewise.
1590 (generate_hsa): Likewise.
1591 (pass_gen_hsail::execute): Likewise.
1592 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1593 (devirtualization_time_bonus): Likewise.
1594 (ipcp_propagate_stage): Likewise.
1595 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1596 (edge_set_predicate): Likewise.
1597 (evaluate_conditions_for_known_args): Likewise.
1598 (evaluate_properties_for_edge): Likewise.
1599 (ipa_fn_summary::reset): Likewise.
1600 (ipa_fn_summary_t::duplicate): Likewise.
1601 (dump_ipa_call_summary): Likewise.
1602 (ipa_dump_fn_summary): Likewise.
1603 (analyze_function_body): Likewise.
1604 (compute_fn_summary): Likewise.
1605 (estimate_edge_devirt_benefit): Likewise.
1606 (estimate_edge_size_and_time): Likewise.
1607 (estimate_calls_size_and_time): Likewise.
1608 (estimate_node_size_and_time): Likewise.
1609 (inline_update_callee_summaries): Likewise.
1610 (remap_edge_change_prob): Likewise.
1611 (remap_edge_summaries): Likewise.
1612 (ipa_merge_fn_summary_after_inlining): Likewise.
1613 (ipa_update_overall_fn_summary): Likewise.
1614 (read_ipa_call_summary): Likewise.
1615 (inline_read_section): Likewise.
1616 (write_ipa_call_summary): Likewise.
1617 (ipa_fn_summary_write): Likewise.
1618 (ipa_free_fn_summary): Likewise.
1619 * ipa-hsa.c (process_hsa_functions): Likewise.
1620 (ipa_hsa_write_summary): Likewise.
1621 (ipa_hsa_read_section): Likewise.
1622 * ipa-icf.c (sem_function::merge): Likewise.
1623 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1624 (do_estimate_edge_time): Likewise.
1625 (estimate_size_after_inlining): Likewise.
1626 (estimate_growth): Likewise.
1627 (growth_likely_positive): Likewise.
1628 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1629 (inline_call): Likewise.
1630 * ipa-inline.c (caller_growth_limits): Likewise.
1631 (can_inline_edge_p): Likewise.
1632 (can_inline_edge_by_limits_p): Likewise.
1633 (compute_uninlined_call_time): Likewise.
1634 (compute_inlined_call_time): Likewise.
1635 (want_inline_small_function_p): Likewise.
1636 (edge_badness): Likewise.
1637 (update_caller_keys): Likewise.
1638 (update_callee_keys): Likewise.
1639 (recursive_inlining): Likewise.
1640 (inline_small_functions): Likewise.
1641 (inline_to_all_callers_1): Likewise.
1642 (dump_overall_stats): Likewise.
1643 (early_inline_small_functions): Likewise.
1644 (early_inliner): Likewise.
1645 * ipa-inline.h (estimate_edge_growth): Likewise.
1646 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1647 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1648 * ipa-prop.h (IPA_NODE_REF): Likewise.
1649 (IPA_EDGE_REF): Likewise.
1650 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1651 (propagate_malloc): Likewise.
1652 * ipa-split.c (execute_split_functions): Likewise.
1653 * symbol-summary.h: Rename get to get_create.
1655 (get_create): Likewise.
1656 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1658 2018-06-08 Martin Liska <mliska@suse.cz>
1660 * symbol-summary.h (release): Move definition out of class
1662 (symtab_removal): Likewise.
1663 (symtab_duplication): Likewise.
1665 2018-06-08 Martin Liska <mliska@suse.cz>
1667 * symbol-summary.h (function_summary): Move constructor
1668 implementation out of class declaration.
1669 (release): Likewise.
1670 (symtab_insertion): Likewise.
1671 (symtab_removal): Likewise.
1672 (symtab_duplication): Likewise.
1675 2018-06-08 Martin Liska <mliska@suse.cz>
1677 * Makefile.in: Remove support for MPX (macros, related functions,
1678 fields in cgraph_node, ...).
1679 * builtin-types.def (BT_BND): Likewise.
1680 (BT_FN_BND_CONST_PTR): Likewise.
1681 (BT_FN_CONST_PTR_BND): Likewise.
1682 (BT_FN_VOID_PTR_BND): Likewise.
1683 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1684 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1685 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1686 (expand_builtin_mempcpy_with_bounds): Likewise.
1687 (expand_builtin_memset_with_bounds): Likewise.
1688 (expand_builtin_memset_args): Likewise.
1689 (std_expand_builtin_va_start): Likewise.
1690 (expand_builtin): Likewise.
1691 (expand_builtin_with_bounds): Likewise.
1692 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1693 (DEF_LIB_BUILTIN_CHKP): Likewise.
1694 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1695 (DEF_CHKP_BUILTIN): Likewise.
1696 (BUILT_IN_MEMCPY): Likewise.
1697 (BUILT_IN_MEMMOVE): Likewise.
1698 (BUILT_IN_MEMPCPY): Likewise.
1699 (BUILT_IN_MEMSET): Likewise.
1700 (BUILT_IN_STPCPY): Likewise.
1701 (BUILT_IN_STRCAT): Likewise.
1702 (BUILT_IN_STRCHR): Likewise.
1703 (BUILT_IN_STRCPY): Likewise.
1704 (BUILT_IN_STRLEN): Likewise.
1705 (BUILT_IN_MEMCPY_CHK): Likewise.
1706 (BUILT_IN_MEMMOVE_CHK): Likewise.
1707 (BUILT_IN_MEMPCPY_CHK): Likewise.
1708 (BUILT_IN_MEMSET_CHK): Likewise.
1709 (BUILT_IN_STPCPY_CHK): Likewise.
1710 (BUILT_IN_STRCAT_CHK): Likewise.
1711 (BUILT_IN_STRCPY_CHK): Likewise.
1712 * calls.c (store_bounds): Likewise.
1713 (emit_call_1): Likewise.
1714 (special_function_p): Likewise.
1715 (maybe_warn_nonstring_arg): Likewise.
1716 (initialize_argument_information): Likewise.
1717 (finalize_must_preallocate): Likewise.
1718 (compute_argument_addresses): Likewise.
1719 (expand_call): Likewise.
1720 * cfgexpand.c (expand_call_stmt): Likewise.
1721 (expand_return): Likewise.
1722 (expand_gimple_stmt_1): Likewise.
1723 (pass_expand::execute): Likewise.
1724 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1725 (cgraph_node::remove): Likewise.
1726 (cgraph_node::dump): Likewise.
1727 (cgraph_node::verify_node): Likewise.
1728 * cgraph.h (chkp_function_instrumented_p): Likewise.
1729 (symtab_node::get_alias_target): Likewise.
1730 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1731 (cgraph_local_p): Likewise.
1732 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1733 (cgraph_edge::rebuild_references): Likewise.
1734 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1735 (walk_polymorphic_call_targets): Likewise.
1736 (cgraph_node::expand_thunk): Likewise.
1737 (symbol_table::output_weakrefs): Likewise.
1738 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1739 (ix86_handle_option): Likewise.
1740 * config/i386/constraints.md: Likewise.
1741 * config/i386/i386-builtin-types.def (BND): Likewise.
1745 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1746 (BDESC_FIRST): Likewise.
1748 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1749 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1750 * config/i386/i386.c (enum reg_class): Likewise.
1751 (ix86_target_string): Likewise.
1752 (ix86_option_override_internal): Likewise.
1753 (ix86_conditional_register_usage): Likewise.
1754 (ix86_valid_target_attribute_inner_p): Likewise.
1755 (ix86_set_indirect_branch_type): Likewise.
1756 (ix86_set_current_function): Likewise.
1757 (ix86_function_arg_regno_p): Likewise.
1758 (init_cumulative_args): Likewise.
1759 (ix86_function_arg_advance): Likewise.
1760 (ix86_function_arg): Likewise.
1761 (ix86_pass_by_reference): Likewise.
1762 (ix86_function_value_regno_p): Likewise.
1763 (ix86_function_value_1): Likewise.
1764 (ix86_function_value_bounds): Likewise.
1765 (ix86_return_in_memory): Likewise.
1766 (ix86_setup_incoming_vararg_bounds): Likewise.
1767 (ix86_va_start): Likewise.
1768 (indirect_thunk_need_prefix): Likewise.
1769 (print_reg): Likewise.
1770 (ix86_print_operand): Likewise.
1771 (ix86_expand_call): Likewise.
1772 (ix86_output_function_return): Likewise.
1773 (reg_encoded_number): Likewise.
1774 (BDESC_VERIFYS): Likewise.
1775 (ix86_init_mpx_builtins): Likewise.
1776 (ix86_init_builtins): Likewise.
1777 (ix86_emit_cmove): Likewise.
1778 (ix86_emit_move_max): Likewise.
1779 (ix86_expand_builtin): Likewise.
1780 (ix86_builtin_mpx_function): Likewise.
1781 (ix86_get_arg_address_for_bt): Likewise.
1782 (ix86_load_bounds): Likewise.
1783 (ix86_store_bounds): Likewise.
1784 (ix86_load_returned_bounds): Likewise.
1785 (ix86_store_returned_bounds): Likewise.
1786 (ix86_class_likely_spilled_p): Likewise.
1787 (ix86_hard_regno_mode_ok): Likewise.
1788 (x86_order_regs_for_local_alloc): Likewise.
1789 (ix86_mitigate_rop): Likewise.
1790 (ix86_bnd_prefixed_insn_p): Likewise.
1791 (ix86_mpx_bound_mode): Likewise.
1792 (ix86_make_bounds_constant): Likewise.
1793 (ix86_initialize_bounds): Likewise.
1794 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1795 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1796 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1797 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1798 (TARGET_CHKP_BOUND_MODE): Likewise.
1799 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1800 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1801 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1802 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1803 * config/i386/i386.h (TARGET_MPX): Likewise.
1804 (TARGET_MPX_P): Likewise.
1805 (VALID_BND_REG_MODE): Likewise.
1806 (FIRST_BND_REG): Likewise.
1807 (LAST_BND_REG): Likewise.
1808 (enum reg_class): Likewise.
1809 (BND_REG_P): Likewise.
1810 (BND_REGNO_P): Likewise.
1811 (BNDmode): Likewise.
1812 (ADJUST_INSN_LENGTH): Likewise.
1813 * config/i386/i386.md: Likewise.
1814 * config/i386/i386.opt: Likewise.
1815 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1816 (defined): Likewise.
1817 (LINK_MPX): Likewise.
1818 (MPX_SPEC): Likewise.
1819 (LIBMPX_SPEC): Likewise.
1820 (LIBMPXWRAPPERS_SPEC): Likewise.
1821 (CHKP_SPEC): Likewise.
1822 * config/i386/predicates.md: Likewise.
1823 * dbxout.c (dbxout_type): Likewise.
1824 * doc/extend.texi: Likewise.
1825 * doc/invoke.texi: Likewise.
1826 * doc/md.texi: Likewise.
1827 * doc/tm.texi: Likewise.
1828 * doc/tm.texi.in: Likewise.
1829 * dwarf2out.c (is_base_type): Likewise.
1830 (gen_formal_types_die): Likewise.
1831 (gen_subprogram_die): Likewise.
1832 (gen_type_die_with_usage): Likewise.
1833 (gen_decl_die): Likewise.
1834 (dwarf2out_late_global_decl): Likewise.
1835 * expr.c (expand_assignment): Likewise.
1836 (emit_storent_insn): Likewise.
1837 (store_expr_with_bounds): Likewise.
1838 (store_expr): Likewise.
1839 (expand_expr_real_1): Likewise.
1840 * expr.h (store_expr_with_bounds): Likewise.
1841 * function.c (use_register_for_decl): Likewise.
1842 (struct bounds_parm_data): Likewise.
1843 (assign_parms_augmented_arg_list): Likewise.
1844 (assign_parm_find_entry_rtl): Likewise.
1845 (assign_parm_is_stack_parm): Likewise.
1846 (assign_parm_load_bounds): Likewise.
1847 (assign_bounds): Likewise.
1848 (assign_parms): Likewise.
1849 (expand_function_start): Likewise.
1850 * gcc.c (CHKP_SPEC): Likewise.
1851 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1852 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1853 (wrestrict_dom_walker::check_call): Likewise.
1854 * gimple.c (gimple_build_call_from_tree): Likewise.
1855 * gimple.h (enum gf_mask): Likewise.
1856 (gimple_call_with_bounds_p): Likewise.
1857 (gimple_call_set_with_bounds): Likewise.
1858 * gimplify.c (gimplify_init_constructor): Likewise.
1859 * ipa-cp.c (initialize_node_lattices): Likewise.
1860 (propagate_constants_across_call): Likewise.
1861 (find_more_scalar_values_for_callers_subset): Likewise.
1862 * ipa-hsa.c (process_hsa_functions): Likewise.
1863 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1864 * ipa-icf.c (sem_function::merge): Likewise.
1865 * ipa-inline.c (early_inliner): Likewise.
1866 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1867 (warn_function_cold): Likewise.
1868 (propagate_pure_const): Likewise.
1869 * ipa-ref.h (enum GTY): Likewise.
1870 * ipa-split.c (find_retbnd): Likewise.
1871 (consider_split): Likewise.
1872 (split_function): Likewise.
1873 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1874 * ipa.c (walk_polymorphic_call_targets): Likewise.
1875 (symbol_table::remove_unreachable_nodes): Likewise.
1876 (process_references): Likewise.
1877 (cgraph_build_static_cdtor_1): Likewise.
1878 * lto-cgraph.c (lto_output_node): Likewise.
1879 (output_refs): Likewise.
1880 (compute_ltrans_boundary): Likewise.
1881 (input_overwrite_node): Likewise.
1882 (input_node): Likewise.
1883 (input_cgraph_1): Likewise.
1884 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1885 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1886 (class pass_chkp_instrumentation_passes): Likewise.
1887 (make_pass_chkp_instrumentation_passes): Likewise.
1888 * passes.def: Likewise.
1889 * rtl.h (struct GTY): Likewise.
1890 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1891 * stor-layout.c (layout_type): Likewise.
1892 * symtab.c: Likewise.
1893 * target.def: Likewise.
1894 * targhooks.c (default_chkp_bound_type): Likewise.
1895 (default_chkp_bound_mode): Likewise.
1896 (default_builtin_chkp_function): Likewise.
1897 (default_chkp_function_value_bounds): Likewise.
1898 (default_chkp_make_bounds_constant): Likewise.
1899 (default_chkp_initialize_bounds): Likewise.
1900 * targhooks.h (default_chkp_bound_type): Likewise.
1901 (default_chkp_bound_mode): Likewise.
1902 (default_builtin_chkp_function): Likewise.
1903 (default_chkp_function_value_bounds): Likewise.
1904 (default_chkp_make_bounds_constant): Likewise.
1905 (default_chkp_initialize_bounds): Likewise.
1906 * toplev.c (compile_file): Likewise.
1907 (process_options): Likewise.
1908 * tree-core.h (DEF_BUILTIN): Likewise.
1909 (DEF_BUILTIN_CHKP): Likewise.
1910 * tree-inline.c (declare_return_variable): Likewise.
1911 (remap_gimple_stmt): Likewise.
1912 (copy_bb): Likewise.
1913 (initialize_inlined_parameters): Likewise.
1914 (expand_call_inline): Likewise.
1915 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1916 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1917 (make_pass_ipa_chkp_produce_thunks): Likewise.
1918 (make_pass_chkp): Likewise.
1919 (make_pass_chkp_opt): Likewise.
1920 (make_pass_chkp_instrumentation_passes): Likewise.
1921 * tree-pretty-print.c (dump_generic_node): Likewise.
1922 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1923 * tree-ssa-dce.c (propagate_necessity): Likewise.
1924 (eliminate_unnecessary_stmts): Likewise.
1925 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1926 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1927 * tree-ssa-sccvn.h: Likewise.
1928 * tree-ssa-strlen.c (get_string_length): Likewise.
1929 (valid_builtin_call): Likewise.
1930 (adjust_last_stmt): Likewise.
1931 (handle_builtin_strchr): Likewise.
1932 (handle_builtin_strcpy): Likewise.
1933 (handle_builtin_stxncpy): Likewise.
1934 (handle_builtin_memcpy): Likewise.
1935 (handle_builtin_strcat): Likewise.
1936 (strlen_check_and_optimize_stmt): Likewise.
1937 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1938 * tree-streamer-in.c: Likewise.
1939 * tree-streamer.c (record_common_node): Likewise.
1940 * tree.c (tree_code_size): Likewise.
1941 (wide_int_to_tree_1): Likewise.
1942 (type_contains_placeholder_1): Likewise.
1943 (build_common_tree_nodes): Likewise.
1944 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1945 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1946 (POINTER_BOUNDS_P): Likewise.
1947 (BOUNDED_TYPE_P): Likewise.
1948 (BOUNDED_P): Likewise.
1949 (CALL_WITH_BOUNDS_P): Likewise.
1950 (pointer_bounds_type_node): Likewise.
1951 * value-prof.c (gimple_ic): Likewise.
1952 * var-tracking.c (vt_add_function_parameters): Likewise.
1953 * varasm.c (make_decl_rtl): Likewise.
1954 (assemble_start_function): Likewise.
1955 (output_constant): Likewise.
1956 (maybe_assemble_visibility): Likewise.
1957 * varpool.c (ctor_for_folding): Likewise.
1958 * chkp-builtins.def: Remove.
1959 * ipa-chkp.c: Remove.
1960 * ipa-chkp.h: Remove.
1961 * rtl-chkp.c: Remove.
1962 * rtl-chkp.h: Remove.
1963 * tree-chkp-opt.c: Remove.
1964 * tree-chkp.c: Remove.
1965 * tree-chkp.h: Remove.
1967 2018-06-07 Carl Love <cel@us.ibm.com>
1969 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1970 vextract_fp_from_shortl): Add BE support.
1972 2018-06-07 Paul Koning <ni1d@arrl.net>
1974 * compare-elim.c (try_merge_compare): Don't merge compare if
1975 address contains a side effect.
1976 (try_eliminate_compare): Likewise.
1978 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1980 * config.gcc: Support "tremont".
1981 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1982 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1984 * config/i386/i386.c (m_TREMONT): Define.
1985 (processor_target_table): Add "tremont".
1986 (PTA_TREMONT): Define.
1987 (ix86_lea_outperforms): Add TARGET_TREMONT.
1988 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1989 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1990 and M_INTEL_GOLDMONT_PLUS.
1991 (fold_builtin_cpu): Add "tremont".
1992 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1993 (ix86_option_override_internal): Add "tremont".
1994 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1995 (processor_type): Add PROCESSOR_TREMONT.
1996 * config/i386/x86-tune.def: Add m_TREMONT.
1997 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1999 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2001 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
2002 symbol defined for msp430i* devices to be lower case.
2004 gcc/testsuite/gcc.target/msp430/
2005 * msp430i-device-symbol.c: New test.
2006 * msp430f-device-symbol.c: New test.
2007 * msp430.h: New test header file.
2009 2018-06-07 Richard Biener <rguenther@suse.de>
2011 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
2012 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
2013 Properly wrap signed arithmetic if overflow wraps.
2015 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2017 PR tree-optimization/69615
2018 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2019 of a cast from a same precision integral SSA_NAME in a bb dominated
2020 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2021 cast to utype if rhs2 has already a compatible type.
2023 2018-06-07 Richard Biener <rguenther@suse.de>
2025 PR tree-optimization/85935
2026 * graphite-scop-detection.c (find_params_in_bb): Analyze
2027 condition operands with respect to the correct loop. Assert
2028 the analysis doesn't fail.
2030 2018-06-04 Carl Love <cel@us.ibm.com>
2032 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2033 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2034 as it is slightly cheaper.
2035 (first_match_or_eos_index_<mode>):
2036 Calculate index using natural element order.
2037 (first_match_index_<mode>):
2038 Calculate index using natural element order.
2039 (first_match_or_eos_index_<mode>):
2040 Calculate index using natural order.
2041 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2042 for BE and LE modes.
2043 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2044 P9V_BUILTIN_VCLZLSBB_V16QI.
2045 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2048 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2050 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2051 indentation and line wrap for many prototypes. Add missing
2052 @smallexample directives around block of prototypes for vec_xl and
2055 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2057 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2058 track if we pass or return IEEE 128-bit floating point.
2059 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2060 C++ mangling that is compatible with GCC 8.1.
2061 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2062 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2063 floating point types.
2064 (rs6000_function_arg_advance_1): Likewise.
2065 (rs6000_mangle_type): Optionally generate mangled names that match
2066 what GCC 8.1 generated for IEEE 128-bit floating point types.
2067 (rs6000_globalize_decl_name): If we have an external function that
2068 passes or returns IEEE 128-bit types, generate a weak reference
2069 from the mangled name used in GCC 8.1 to the current mangled
2071 (rs6000_init_builtins): Make __ibm128 use the long double type if
2072 long double is IBM extended double. Make __float128 use the long
2073 double type if long double is IEEE 128-bit.
2076 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2077 macro for __ibm128 built-in functions.
2078 (PACK_IF): Add __ibm128 pack/unpack functions.
2079 (UNPACK_IF): Likewise.
2080 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2081 enable long double built-in functions if long double is IEEE
2082 128-bit floating point.
2083 (rs6000_invalid_builtin): Update long double built-in function
2085 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2086 functions, adjust the built-in function to use the long double
2087 built-in function if __ibm128 and long double are the same type.
2088 * doc/extend.texi (PowerPC builtins): Update documention for
2089 __builtin_{,un}pack_longdouble. Add documentation for
2090 __builtin_{,un}pack_ibm128.
2092 2018-06-06 Jim Wilson <jimw@sifive.com>
2094 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2095 (struct machine_function): New field interrupt_mode.
2096 (riscv_handle_type_attribute): New function. Add forward declaration.
2097 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2098 (riscv_expand_epilogue): Check interrupt_mode field.
2099 (riscv_set_current_function): Check interrupt attribute args and
2100 set interrupt_mode field.
2101 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2102 (riscv_sret, riscv_uret): New.
2103 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2104 new arguments to interrupt attribute.
2106 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2109 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2110 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2112 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2114 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2115 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2116 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2117 several redundant entries.
2119 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2121 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2122 type from "rtx" to "rtx_insn *".
2123 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2124 for local "call_insn", removing cast.
2125 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2127 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2129 PR tree-optimization/86066
2130 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2131 for BIT_INSERT_EXPR stores.
2133 2018-06-06 Richard Biener <rguenther@suse.de>
2135 PR tree-optimization/86062
2136 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2137 component refs ontop
2138 of to be offsetted base.
2140 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2142 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2143 to be static and remove check on interrupt attribute name.
2145 gcc/testsuite/gcc.target/msp430/
2146 * function-attributes-4.c: New test.
2147 * static-interrupts.c: New test.
2149 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2151 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2152 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2154 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2157 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2159 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2160 Remove second argument, change how error is called.
2161 (aarch64_layout_arg): Remove second argument from
2162 aarch64_err_no_fpadvsimd call.
2163 (aarch64_init_cumulative_args): Ditto.
2164 (aarch64_gimplify_va_arg_expr): Ditto.
2165 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2167 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2169 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2170 (*simple_return_indirect_internal<mode>): Rename from
2171 simple_return_indirect_internal. Use W mode iterator.
2172 (rstorssp): New expander.
2173 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2174 (clrssbsy): New expander.
2175 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2177 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2179 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2181 (cmse_check_pointed_object): Likewise.
2183 2018-06-05 Martin Liska <mliska@suse.cz>
2185 PR gcov-profile/47618
2186 * doc/invoke.texi: Document how -fprofile-dir format
2189 2018-06-05 Richard Biener <rguenther@suse.de>
2191 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2192 removal pretend DOM info isn't available so we do not update
2193 it and only remove edges, not dominated blocks. Actually free
2194 DOM info in case we removed something. Remove unreachable blocks.
2195 (mfb_keep_latches): Work with either DOM info or marked backedges.
2196 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2197 first. Mark backedges if DOM info isn't available.
2198 (Re-)compute DOM info after cleanup_control_flow_pre.
2200 2018-06-05 Richard Biener <rguenther@suse.de>
2202 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2203 (locus_discrim_hasher::hash): Adjust.
2204 (locus_discrim_hasher::equal): Likewise.
2205 (next_discriminator_for_locus): Work on line directly.
2206 (same_line_p): Pass in expanded locus1 as well.
2207 (assign_discriminators): Avoid redundant location expansions.
2209 2018-06-05 Richard Biener <rguenther@suse.de>
2211 PR tree-optimization/86046
2212 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2213 if required after clearing TREE_ADDRESSABLE.
2215 2018-06-05 Richard Biener <rguenther@suse.de>
2217 PR tree-optimization/86047
2218 * tree-ssa-loop.c (for_each_index): Glob handling of all
2219 decls and constants and really handle all of them.
2221 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2224 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2225 qualifier_void_pointer and qualifier_const_void_pointer.
2226 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2227 (arm_init_builtins): Handle the above.
2228 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2229 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2230 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2233 2018-06-05 Martin Liska <mliska@suse.cz>
2235 * auto-profile.c (read_autofdo_file): Do not use
2236 gcov_ctr_summary struct.
2237 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2238 * coverage.c (struct counts_entry): Likewise.
2239 (read_counts_file): Read just single summary entry.
2240 (get_coverage_counts): Use gcov_summary struct.
2241 * coverage.h (get_coverage_counts): Likewise.
2242 * gcov-dump.c (dump_working_sets): Likewise.
2243 (tag_summary): Dump just single summary.
2244 * gcov-io.c (gcov_write_summary): Write just histogram
2246 (gcov_read_summary): Read just single summary.
2247 (compute_working_sets): Use gcov_summary struct.
2248 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2249 of GCOV_COUNTERS_SUMMABLE.
2250 (GCOV_COUNTERS_SUMMABLE): Remove.
2251 (GCOV_FIRST_VALUE_COUNTER): Replace with
2252 GCOV_COUNTER_V_INTERVAL.
2253 (struct gcov_ctr_summary): Remove.
2254 (struct gcov_summary): Directly use fields of former
2256 (compute_working_sets): Use gcov_summary struct.
2257 * gcov.c (read_count_file): Do not use ctrs fields.
2258 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2260 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2262 * profile.c: Likewise.
2263 * profile.h: Likewise.
2265 2018-06-05 Martin Liska <mliska@suse.cz>
2267 PR gcov-profile/84846
2268 * gcov.c (output_lines): Print working directory only
2269 in intermediate format.
2271 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2273 * config/s390/s390-builtin-types.def: Add void function type.
2274 * config/s390/s390-builtins.def: Use the function type for the
2277 2018-06-04 Jim Wilson <jimw@sifive.com>
2279 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2281 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2282 and maybe_eh_return. Change regno to unsigned int. Use new args to
2283 handle EH_RETURN_DATA_REGNO registers properly.
2284 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2285 (riscv_expand_epilogue): Update comment. Change argument name and
2286 type. Update code to use new name and type. Pass new args to
2287 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2289 * config/riscv/riscv.md (NORMAL_RETURN): New.
2290 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2291 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2292 (eh_return): Call gen_eh_return_internal and emit barrier.
2293 (eh_return_internal): Call riscv_expand_epilogue.
2295 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2297 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2298 bit_insertion field and declare can_be_merged_into method.
2299 (merged_store_group::can_be_merged_into): New method.
2300 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2301 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2302 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2304 2018-06-04 Richard Biener <rguenther@suse.de>
2306 PR tree-optimization/85955
2307 * builtins.c (fold_builtin_sincos): Convert pointers to
2308 destination to appropriate type before dereferencing.
2310 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2312 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2314 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2316 * expr.c (expand_expr_real_1): Force the operand into memory if
2317 its TYPE_MODE is BLKmode and if there is no integer mode for
2318 the number of bits being extracted.
2320 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2324 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2325 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2327 2018-06-04 Richard Biener <rguenther@suse.de>
2329 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2330 (cleanup_tree_cfg_noloop): ... single caller. Do
2331 start_recording_case_labels later.
2333 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2335 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2336 to _IMMINTRIN_H_INCLUDED.
2337 * config/i386/pconfigintrin.h: Ditto.
2338 * config/i386/waitpkgintrin.h: Ditto.
2339 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2340 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2341 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2342 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2343 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2344 waitpkgintrin.h and cldemoteintrin.h.
2346 2018-06-04 Richard Biener <rguenther@suse.de>
2348 PR tree-optimization/86038
2349 * tracer.c (find_best_successor): Check probability for
2350 being initialized, bail out if not.
2352 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2355 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2356 of bits to ignore when comparing architectures.
2358 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2360 PR tree-optimization/69615
2361 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2362 maximum or minimum of the type, try to merge it also as if
2363 range1 is + [-, x - 1] or + [x + 1, -].
2366 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2368 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2370 PR tree-optimization/86034
2371 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2372 the unsigned bitfield type in a bit insertion sequence if it does not
2373 have a larger precision than the bitfield size.
2374 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2376 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2378 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2380 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2382 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2383 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2384 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2385 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2387 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2389 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2390 Disable -fdelete-null-pointer-checks for ELF toolchain.
2392 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2393 Kito Cheng <kito.cheng@gmail.com>
2395 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2396 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2397 (nds32*-*-*): Add glibc and uclibc conditions.
2398 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2399 (TARGET_EXCEPT_UNWIND_INFO): Define.
2400 * config/nds32/elf.h: New file.
2401 * config/nds32/linux.h: New file.
2402 * config/nds32/nds32-elf.opt: New file.
2403 * config/nds32/nds32-linux.opt: New file.
2404 * config/nds32/nds32-fp-as-gp.c
2405 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2406 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2408 (nds32_asm_file_end): Ditto.
2409 (nds32_print_operand): Ditto.
2410 (nds32_insert_attributes): Ditto.
2411 (nds32_init_libfuncs): New function.
2412 (TARGET_HAVE_TLS): Define.
2413 (TARGET_INIT_LIBFUNCS): Define.
2414 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2416 (TARGET_ELF): Apply different mcmodel setting.
2417 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2418 been migrated into elf.h and linux.h files.
2419 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2420 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2421 (mcmodel): The content has been migrated into nds32-elf.opt and
2422 nds32-linux.opt files.
2423 * config/nds32/t-elf: New file.
2424 * config/nds32/t-linux: New file.
2426 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2427 Shiva Chen <shiva0217@gmail.com>
2429 * config/nds32/constants.md (unspec_volatile_element): Add
2430 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2431 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2433 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2434 (make_pass_nds32_fp_as_gp): Declare.
2435 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2437 (nds32_asm_function_end_prologue): Remove unused asm output.
2438 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2439 (nds32_asm_file_start): Output necessary fp_as_gp information.
2440 (nds32_option_override): Adjust register usage.
2441 (nds32_expand_prologue): Consider fp_as_gp situation.
2442 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2443 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2446 (simple_return): Ditto.
2447 (omit_fp_begin): Output special directive for fp_as_gp.
2448 (omit_fp_end): Output special directive for fp_as_gp.
2449 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2450 mforbid-fp-as-gp): New options.
2452 2018-06-01 Mark Wielaard <mark@klomp.org>
2454 * dwarf2out.c (dwarf2out_finish): Remove generation of
2455 DW_AT_loclists_base.
2457 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2459 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2460 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2461 (struct merged_store_group): Add bit_insertion field.
2462 (dump_char_array): Use standard hexadecimal format.
2463 (merged_store_group::merged_store_group): Set bit_insertion to false.
2464 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2465 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2466 also print the mask in the dump file.
2467 (pass_store_merging::gate): Minor tweak.
2468 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2469 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2470 stores with INTEGER_CST stores.
2471 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2472 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2473 and use it throughout. Generate bit insertion sequences if need be.
2474 (pass_store_merging::process_store): Remove redundant condition.
2475 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2477 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2479 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2480 the 128-bit floating point types. Fix function comment.
2482 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2484 * config/aarch64/aarch64-simd.md
2485 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2487 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2490 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2492 PR tree-optimization/85989
2493 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2495 (backprop::intersect_uses): Check it when deciding whether this
2496 is a backedge reference.
2497 (backprop::process_block): Add each phi to m_visited_phis
2498 after visiting it, then clear it at the end.
2500 2018-06-01 Richard Biener <rguenther@suse.de>
2502 * tree-vectorizer.h (vect_dr_stmt): New function.
2503 (vect_get_load_cost): Adjust.
2504 (vect_get_store_cost): Likewise.
2505 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2506 Use vect_dr_stmt instead of DR_SMTT.
2507 (vect_record_base_alignments): Likewise.
2508 (vect_calculate_target_alignment): Likewise.
2509 (vect_compute_data_ref_alignment): Likewise and make static.
2510 (vect_update_misalignment_for_peel): Likewise.
2511 (vect_verify_datarefs_alignment): Likewise.
2512 (vector_alignment_reachable_p): Likewise.
2513 (vect_get_data_access_cost): Likewise. Pass down
2514 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2515 (vect_get_peeling_costs_all_drs): Likewise.
2516 (vect_peeling_hash_get_lowest_cost): Likewise.
2517 (vect_enhance_data_refs_alignment): Likewise.
2518 (vect_find_same_alignment_drs): Likewise.
2519 (vect_analyze_data_refs_alignment): Likewise.
2520 (vect_analyze_group_access_1): Likewise.
2521 (vect_analyze_group_access): Likewise.
2522 (vect_analyze_data_ref_access): Likewise.
2523 (vect_analyze_data_ref_accesses): Likewise.
2524 (vect_vfa_segment_size): Likewise.
2525 (vect_small_gap_p): Likewise.
2526 (vectorizable_with_step_bound_p): Likewise.
2527 (vect_prune_runtime_alias_test_list): Likewise.
2528 (vect_analyze_data_refs): Likewise.
2529 (vect_supportable_dr_alignment): Likewise.
2530 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2531 (vect_gen_prolog_loop_niters): Likewise.
2532 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2533 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2535 (vect_recog_mask_conversion_pattern): Likewise.
2536 (vect_try_gather_scatter_pattern): Likewise.
2537 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2538 to vect_get_store_cost.
2539 (vect_get_store_cost): Get stmt_info instead of DR.
2540 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2541 (vect_get_load_cost): Get stmt_info instead of DR.
2543 2018-06-01 Richard Biener <rguenther@suse.de>
2546 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2547 vars wrapped in MEM_REFs.
2549 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2551 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2552 Fix subreg tests so that we only return a choice between
2553 GENERAL_REGS and FP_REGS if the original classes included both.
2555 2018-06-01 Richard Biener <rguenther@suse.de>
2558 * tree-ssa-structalias.c (get_function_part_constraint):
2559 Handle NULL fi->decl.
2560 (find_func_aliases_for_call): Properly handle indirect
2561 fi from direct call.
2562 (find_func_clobbers): Likewise.
2563 (ipa_pta_execute): Likewise.
2564 (create_variable_info_for): For functions that are ifunc_resolver
2565 resolve to a varinfo that contains the result of the resolver call.
2566 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2569 2018-05-31 Michael Collison <michael.collison@arm.com>
2571 * config/aarch64/aarch64.md:
2572 (*fix_to_zero_extenddfdi2): New pattern.
2573 * gcc.target/aarch64/fix_extend1.c: New testcase.
2575 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2579 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2580 and BUILT_IN_STRNCMP_EQ.
2581 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2582 BUILT_IN_STRNCMP_EQ.
2583 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2584 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2585 (gimple_fold_builtin): Likewise.
2586 * tree-ssa-strlen.c (compute_string_length): New function.
2587 (determine_min_obsize): New function.
2588 (handle_builtin_string_cmp): New function to handle calls to
2589 string compare functions.
2590 (strlen_optimize_stmt): Add handling to builtin string compare
2592 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2593 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2594 * tree.c (build_common_builtin_nodes): Add new defines of
2595 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2597 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2600 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2601 functions with naked attribute.
2603 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2605 * config/i386/sse.md (avx_vec_concat<mode>):
2606 Substitute concat_tg_mode mode attribute with xtg_mode.
2607 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2608 (concat_tg_mode): Remove mode attribute.
2610 2018-05-31 Martin Sebor <msebor@redhat.com>
2613 * calls.c (alloc_max_size): Correct a logic error/typo.
2614 Treat excessive arguments as infinite. Warn for invalid arguments.
2615 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2617 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2620 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2621 and movx for Haswell.
2623 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2624 Cesar Philippidis <cesar@codesourcery.com>
2627 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2628 when emitting error on private/firstprivate reductions.
2629 * omp-low.c (lower_omp_target): Avoid reference-type processing
2630 on pointers for firstprivate clause.
2632 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2634 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2637 * config/aarch64/aarch64-simd.md
2638 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2639 (aarch64_ld1_x3_<mode>): Likewise
2640 (aarch64_st1x2<VALLDIF:mode>): Likewise
2641 (aarch64_st1_x2_<mode>): Likewise
2642 (aarch64_st1x3<VALLDIF:mode>): Likewise
2643 (aarch64_st1_x3_<mode>): Likewise
2644 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2645 (vld1_s8_x3): Likewise.
2646 (vld1_u16_x3): Likewise.
2647 (vld1_s16_x3): Likewise.
2648 (vld1_u32_x3): Likewise.
2649 (vld1_s32_x3): Likewise.
2650 (vld1_u64_x3): Likewise.
2651 (vld1_s64_x3): Likewise.
2652 (vld1_f16_x3): Likewise.
2653 (vld1_f32_x3): Likewise.
2654 (vld1_f64_x3): Likewise.
2655 (vld1_p8_x3): Likewise.
2656 (vld1_p16_x3): Likewise.
2657 (vld1_p64_x3): Likewise.
2658 (vld1q_u8_x3): Likewise.
2659 (vld1q_s8_x3): Likewise.
2660 (vld1q_u16_x3): Likewise.
2661 (vld1q_s16_x3): Likewise.
2662 (vld1q_u32_x3): Likewise.
2663 (vld1q_s32_x3): Likewise.
2664 (vld1q_u64_x3): Likewise.
2665 (vld1q_s64_x3): Likewise.
2666 (vld1q_f16_x3): Likewise.
2667 (vld1q_f32_x3): Likewise.
2668 (vld1q_f64_x3): Likewise.
2669 (vld1q_p8_x3): Likewise.
2670 (vld1q_p16_x3): Likewise.
2671 (vld1q_p64_x3): Likewise.
2672 (vst1_s64_x2): Likewise.
2673 (vst1_u64_x2): Likewise.
2674 (vst1_f64_x2): Likewise.
2675 (vst1_s8_x2): Likewise.
2676 (vst1_p8_x2): Likewise.
2677 (vst1_s16_x2): Likewise.
2678 (vst1_p16_x2): Likewise.
2679 (vst1_s32_x2): Likewise.
2680 (vst1_u8_x2): Likewise.
2681 (vst1_u16_x2): Likewise.
2682 (vst1_u32_x2): Likewise.
2683 (vst1_f16_x2): Likewise.
2684 (vst1_f32_x2): Likewise.
2685 (vst1_p64_x2): Likewise.
2686 (vst1q_s8_x2): Likewise.
2687 (vst1q_p8_x2): Likewise.
2688 (vst1q_s16_x2): Likewise.
2689 (vst1q_p16_x2): Likewise.
2690 (vst1q_s32_x2): Likewise.
2691 (vst1q_s64_x2): Likewise.
2692 (vst1q_u8_x2): Likewise.
2693 (vst1q_u16_x2): Likewise.
2694 (vst1q_u32_x2): Likewise.
2695 (vst1q_u64_x2): Likewise.
2696 (vst1q_f16_x2): Likewise.
2697 (vst1q_f32_x2): Likewise.
2698 (vst1q_f64_x2): Likewise.
2699 (vst1q_p64_x2): Likewise.
2700 (vst1_s64_x3): Likewise.
2701 (vst1_u64_x3): Likewise.
2702 (vst1_f64_x3): Likewise.
2703 (vst1_s8_x3): Likewise.
2704 (vst1_p8_x3): Likewise.
2705 (vst1_s16_x3): Likewise.
2706 (vst1_p16_x3): Likewise.
2707 (vst1_s32_x3): Likewise.
2708 (vst1_u8_x3): Likewise.
2709 (vst1_u16_x3): Likewise.
2710 (vst1_u32_x3): Likewise.
2711 (vst1_f16_x3): Likewise.
2712 (vst1_f32_x3): Likewise.
2713 (vst1_p64_x3): Likewise.
2714 (vst1q_s8_x3): Likewise.
2715 (vst1q_p8_x3): Likewise.
2716 (vst1q_s16_x3): Likewise.
2717 (vst1q_p16_x3): Likewise.
2718 (vst1q_s32_x3): Likewise.
2719 (vst1q_s64_x3): Likewise.
2720 (vst1q_u8_x3): Likewise.
2721 (vst1q_u16_x3): Likewise.
2722 (vst1q_u32_x3): Likewise.
2723 (vst1q_u64_x3): Likewise.
2724 (vst1q_f16_x3): Likewise.
2725 (vst1q_f32_x3): Likewise.
2726 (vst1q_f64_x3): Likewise.
2727 (vst1q_p64_x3): Likewise.
2729 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2731 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2732 user_label_prefix to name.
2734 * tree-core.h: Update comment about the format of NAME string
2735 passed to handler in attribute_spec.
2737 * config/msp430/msp430.md: Remove erroneous subreg expression from
2738 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2739 zero_extend{q,h}isi2.
2741 2018-05-30 Borislav Petkov <bp@suse.de>
2743 * doc/extend.texi: Document some architecture specific
2744 constraints and sort entries.
2746 2018-05-30 Martin Sebor <msebor@redhat.com>
2749 * builtins.c (expand_builtin_stpcpy_1): New function.
2750 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2751 only if the former succeeds.
2753 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2755 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2758 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2760 * doc/invoke.texi (-flinker-output): Document
2762 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2764 * passes.c (ipa_write_summaries): Only modify statements if body
2766 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2767 incrementally linking.
2768 (ipa_passes): Likewise.
2769 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2770 pass down resolution info.
2771 * common.opt (flag_incremental_link): Update info.
2772 * gcc.c (plugin specs): Turn flinker-output=* to
2773 -plugin-opt=-linker-output-known
2774 * toplev.c (compile_file): Also cut compilation when doing incremental
2776 * flag-types. (enum lto_partition_model): Add
2777 LTO_LINKER_OUTPUT_NOLTOREL.
2778 (invoke.texi): Add -flinker-output docs.
2779 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2780 link same way as WPA; do not stream in dead initializers.
2782 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2783 darf2out_register_external_decl): Support incremental link.
2785 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2787 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2789 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2791 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2792 it down to simple_object_copy_lto_debug_sections.
2793 (run_gcc): Determine incremental LTO link time and configure
2794 lto1 into non-wpa mode, disable renaming of debug sections.
2796 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2798 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2799 descriptions of various incorrectly documented functions.
2801 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2804 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2805 address check not strict.
2807 2018-05-30 Richard Biener <rguenther@suse.de>
2809 PR tree-optimization/85964
2810 * tracer.c (better_p): Drop initialized count check, we only
2811 call the function with initialized counts now.
2812 (find_best_successor): Do find a best edge if one
2813 has uninitialized count.
2814 (find_best_predecessor): Likewise. Do BB frequency check only
2815 if count is initialized.
2817 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2819 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2820 (aarch64_ldrstr_offset_compare): New.
2821 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2822 load/store orderings.
2823 (aarch64_gen_adjusted_ldpstp): Likewise.
2825 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2827 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2828 Check for subset of GENERAL_REGS and FP_REGS.
2829 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2832 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2834 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2835 and wi::to_poly_offset. Add the current offset and then check
2836 whether the sum fits, rather than using an unchecked addition of
2837 a checked term. Check for a shwi rather than a uhwi.
2838 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2839 (store_constructor): Use poly_int_tree_p.
2840 (expand_expr_real_1): Likewise.
2841 * function.c (assign_temp): Likewise.
2842 * fold-const.c (const_binop): Use poly_int_tree_p and
2844 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2846 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2847 to_poly_offset for MEM offsets.
2848 * ipa-icf.c (sem_variable::equals): Likewise.
2849 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2850 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2851 wi::to_poly_offset for BIT_FIELD_REF offsets.
2852 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2854 * var-tracking.c (emit_note_insn_var_location): Use
2855 tree_to_poly_uint64.
2857 2018-05-29 Jim Wilson <jimw@sifive.com>
2859 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2861 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2864 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2865 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2867 (sse4_1_round<mode>2): Use nonimmediate_operand
2868 for operand 1 predicate.
2870 2018-05-29 Martin Sebor <msebor@redhat.com>
2871 Richard Biener <rguenther@suse.de>
2874 * calls.c (get_size_range): Call determine_value_range instead
2875 of get_value_range..
2876 * tree-vrp.h (determine_value_range): Declared new function.
2877 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2879 2018-05-29 Richard Biener <rguenther@suse.de>
2881 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2882 sure to use non-pattern stmts for get_earlier_stmt arguments.
2883 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2884 called on pattern stmts.
2885 (get_later_stmt): Likewise.
2887 2018-05-29 Martin Liska <mliska@suse.cz>
2889 PR gcov-profile/85759
2890 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2893 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2895 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2897 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2901 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2902 VEC_PACK_FLOAT_EXPR): New tree codes.
2903 * tree-pretty-print.c (op_code_prio): Handle
2904 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2905 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2906 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2907 * tree-inline.c (estimate_operator_cost): Likewise.
2908 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2909 * fold-const.c (const_binop): Likewise.
2910 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2911 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2912 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2913 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2914 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2915 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2916 * expr.c (expand_expr_real_2): Likewise.
2917 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2918 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2919 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2921 * optabs.c (expand_widen_pattern_expr): For
2922 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2923 sign from result type rather than operand's type.
2924 (expand_binop_directly): For vec_packu_float_optab and
2925 vec_packs_float_optab allow result type to be different from operand's
2927 * optabs-tree.c (optab_for_tree_code): Handle
2928 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2929 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2930 * tree-vect-generic.c (expand_vector_operations_1): Handle
2931 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2932 VEC_PACK_FLOAT_EXPR.
2933 * tree-vect-stmts.c (supportable_widening_operation): Handle
2935 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2936 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2937 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2938 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2939 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2941 (vec_pack<floatprefix>_float_<mode>): New expander.
2942 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2944 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2945 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2946 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2947 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2948 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2950 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2951 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2952 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2953 VEC_PACK_FLOAT_EXPR): Document.
2955 2018-05-29 Richard Biener <rguenther@suse.de>
2957 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2959 (stmt_vec_info_vec): Make pointer.
2960 (init_stmt_vec_info_vec): Remove.
2961 (free_stmt_vec_info_vec): Likewise.
2962 (set_stmt_vec_info_vec): New function.
2963 (free_stmt_vec_infos): Likewise.
2964 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2965 (set_vinfo_for_stmt): Likewise.
2966 (get_earlier_stmt): Likewise.
2967 (get_later_stmt): Likewise.
2968 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2969 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2970 (vec_info::~vec_info): Free stmt_vec_infos.
2971 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2972 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2973 (pass_slp_vectorize::execute): Likewise.
2974 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2975 (free_stmt_vec_info_vec): Likewise.
2976 (set_stmt_vec_info_vec): New function.
2977 (free_stmt_vec_infos): Likewise.
2978 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2979 the global stmt_vec_info_vec.
2980 * tree-parloops.c (gather_scalar_reductions): Use
2981 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2984 2018-05-29 Richard Biener <rguenther@suse.de>
2986 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2988 2018-05-29 Martin Liska <mliska@suse.cz>
2989 David Malcolm <dmalcolm@redhat.com>
2991 * vec.c (test_reverse): New.
2992 (vec_c_tests): Add new test.
2993 * vec.h (vl_ptr>::reverse): New function.
2995 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2997 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2999 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
3002 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
3004 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
3006 2018-05-28 Richard Biener <rguenther@suse.de>
3008 PR tree-optimization/85933
3009 * tree-vect-data-refs.c (vect_record_base_alignments): Only
3010 look at stmts marked as vectorizable.
3012 2018-05-28 Richard Biener <rguenther@suse.de>
3014 PR tree-optimization/85934
3015 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3016 vector boolean check before scalar optimization.
3018 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3020 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3023 2018-05-28 Mark Wielaard <mark@klomp.org>
3025 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3026 if it is an expression containing a minus sign.
3028 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3030 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3032 2018-05-27 Paul Koning <ni1d@arrl.net>
3034 * config/pdp11/pdp11.md (truncsihi2): Remove.
3036 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3037 Chung-Ju Wu <jasonwucj@gmail.com>
3039 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3041 (unaligned_store_dw): Ditto.
3042 * config/nds32/nds32-memory-manipulation.c
3043 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3044 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3045 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3046 (emit_setmem_word_loop): Rename to ...
3047 (emit_setmem_doubleword_loop): ... this.
3048 (nds32_gen_dup_4_byte_to_word_value): New function.
3049 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3050 (nds32_expand_setmem_loop): Refine implementation.
3051 (nds32_expand_setmem_loop_v3m): Ditto.
3052 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3055 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3057 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3059 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3061 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3062 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3063 machine->attr_no_prologue_p.
3064 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3065 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3066 (nds32_expand_epilogue): Consider attr_naked_p.
3067 (nds32_expand_epilogue_v3pop): Likewise.
3068 (nds32_can_use_return_insn): Likewise.
3069 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3070 attr_no_prologue_p fields.
3071 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3073 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3076 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3078 * config/i386/sse.md
3079 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3081 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3083 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3085 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3087 (*<floatsuffix>floatv2div2sf2): Rename to ...
3088 (*float<floatunssuffix>v2div2sf2): ... this.
3089 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3090 (float<floatunssuffix>v2div2sf2_mask): ... this.
3091 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3092 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3093 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3095 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3097 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3099 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3101 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3103 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3105 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3106 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3107 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3108 gen_ufix_truncv8dfv8si2.
3109 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3110 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3111 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3112 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3113 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3114 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3115 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3116 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3118 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3122 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3124 2018-05-25 Jim Wilson <jimw@sifive.com>
3126 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3127 * config/riscv/riscv.c (struct machine_function): Add
3128 interrupt_handler_p and attribute_checked_p fields.
3129 (riscv_attribute_table): Add interrupt.
3130 (riscv_interrupt_type_p): New.
3131 (riscv_save_reg_p): Save extra regs for interrupt handler.
3132 (riscv_use_save_libcall): Return false for interrupt handler.
3133 (riscv_first_stack_step): Add forward declaration.
3134 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3135 for interrupt handler with large frame. Use it for saved reg list.
3136 (riscv_expand_prologue): Move flag_stack_usage_info support to
3137 eliminate duplication.
3138 (riscv_expand_epilogue): Generate mret for interrupt handler.
3139 (riscv_epilogue_uses): New.
3140 (riscv_can_use_return_insn): Return false for interrupt handler.
3141 (riscv_function_ok_for_sibcall): Likewise.
3142 (riscv_set_current_function): Add interrupt handler support.
3143 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3144 * config/riscv/riscv.md (UNSPECV_MRET): New.
3146 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3148 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3150 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3152 PR tree-optimization/85712
3153 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3154 this candidate has already been replaced in-situ by a copy.
3156 2018-05-25 Jason Merrill <jason@redhat.com>
3158 PR c++/80485 - inline function non-zero address.
3159 * symtab.c (nonzero_address): Check DECL_COMDAT.
3161 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3164 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3165 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3166 extension of SImode operation. Use const123_operand predicate.
3167 (*saddsi_1): Remove.
3168 (*saddl_se_1): Ditto.
3170 (*ssubl_se_1): Ditto.
3171 * config/alpha/predicates.md (const123_operand): New predicate.
3172 * config/alpha/constraints.md (P): Use IN_RANGE.
3174 2018-05-25 Richard Biener <rguenther@suse.de>
3176 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3178 (ref_maybe_used_by_stmt_p): Likewise.
3179 (stmt_may_clobber_ref_p): Likewise.
3180 (stmt_may_clobber_ref_p_1): Likewise.
3181 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3183 (ref_maybe_used_by_stmt_p): Likewise.
3184 (stmt_may_clobber_ref_p): Likewise.
3185 (stmt_may_clobber_ref_p_1): Likewise.
3186 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3187 the alias oracle to disambiguate DRs with stmts DR analysis
3189 (vect_analyze_data_refs): Do not give up on not analyzable
3190 DRs for BB vectorization. Remove code truncating the dataref
3193 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3196 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3197 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3198 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3200 2018-05-25 Richard Biener <rguenther@suse.de>
3202 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3203 function, combining stmt data ref gathering and fatal analysis
3205 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3206 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3207 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3208 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3209 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3211 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3213 PR tree-optimization/85720
3214 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3215 SCC if all partitions are builtins.
3216 (version_loop_by_alias_check): New parameter. Generate cancelable
3217 runtime alias check if all partitions are builtins.
3218 (distribute_loop): Update call to above function.
3220 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3222 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3223 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3224 (parm_default_def_partition_arg): Ditto.
3225 (set_parm_default_def_partition): Ditto.
3226 (get_parm_default_def_partitions): Ditto and make it static.
3227 (get_undefined_value_partitions): Ditto and make it static.
3228 (remove_ssa_form): Refactor call to init_var_map here.
3229 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3230 computation for loop region.
3231 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3232 (register_default_def): Delete.
3233 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3234 (parm_default_def_partition_arg): Ditto.
3235 (set_parm_default_def_partition): Ditto.
3236 (get_parm_default_def_partitions): Ditto and make it static.
3237 (get_undefined_value_partitions): Ditto and make it static.
3238 (coalesce_with_default, coalesce_with_default): Update comment.
3239 (create_coalesce_list_for_region): New func factored out from
3240 create_outofssa_var_map.
3241 (populate_coalesce_list_for_outofssa): New func factored out from
3242 create_outofssa_var_map and coalesce_ssa_name.
3243 (create_outofssa_var_map): Delete.
3244 (coalesce_ssa_name): Refactor to support live range computation.
3245 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3246 (get_parm_default_def_partitions): Delete.
3247 (get_undefined_value_partitions): Ditto.
3248 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3249 computation for loop region.
3250 (new_tree_live_info, loe_visit_block): Ditto.
3251 (live_worklist, set_var_live_on_entry): Ditto.
3252 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3253 * tree-ssa-live.h (struct _var_map): New fields.
3254 (init_var_map): Change decl.
3255 (region_contains_p): New.
3257 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3259 * tree-ssa-live.h (live_merge_and_clear): Delete.
3261 2018-05-25 Richard Biener <rguenther@suse.de>
3264 * tree-dump.c (dequeue_and_dump): Remove access to removed
3265 operand 2 of a SWITCH_EXPR.
3267 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3269 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3270 multiplication and division.
3271 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3272 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3273 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3274 (cond_udiv_optab, cond_umod_optab): New optabs.
3275 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3276 (IFN_COND_RDIV): New internal functions.
3277 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3278 TRUNC_MOD_EXPR and RDIV_EXPR.
3279 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3280 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3282 (SVE_INT_BINARY): Include mult.
3283 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3284 (optab, sve_int_op): Handle mult.
3285 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3287 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3288 for SVE_INT_BINARY_SD.
3290 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3292 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3293 (optab, sve_int_op): Handle div and udiv.
3294 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3295 for SVE_INT_BINARY_SD.
3296 (*<optab><mode>3): New insn for the same.
3298 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3300 * tree-vect-patterns.c: Include predict.h.
3301 (vect_recog_divmod_pattern): Restrict check for division support
3302 to when optimizing for size.
3304 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3306 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3307 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3308 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3309 (gimple_match_op::set_op): Likewise.
3310 (gimple_resimplify4): Declare.
3311 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3312 (expr::gen_transform): Likewise.
3313 (decision_tree::gen): Generate a simplification routine for 4 operands.
3314 * gimple-match-head.c (gimple_simplify): Add an overload for
3315 4 operands. In the top-level function, handle up to 4 call
3316 arguments and call gimple_resimplify4.
3317 (gimple_resimplify4): New function.
3318 (build_call_internal): Pass a fourth operand.
3319 (maybe_push_to_seq): Likewise.
3320 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3321 Fold VEC_COND_EXPRs of an operation and a default value into
3322 an IFN_COND_* function if possible.
3323 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3325 (SVE_COND_FP_BINARY): Include them.
3326 (optab, sve_fp_op): Handle them.
3327 (SVE_INT_BINARY_REV): New code iterator.
3328 (SVE_COND_FP_BINARY_REV): New int iterator.
3329 (commutative): New int attribute.
3330 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3332 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3334 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3335 (*cond_<optab><mode>): New patterns for reversed operands.
3337 2018-05-25 Richard Biener <rguenther@suse.de>
3339 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3340 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3341 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3342 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3343 * tree-vect-data-refs.c (everywhere): Adjust users.
3344 * tree-vect-loop.c (everywhere): Likewise.
3345 * tree-vect-slp.c (everywhere): Likewise.
3346 * tree-vect-stmts.c (everywhere): Likewise.
3347 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3349 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3351 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3353 (gcc_cv_as_section_exclude): ... this.
3354 Try Solaris as #exclude syntax.
3355 * configure: Regenerate.
3356 * config.in: Regenerate.
3357 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3359 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3360 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3362 * varasm.c (default_elf_asm_named_section): Don't check if
3363 HAVE_GAS_SECTION_EXCLUDE is defined.
3365 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3367 * doc/md.texi: Update the documentation of the cond_* optabs
3368 to mention the new final operand. Fix GET_MODE_NUNITS call.
3369 Describe the scalar case too.
3370 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3371 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3373 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3374 (get_conditional_internal_fn): Update comment.
3375 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3376 accumulator value as a final argument to conditional functions.
3377 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3378 a define_expand and add an "else" operand. Assert for now that
3379 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3380 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3381 (*cond_<optab><mode>): New patterns.
3382 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3383 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3384 (UNSPEC_COND_EOR): Delete.
3385 (optab): Remove associated mappings.
3386 (SVE_INT_BINARY): New code iterator.
3387 (sve_int_op): Remove int attribute and add "minus" to the code
3389 (SVE_COND_INT_OP): Delete.
3390 (SVE_COND_FP_OP): Rename to...
3391 (SVE_COND_FP_BINARY): ...this.
3393 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3395 * optabs.c (can_reuse_operands_p): New function.
3396 (maybe_legitimize_operands): Try to reuse the results for
3399 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3401 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3402 Add {q} suffix to insn mnemonic.
3404 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3406 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3407 (msp430_warn_func_return): New.
3409 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3411 * fold-const.c (tree_nonzero_bits): New function.
3412 * fold-const.h (tree_nonzero_bits): Likewise.
3413 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3414 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3416 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3420 * varasm.c (assemble_alias): Check ifunc_resolver only on
3423 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3426 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3427 when memory input operand is handled.
3429 2018-05-24 Luis Machado <luis.machado@linaro.org>
3431 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3433 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3435 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3437 * match.pd: Delay FMA folds until after vectorization.
3439 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3442 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3443 address check not strict.
3445 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3447 * gimple-match.h (gimple_match_op): New class.
3448 (mprts_hook): Replace parameters with a gimple_match_op *.
3449 (maybe_build_generic_op): Likewise.
3450 (gimple_simplified_result_is_gimple_val): Replace parameters with
3451 a const gimple_match_op *.
3452 (gimple_simplify): Replace code_helper * and tree * parameters with
3453 a gimple_match_op * parameter.
3454 (gimple_resimplify1): Replace code_helper *, tree and tree *
3455 parameters with a gimple_match_op * parameter.
3456 (gimple_resimplify2): Likewise.
3457 (gimple_resimplify3): Likewise.
3458 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3459 parameters with a gimple_match_op * parameter.
3460 * gimple-match-head.c (gimple_simplify): Change prototypes of
3461 auto-generated functions to take a gimple_match_op * instead of
3462 separate code_helper * and tree * parameters. Make the same
3463 change in the top-level overload and update calls to the
3464 gimple_resimplify routines. Update calls to the auto-generated
3465 functions and to maybe_push_res_to_seq in the publicly-facing
3466 operation-specific gimple_simplify overloads.
3467 (gimple_match_op::MAX_NUM_OPS): Define.
3468 (gimple_resimplify1): Replace rcode and ops with a single res_op
3469 parameter. Update call to gimple_simplify.
3470 (gimple_resimplify2): Likewise.
3471 (gimple_resimplify3): Likewise.
3472 (mprts_hook): Replace parameters with a gimple_match_op *.
3473 (maybe_build_generic_op): Likewise.
3474 (build_call_internal): Replace type, nargs and ops with
3475 a gimple_match_op *.
3476 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3477 with a single gimple_match_op *. Update calls to mprts_hook,
3478 build_call_internal and gimple_simplified_result_is_gimple_val.
3479 Factor out code that is common to the tree_code and combined_fn cases.
3480 * genmatch.c (expr::gen_transform): Replace tem_code and
3481 tem_ops with a gimple_match_op called tem_op. Update calls
3482 to the gimple_resimplify functions and maybe_push_res_to_seq.
3483 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3484 res_ops. Update call to the gimple_resimplify functions.
3485 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3486 (decision_tree::gen): Make the functions take a gimple_match_op *
3487 called res_op instead of separate res_code and res_ops parameters.
3488 Update call accordingly.
3489 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3490 and ops with a single res_op parameter. Update calls to
3491 maybe_build_generic_op and maybe_push_res_to_seq.
3492 (fold_stmt_1): Update calls to gimple_simplify and
3493 replace_stmt_with_simplification.
3494 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3495 and gimple_simplified_result_is_gimple_val.
3496 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3498 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3499 with a gimple_match_op *.
3500 (vn_nary_build_or_lookup): Likewise. Update call to
3501 vn_nary_build_or_lookup_1.
3502 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3503 gimple_match_op *. Update calls to the gimple_resimplify routines
3504 and to gimple_simplified_result_is_gimple_val.
3505 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3506 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3507 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3508 (visit_nary_op): Likewise.
3509 (visit_reference_op_load): Likewise.
3511 2018-05-23 Luis Machado <luis.machado@linaro.org>
3513 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3514 modifier for printing the step amount.
3516 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3519 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3522 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3524 * doc/sourcebuild.texi (Endianness): New subsubsection.
3526 2018-05-23 Luis Machado <luis.machado@linaro.org>
3528 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3529 <prefetch_dynamic_strides>: New const bool field.
3530 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3531 prefetch_dynamic_strides.
3532 (exynosm1_prefetch_tune): Likewise.
3533 (thunderxt88_prefetch_tune): Likewise.
3534 (thunderx_prefetch_tune): Likewise.
3535 (thunderx2t99_prefetch_tune): Likewise.
3536 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3538 (aarch64_override_options_internal): Update to set
3539 PARAM_PREFETCH_DYNAMIC_STRIDES.
3540 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3541 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3542 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3543 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3544 prefetch-dynamic-strides setting.
3546 2018-05-23 Luis Machado <luis.machado@linaro.org>
3548 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3549 <minimum_stride>: New const int field.
3550 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3551 minimum_stride field defaulting to -1.
3552 (exynosm1_prefetch_tune): Likewise.
3553 (thunderxt88_prefetch_tune): Likewise.
3554 (thunderx_prefetch_tune): Likewise.
3555 (thunderx2t99_prefetch_tune): Likewise.
3556 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3557 <default_opt_level>: Set to 3.
3558 (aarch64_override_options_internal): Update to set
3559 PARAM_PREFETCH_MINIMUM_STRIDE.
3560 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3561 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3562 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3563 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3564 stride is constant and is below the minimum stride threshold.
3566 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3568 * config/arm/arm-cpus.in (mode26): Delete.
3569 (armv4): Delete mode26 reference.
3570 * config/arm/arm.c (arm_configure_build_target): Delete use of
3573 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3575 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3577 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3578 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3580 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3581 (floatunsdidf2): Ditto.
3583 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3585 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3586 (fixuns_trunc<mode>si2_avx512f): Ditto.
3587 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3588 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3589 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3591 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3593 PR rtl-optimization/79985
3594 * df-scan.c (df_insn_refs_collect): Remove special case for
3595 global registers and asm statements.
3597 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3599 * extend.texi (Global Register Variables): Rewrite the bullet list.
3600 Note that the register is available for allocation. Note that access
3601 via inline asm must use constraints. Add note about async-signal
3602 handlers. Remove paragraph about automagic register selection.
3604 2018-05-23 Richard Biener <rguenther@suse.de>
3606 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3607 of fixed offset from memset VN.
3609 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3611 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3613 (alloc_cand_and_find_basis): Initialize first_interp field.
3614 (slsr_process_mul): Modify first_interp field.
3615 (slsr_process_add): Likewise.
3616 (slsr_process_cast): Modify first_interp field for each new
3618 (slsr_process_copy): Likewise.
3619 (dump_candidate): Dump first_interp field.
3620 (replace_mult_candidate): Process all interpretations, not just
3622 (replace_rhs_if_not_dup): Likewise.
3623 (replace_one_candidate): Likewise.
3625 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3627 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3629 (aarch64_needs_frame_chain): New function.
3630 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3632 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3635 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3636 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3637 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3638 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3639 as true for strict-align.
3640 (aarch64_can_inline_p): Perform checks even when callee has no
3641 attributes to check for strict alignment.
3642 * doc/extend.texi (AArch64 Function Attributes): Document
3644 * doc/invoke.texi: (AArch64 Options): Likewise.
3646 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3648 PR tree-optimization/85853
3649 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3650 the handling of the root of the node to...
3651 (vect_slp_analyze_node_operations_1): ...this new function,
3652 and run the whole thing with the child nodes' def types
3653 set according to their SLP node's def type.
3655 2018-05-23 Richard Biener <rguenther@suse.de>
3658 * tree-data-ref.c (create_runtime_alias_checks): Defer
3659 and ignore overflow warnings.
3661 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3663 PR tree-optimization/85822
3664 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3667 2018-05-23 Richard Biener <rguenther@suse.de>
3669 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3670 memset constants via native_interpret_expr.
3672 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3675 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3677 (cgraph_node::create_alias): Likewise.
3678 (cgraph_node::get_availability): Check ifunc_resolver instead
3679 of looking up ifunc attribute.
3680 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3681 * varasm.c (do_assemble_alias): Likewise.
3682 (assemble_alias): Likewise.
3683 (default_binds_local_p_3): Likewise.
3684 * cgraph.h (cgraph_node): Add ifunc_resolver.
3685 (cgraph_node::only_called_directly_or_aliased_p): Return false
3687 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3689 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3690 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3691 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3692 instead of looking up ifunc attribute.
3694 2018-05-22 Luis Machado <luis.machado@linaro.org>
3696 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3698 2018-05-22 Martin Sebor <msebor@redhat.com>
3701 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3702 only when expasion succeeds.
3703 (expand_builtin_strcmp): Same.
3704 (expand_builtin_strncmp): Same.
3706 2018-05-22 Martin Sebor <msebor@redhat.com>
3708 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3710 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3711 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3713 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3714 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3716 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3717 Add check for legitimate_address.
3718 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3719 (aarch64_swap_ldrstr_operands): New.
3720 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3723 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3724 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3726 * config/aarch64/aarch64.md: New patterns to generate stp
3728 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3729 single words and double words.
3730 (load_pair_sw, load_pair_dw): Likewise.
3731 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3733 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3735 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3736 for different mode ldpstp and add peephole for merged zero stores.
3738 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3740 (aarch64_gen_store_pair): Rename calls to match new patterns.
3741 (aarch64_gen_load_pair): Rename calls to match new patterns.
3742 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3743 (load_pair<DREG:mode><DREG2:mode>): ... This.
3744 (store_pair<mode>): Rename to...
3745 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3746 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3748 (V_INT_EQUIV): Handle SImode.
3749 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3752 2018-05-22 Martin Sebor <msebor@redhat.com>
3755 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3756 or ajust the presumed bound on an operation to avoid unnecessary
3759 2018-05-22 Martin Sebor <msebor@redhat.com>
3761 PR tree-optimization/85826
3762 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3763 assuming that a DECL necesarily has a constant size.
3765 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3768 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3770 2018-05-22 Richard Biener <rguenther@suse.de>
3772 PR tree-optimization/85834
3773 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3774 non-constant and non-zero memset arguments.
3776 2018-05-22 Martin Liska <mliska@suse.cz>
3779 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3781 2018-05-22 Richard Biener <rguenther@suse.de>
3783 PR tree-optimization/85863
3784 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3785 comparisons when vectype is specified.
3786 (vectorizable_condition): Do not specify vectype for
3787 vect_is_simple_cond when SLP vectorizing.
3789 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3792 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3793 define __ibm128 as long double.
3794 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3795 as a distinct type when IEEE 128-bit support is enabled.
3796 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3797 128-bit types to use the correct functions.
3798 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3799 convert between 128-bit floating point types that have different
3800 modes but the same representation, instead of using gen_lowpart to
3802 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3804 (IFKF_reg): New attributes to give the register constraints for
3806 (extend<mode>tf2_internal): New insns to mark an explicit
3807 conversion between 128-bit floating point types that have a
3808 different mode but share the same representation.
3810 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3812 PR tree-optimization/85814
3813 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3814 a null return from get_strinfo when unsharing the next
3815 strinfo in the chain.
3817 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3820 * varasm.c (weak_finish): Clean up weak_decls.
3822 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3824 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3825 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3826 UNSPEC_UADALP values.
3827 * config/aarch64/iterators.md (ABAL): New int iterator.
3830 (sur): Add mappings for the above.
3831 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3833 (aarch64_<sur>abal<mode>_4): Likewise.
3834 (aarch64_<sur>adalp<mode>_3): Likewise.
3835 (<sur>sadv16qi): New define_expand.
3837 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3839 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3840 (*movdf_internal): Ditto.
3841 (*rcpsf2_sse): Ditto.
3842 (*rsqrtsf2_sse): Ditto.
3843 (*sqrt<mode>2_sse): Ditto.
3845 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3847 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3849 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3850 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3851 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3852 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3854 * config/aarch64/arm_neon.h: Likewise.
3855 * config/aarch64/iterators.md (VQ_I): New.
3857 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3859 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3861 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3864 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3866 * config/nds32/constraints.md (S): New constraint.
3867 * config/nds32/nds32.md (call_internal): Use constraint S.
3868 (call_value_internal): Likewise.
3869 (sibcall_internal): Likewise.
3870 (sibcall_value_internal): Likewise.
3872 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3873 Chung-Ju Wu <jasonwucj@gmail.com>
3875 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3878 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3879 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3882 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3883 (nds32_rtx_costs_impl): Simplify.
3884 (nds32_address_cost_impl): Simplify.
3885 (nds32_init_rtx_costs): New function.
3886 (nds32_rtx_costs_speed_prefer): Likewise.
3887 (nds32_rtx_costs_size_prefer): Likewise.
3888 (nds32_address_cost_speed_prefer): Likewise.
3889 (nds32_address_cost_speed_fwprop): Likewise.
3890 (nds32_address_cost_size_prefer): Likewise.
3891 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3892 * config/nds32/nds32.c (nds32_option_override): Use
3893 nds32_init_rtx_costs function.
3895 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3897 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3898 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3899 (TARGET_PIPELINE_N8): Likewise.
3900 (TARGET_PIPELINE_N10): Likewise.
3901 (TARGET_PIPELINE_N13): Likewise.
3902 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3904 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3906 * config/nds32/nds32-fpu.md: Update copyright year.
3908 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3912 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3914 * config/nds32/nds32.c
3915 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3916 * config/nds32/nds32.opt (minline-asm-r15): New option.
3918 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3920 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3922 * config/nds32/nds32.md (abssi2): New pattern.
3924 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3926 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3927 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3928 Merge insn pattern from sse_cvtsi2ss<round_name> and
3929 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3930 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3931 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3932 using SWI48 mode iterator.
3933 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3934 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3935 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3936 pattern from sse_cvttss2si<round_saeonly_name>
3937 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3938 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3939 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3940 using SWI48 mode iterator.
3941 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3942 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3943 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3944 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3945 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3946 using SWI48 mode iterator.
3947 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3948 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3949 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3950 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3951 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3952 SWI48 mode iterator.
3953 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3954 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3955 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3956 pattern from sse_cvttsd2si<round_saeonly_name>
3957 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3959 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3961 * config/nds32/nds32-md-auxiliary.c
3962 (nds32_valid_smw_lwm_base_p): Refine.
3963 (nds32_output_smw_single_word): Refine.
3964 (nds32_output_smw_double_word): New.
3965 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3967 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3969 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3970 (nds32_output_stack_pop): Refine.
3971 (nds32_expand_unaligned_load): Refine.
3972 (nds32_expand_unaligned_store): Refine.
3974 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3975 Chung-Ju Wu <jasonwucj@gmail.com>
3977 * config/nds32/constants.md: Add TP_REGNUM constant.
3978 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3979 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3981 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3982 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3983 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3984 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3985 and PIC code generation.
3986 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3988 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3990 * config/nds32/nds32.md: Support TLS and PIC.
3991 * config/nds32/nds32.c: Support TLS and PIC.
3992 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3993 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3996 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3998 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3999 mode with E_ prefix.
4001 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4002 Chung-Ju Wu <jasonwucj@gmail.com>
4004 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
4005 * config/nds32/nds32-md-auxiliary.c
4006 (symbolic_reference_mentioned_p): New.
4007 (nds32_legitimize_ict_address): New.
4008 (nds32_expand_ict_move): New.
4009 (nds32_indirect_call_referenced_p): New.
4010 (nds32_symbol_binds_local_p): Delete.
4011 (nds32_long_call_p): Modify.
4012 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
4013 * config/nds32/nds32-protos.h
4014 (symbolic_reference_mentioned_p): Declare.
4015 (nds32_legitimize_ict_address): Declare.
4016 (nds32_expand_ict_move): Declare.
4017 (nds32_indirect_call_referenced_p): Declare.
4018 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4019 (nds32_relax_group): Use nds32_ict_const_p as condition.
4020 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4021 (nds32_asm_file_start): Output ict_model directive in asm code.
4022 (nds32_legitimate_address_p): Consider indirect call.
4023 (nds32_print_operand): Consider indirect call.
4024 (nds32_print_operand_address): Consider indirect call.
4025 (nds32_insert_attributes): Handle "indirect_call" attribute.
4026 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4027 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4028 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4029 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4030 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4031 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4032 (TARGET_ICT_MODEL_SMALL): Define.
4033 (TARGET_ICT_MODEL_LARGE): Define.
4034 * config/nds32/nds32.md (movsi): Consider ict model.
4035 (call, call_value): Consider ict model.
4036 (sibcall, sibcall_value): Consider ict model.
4037 * config/nds32/nds32.opt (mict-model): New option.
4038 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4041 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4042 Monk Chiang <sh.chiang04@gmail.com>
4043 Jim Wilson <jimw@sifive.com>
4045 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4046 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4047 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4048 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4049 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4050 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4051 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4052 compute save_libcall_adjustment properly.
4053 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4054 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4055 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4056 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4057 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4058 (ABI_SPEC): Handle mabi=ilp32e.
4059 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4060 (RVE): Add RVE mask.
4061 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4062 <-march>: Add rv32e as an example.
4064 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4067 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4068 (intra_create_variable_infos): Handle C++ constructors.
4070 2018-05-18 Martin Liska <mliska@suse.cz>
4072 * passes.def: Remove a redundant pass.
4074 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4077 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4079 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4081 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4083 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4085 (armv2, armv2a, armv3, armv3m): Delete architectures.
4086 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4087 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4088 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4090 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4091 (*mulsidi3adddi): Likewise.
4092 (mulsidi3): Likewise.
4093 (*mulsidi3_nov6): Likewise.
4094 (umulsidi3): Likewise.
4095 (umulsidi3_nov6): Likewise.
4096 (umaddsidi4): Likewise.
4097 (*umulsidi3adddi): Likewise.
4098 (smulsi3_highpart): Likewise.
4099 (*smulsi3_highpart_nov6): Likewise.
4100 (umulsi3_highpart): Likewise.
4101 (*umulsi3_highpart_nov6): Likewise.
4102 * config/arm/arm.h (arm_arch3m): Delete.
4103 * config/arm/arm.c (arm_arch3m): Delete.
4104 (arm_option_override_internal): Update armv3-related comment.
4105 (arm_configure_build_target): Delete use of isa_bit_mode32.
4106 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4107 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4108 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4110 (mulusa3): Likewise.
4111 * config/arm/arm-protos.h (arm_arch3m): Delete.
4112 * config/arm/arm-tables.opt: Regenerate.
4113 * config/arm/arm-tune.md: Likewise.
4114 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4115 deleted architectures.
4117 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4119 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4120 (armv5t, armv5te): New features.
4121 (ARMv5, ARMv5e): Delete fgroups.
4122 (ARMv5t, ARMv5te): Adjust for above changes.
4124 (armv5, armv5e): Delete arches.
4125 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4127 (*call_reg_arm): Likewise.
4128 (*call_value_reg_armv5): Likewise.
4129 (*call_value_reg_arm): Likewise.
4130 (*call_symbol): Likewise.
4131 (*call_value_symbol): Likewise.
4132 (*sibcall_insn): Likewise.
4133 (*sibcall_value_insn): Likewise.
4135 (prefetch): Likewise.
4136 (define_split and define_peephole2 dependent on arm_arch5):
4138 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4140 (TARGET_ARM_QBIT): Likewise.
4141 (TARGET_DSP_MULTIPLY): Likewise.
4142 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4143 (arm_arch5, arm_arch5e): Delete.
4144 (arm_arch5t, arm_arch5te): Declare.
4145 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4146 (arm_arch5t): Declare.
4147 (arm_option_reconfigure_globals): Update for the above.
4148 (arm_options_perform_arch_sanity_checks): Update comment, replace
4149 use of arm_arch5 with arm_arch5t.
4150 (use_return_insn): Likewise.
4151 (arm_emit_call_insn): Likewise.
4152 (output_return_instruction): Likewise.
4153 (arm_final_prescan_insn): Likewise.
4154 (arm_coproc_builtin_available): Likewise.
4155 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4156 arm_arch5e with arm_arch5t and arm_arch5te.
4157 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4158 (arm_arch5t, arm_arch5te): Declare.
4159 * config/arm/arm-tables.opt: Regenerate.
4160 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4161 * config/arm/t-multilib: Likewise.
4162 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4163 instead of arm_arch5.
4164 (*call_reg_thumb1): Likewise.
4165 (*call_value_reg_thumb1_v5): Likewise.
4166 (*call_value_reg_thumb1): Likewise.
4167 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4169 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4171 2018-05-18 Martin Liska <mliska@suse.cz>
4173 PR gcov-profile/84846
4174 * doc/gcov.texi: Document -t option of gcov tool.
4176 2018-05-18 Martin Liska <mliska@suse.cz>
4178 PR gcov-profile/84846
4179 * gcov.c (print_usage): Add new -t option.
4180 (process_args): Handle the option.
4181 (generate_results): Use stdout as output when requested by
4184 2018-05-18 Martin Liska <mliska@suse.cz>
4186 PR gcov-profile/84846
4187 * coverage.c (coverage_init): Write PWD to .gcno file.
4188 * doc/gcov.texi: Document how working directory is printed.
4189 * gcov-dump.c (dump_gcov_file): Print PWD.
4190 * gcov.c (output_intermediate_file): Likewise.
4191 (read_graph_file): Read PWD string.
4192 (output_lines): Print PWD.
4194 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4197 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4198 for retval and return false if all args to phi are zero.
4200 2018-05-18 Richard Biener <rguenther@suse.de>
4202 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4204 (evrp_dom_walker::before_dom_children): Call it.
4206 2018-05-18 Richard Biener <rguenther@suse.de>
4208 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4209 results when processing array refs with variable index.
4211 2018-05-18 Toon Moene <toon@moene.org>
4213 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4214 directly after that of -floop-interchange. Indicate that both
4215 options are enabled by default when specifying -O3.
4217 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4219 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4220 iterator. Delete separate integer-mode vec_set<mode> expander.
4221 (aarch64_simd_vec_setv2di): Delete.
4222 (vec_setv2di): Delete.
4223 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4224 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4225 the "w, r" alternative.
4227 2018-05-18 Martin Liska <mliska@suse.cz>
4229 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4230 * tree-pass.h (make_pass_lower_switch_O0): New function.
4231 * tree-switch-conversion.c (node_has_low_bound): Remove.
4232 (node_has_high_bound): Likewise.
4233 (node_is_bounded): Likewise.
4234 (class pass_lower_switch): Make it a template type and create
4236 (pass_lower_switch::execute): Add template argument.
4237 (make_pass_lower_switch): New function.
4238 (make_pass_lower_switch_O0): New function.
4239 (do_jump_if_equal): Remove.
4240 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4241 all the hard work to tree optimization passes.
4243 2018-05-18 Martin Liska <mliska@suse.cz>
4245 * dbgcnt.c (limit_low): Renamed from limit.
4246 (limit_high): New variable.
4247 (dbg_cnt_is_enabled): Check for upper limit.
4248 (dbg_cnt): Adjust dumping.
4249 (dbg_cnt_set_limit_by_index): Add new argument for high
4251 (dbg_cnt_set_limit_by_name): Likewise.
4252 (dbg_cnt_process_single_pair): Parse new format.
4253 (dbg_cnt_process_opt): Use strtok.
4254 (dbg_cnt_list_all_counters): Remove 'value' and add
4256 * doc/invoke.texi: Document changes.
4258 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4260 * doc/sourcebuild.texi (scalar_all_fma): Document.
4261 * tree.def (FMA_EXPR): Delete.
4262 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4263 * internal-fn.c (ternary_direct): New macro.
4264 (expand_ternary_optab_fn): Likewise.
4265 (direct_ternary_optab_supported_p): Likewise.
4266 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4267 * builtins.c (fold_builtin_fma): Delete.
4268 (fold_builtin_3): Don't call it.
4269 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4270 * expr.c (expand_expr_real_2): Likewise.
4271 * fold-const.c (operand_equal_p): Likewise.
4272 (fold_ternary_loc): Likewise.
4273 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4274 * gimple.c (DEFTREECODE): Likewise.
4275 * gimplify.c (gimplify_expr): Likewise.
4276 * optabs-tree.c (optab_for_tree_code): Likewise.
4277 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4278 * tree-eh.c (operation_could_trap_p): Likewise.
4279 (stmt_could_throw_1_p): Likewise.
4280 * tree-inline.c (estimate_operator_cost): Likewise.
4281 * tree-pretty-print.c (dump_generic_node): Likewise.
4282 (op_code_prio): Likewise.
4283 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4284 * tree-ssa-operands.c (get_expr_operands): Likewise.
4285 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4286 * fold-const-call.h (fold_fma): Delete.
4287 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4288 CFN_FNMA and CFN_FNMS.
4290 * genmatch.c (combined_fn): New enum.
4291 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4292 (commutative_op): New function.
4293 (commutate): Use it. Handle more than 2 operands.
4294 (dt_operand::gen_gimple_expr): Use commutative_op.
4295 (parser::parse_expr): Allow :c to be used with non-binary
4296 operators if the commutative operand is known.
4297 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4298 CFN_FMS, CFN_FNMA and CFN_FNMS.
4299 (backprop::process_assign_use): Remove FMA_EXPR handling.
4300 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4301 (gen_hsa_fma): New function.
4302 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4303 IFN_FNMA and IFN_FNMS.
4304 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4305 * gimple-fold.h (follow_all_ssa_edges): Declare.
4306 * gimple-fold.c (follow_all_ssa_edges): New function.
4307 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4308 gimple_build interface and use follow_all_ssa_edges to fold the result.
4309 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4310 instead of checking for optabs directly.
4311 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4312 rather than FMA_EXPRs.
4313 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4314 call to IFN_FMA instead of an FMA_EXPR.
4316 2018-05-17 Jim Wilson <jimw@sifive.com>
4318 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4319 sign extended promoted subreg, and the range does not have the sign bit
4320 set, then do a sign extend.
4322 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4323 test, check for sign extended subreg and/or constant operands, and
4324 do a sign extend in that case.
4326 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4328 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4329 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4331 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4332 Change logics_shift_reg to logics_shift_imm.
4333 (thunderx2t99_fp_loadpair_basic): Delete.
4334 (thunderx2t99_fp_storepair_basic): Delete.
4335 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4336 (thunderx2t99_asimd_polynomial): Delete.
4337 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4338 and neon_fp_mul_d_scalar_q.
4339 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4340 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4341 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4342 (thunderx2t99_asimd_lut): Add missing tbl types.
4343 (thunderx2t99_asimd_ext): Delete.
4344 (thunderx2t99_asimd_load1_1_mult): Delete.
4345 (thunderx2t99_asimd_load1_2_mult): Delete.
4346 (thunderx2t99_asimd_load1_ldp): New.
4347 (thunderx2t99_asimd_load1): New.
4348 (thunderx2t99_asimd_load2): Add missing *load2* types.
4349 (thunderx2t99_asimd_load3): New.
4350 (thunderx2t99_asimd_load4): New.
4351 (thunderx2t99_asimd_store1_1_mult): Delete.
4352 (thunderx2t99_asimd_store1_2_mult): Delete.
4353 (thunderx2t99_asimd_store2_mult): Delete.
4354 (thunderx2t99_asimd_store2_onelane): Delete.
4355 (thunderx2t99_asimd_store_stp): New.
4356 (thunderx2t99_asimd_store1): New.
4357 (thunderx2t99_asimd_store2): New.
4358 (thunderx2t99_asimd_store3): New.
4359 (thunderx2t99_asimd_store4): New.
4361 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4363 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4364 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4366 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4367 Segher Boessenkool <segher@kernel.crashing.org>
4370 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4373 2018-05-17 Richard Biener <rguenther@suse.de>
4375 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4376 for pruning loop and prune defs feeding only already visited PHIs.
4378 2018-05-17 Richard Biener <rguenther@suse.de>
4380 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4382 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4383 Richard Biener <rguenther@suse.de>
4385 PR tree-optimization/85793
4386 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4387 for VMAT_ELEMENTWISE.
4389 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4391 * internal-fn.h (lookup_internal_fn): Declare
4392 * internal-fn.c (lookup_internal_fn): New function.
4393 * gimple.c (gimple_build_call_from_tree): Handle calls to
4395 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4396 internal function names.
4397 * tree-pretty-print.c (dump_generic_node): Likewise.
4398 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4400 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4402 * gimple-fold.h (gimple_build): Make the function forms take
4403 combined_fn rather than built_in_function.
4404 (gimple_simplify): Likewise.
4405 * gimple-match-head.c (gimple_simplify): Likewise.
4406 * gimple-fold.c (gimple_build): Likewise.
4407 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4408 rather than gimple_build_call_internal.
4409 (get_initial_defs_for_reduction): Likewise.
4410 (vect_create_epilog_for_reduction): Likewise.
4411 (vectorizable_live_operation): Likewise.
4413 2018-05-17 Martin Liska <mliska@suse.cz>
4415 * gimple-ssa-sprintf.c (format_directive): Do not use
4416 space in between 'G_' and '('.
4418 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4421 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4422 even if the mask is not all ones.
4425 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4427 (ix86_gimple_fold_builtin): Likewise.
4430 * config/i386/i386.c: Include tree-vector-builder.h.
4431 (ix86_vector_shift_count): New function.
4432 (ix86_fold_builtin): Fold shift builtins by scalar count.
4433 (ix86_gimple_fold_builtin): Likewise.
4435 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4436 _mm512_setzero): New intrinsics.
4438 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4439 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4441 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4442 code generation for cases where splatting a value is not useful.
4443 * simplify-rtx.c (simplify_ternary_operation): Simplify
4444 vec_merge across a vec_duplicate and a paradoxical subreg forming
4445 a vector mode to a vec_concat.
4447 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4449 * config.gcc: Support "goldmont-plus".
4450 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4452 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4453 PROCESSOR_GOLDMONT_PLUS.
4454 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4455 (processor_target_table): Add "goldmont-plus".
4456 (PTA_GOLDMONT_PLUS): Define.
4457 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4458 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4459 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4460 (fold_builtin_cpu): Add "goldmont-plus".
4461 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4462 (ix86_option_override_internal): Add "goldmont-plus".
4463 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4464 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4465 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4466 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4468 2018-05-17 Richard Biener <rguenther@suse.de>
4470 PR tree-optimization/85757
4471 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4472 remove defs that only feed that PHI from further processing.
4474 2018-05-16 Jim Wilson <jimw@sifive.com>
4476 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4478 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4480 2018-05-16 Mark Wielaard <mark@klomp.org>
4482 * dwarf2out.c (count_index_strings): New function.
4483 (output_indirect_strings): Call count_index_strings and generate
4484 header for dwarf_version >= 5.
4486 2018-05-16 Mark Wielaard <mark@klomp.org>
4488 * dwarf2out.c (dwarf_FORM): New function.
4489 (set_indirect_string): Use dwarf_FORM.
4490 (reset_indirect_string): Likewise.
4491 (size_of_die): Likewise.
4492 (value_format): Likewise.
4493 (output_die): Likewise.
4494 (add_skeleton_AT_string): Likewise.
4495 (output_macinfo_op): Likewise.
4496 (index_string): Likewise.
4497 (output_index_string_offset): Likewise.
4498 (output_index_string): Likewise.
4499 (count_index_strings): Likewise.
4501 2018-05-16 Carl Love <cel@us.ibm.com>
4503 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4504 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4506 2018-05-16 Martin Jambor <mjambor@suse.cz>
4508 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4509 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4511 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4513 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4514 (fnma<mode>4): Likewise.
4515 (fms<mode>4): Likewise.
4516 (fnms<mode>4): Likewise.
4517 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4518 (aarch64_fnma<mode>4): Likewise.
4519 (aarch64_fms<mode>4): Likewise.
4520 (aarch64_fnms<mode>4): Likewise.
4521 (aarch64_fnmadd<mode>4): Likewise.
4523 2018-05-16 Jason Merrill <jason@redhat.com>
4525 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4527 2018-05-16 Richard Biener <rguenther@suse.de>
4529 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4530 (dump_stmt_cost): Declare.
4531 (add_stmt_cost): Dump cost we add.
4532 (add_stmt_costs): New function.
4533 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4535 (vect_analyze_stmt): Adjust prototype.
4536 (vectorizable_condition): Likewise.
4537 (vectorizable_live_operation): Likewise.
4538 (vectorizable_reduction): Likewise.
4539 (vectorizable_induction): Likewise.
4540 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4541 cost vector to pass to vectorizable_ and record afterwards.
4542 (vect_model_reduction_cost): Take cost vector argument and adjust.
4543 (vect_model_induction_cost): Likewise.
4544 (vectorizable_reduction): Likewise.
4545 (vectorizable_induction): Likewise.
4546 (vectorizable_live_operation): Likewise.
4547 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4548 SLP_TREE_NUMBER_OF_VEC_STMTS.
4549 (vect_analyze_slp_cost_1): Remove.
4550 (vect_analyze_slp_cost): Likewise.
4551 (vect_slp_analyze_node_operations): Take visited args and
4552 a target cost vector. Avoid processing already visited stmt sets.
4553 (vect_slp_analyze_operations): Use a local cost vector to gather
4554 costs and register those of non-discarded instances.
4555 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4556 (vect_schedule_slp_instance): Remove copying of
4557 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4559 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4560 adding cost. Record cost entry location.
4561 (vect_prologue_cost_for_slp_op): Function to compute cost of
4562 a constant or invariant generated for SLP vect in the prologue,
4563 split out from vect_analyze_slp_cost_1.
4564 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4565 (vect_model_promotion_demotion_cost): Likewise.
4566 (vect_model_store_cost): Likewise, make static.
4567 (vect_model_load_cost): Likewise.
4568 (vectorizable_bswap): Add cost vector arg and adjust.
4569 (vectorizable_call): Likewise.
4570 (vectorizable_simd_clone_call): Likewise.
4571 (vectorizable_conversion): Likewise.
4572 (vectorizable_assignment): Likewise.
4573 (vectorizable_shift): Likewise.
4574 (vectorizable_operation): Likewise.
4575 (vectorizable_store): Likewise.
4576 (vectorizable_load): Likewise.
4577 (vectorizable_condition): Likewise.
4578 (vectorizable_comparison): Likewise.
4579 (can_vectorize_live_stmts): Likewise.
4580 (vect_analyze_stmt): Likewise.
4581 (vect_transform_stmt): Adjust calls to vectorizable_*.
4582 * tree-vectorizer.c: Include gimple-pretty-print.h.
4583 (dump_stmt_cost): New function.
4585 2018-05-16 Richard Biener <rguenther@suse.de>
4587 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4588 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4589 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4590 (check_name): New callback.
4591 (dse_classify_store): Track cycles via a visited bitmap of PHI
4592 defs and simplify handling of in-loop and across loop dead stores
4593 and properly fail for loop-variant refs. Handle byte-tracking with
4594 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4597 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4599 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4600 (vect_get_mask_type_for_stmt): Likewise.
4601 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4603 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4604 to determine the statement's vector type and the vector type that
4605 should be used for calculating nunits. Deal with cases in which
4606 the type has to be deferred.
4607 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4608 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4609 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4610 (vect_determine_vf_for_stmt): New functions, split out from...
4611 (vect_determine_vectorization_factor): ...here.
4612 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4613 (vect_get_mask_type_for_stmt): New functions, split out from
4614 vect_determine_vectorization_factor.
4616 2018-05-16 Richard Biener <rguenther@suse.de>
4618 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4619 verify the [VEC_]COND_EXPR embedded comparison.
4621 2018-05-15 Martin Sebor <msebor@redhat.com>
4623 PR tree-optimization/85753
4624 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4625 RECORD_TYPE in addition to ARRAY_TYPE.
4627 2018-05-15 Martin Sebor <msebor@redhat.com>
4630 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4632 2018-05-15 Richard Biener <rguenther@suse.de>
4634 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4635 add by_clobber_p one. Change algorithm to collect all defs
4636 representing uses we need to walk and try reducing them to
4637 a single one before failing.
4638 (dse_dom_walker::dse_optimize_stmt): Adjust.
4640 2018-05-13 Mark Wielaard <mark@klomp.org>
4642 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4643 (size_of_loc_descr): Likewise.
4644 (output_loc_operands): Likewise.
4645 (output_loc_operands_raw): Likewise.
4646 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4647 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4648 (hash_loc_operands): Likewise.
4649 (compare_loc_operands): Likewise.
4651 2018-05-14 Mark Wielaard <mark@klomp.org>
4653 * dwarf2out.c (count_index_addrs): New function.
4654 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4656 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4658 PR tree-optimization/83648
4659 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4660 return value as malloc candidate.
4662 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4665 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4666 param as true in call to suggest_attribute.
4668 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4670 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4671 -mreadonly-in-sdata.
4673 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4675 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4677 (aarch64_crypto_aesd_fused): Likewise.
4679 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4681 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4682 (movsi_aarch64): Likewise.
4683 (load_pairsi): Likewise.
4684 (load_pairdi): Likewise.
4685 (store_pairsi): Likewise.
4686 (store_pairdi): Likewise.
4687 (load_pairsf): Likewise.
4688 (load_pairdf): Likewise.
4689 (store_pairsf): Likewise.
4690 (store_pairdf): Likewise.
4691 (zero_extend): Likewise.
4692 (trunc): Swap alternatives.
4693 (fcvt_target): Add '?' to prefer w over r.
4695 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4698 * config/i386/i386.md: Disallow non-commutative arithmetics in
4699 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4700 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4701 in the peephole2 before it.
4703 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4705 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4706 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4707 (ix86_handle_option): Handle -mcldemote.
4708 * config.gcc: New header.
4709 * config/i386/cldemoteintrin.h: New file.
4710 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4711 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4713 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4714 OPTION_MASK_ISA_CLDEMOTE.
4715 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4716 (ix86_valid_target_attribute_inner_p): Ditto.
4717 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4718 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4719 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4720 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4721 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4723 * config/i386/i386.opt: Add -mcldemote.
4724 * config/i386/x86intrin.h: New header.
4725 * doc/invoke.texi: Add -mcldemote.
4727 2018-05-14 Richard Biener <rguenther@suse.de>
4729 * doc/match-and-simplify.texi: Adjust :s documentation.
4731 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4733 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4734 intended memcpy size.
4735 (REORDER_45): Likewise.
4737 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4739 * sort.cc: New file.
4740 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4741 * vec.c (qsort_chk): Use gcc_qsort.
4742 * Makefile.in (OBJS-libcommon): Add sort.o.
4743 (build/sort.o): New target. Use it...
4744 (BUILD_RTL): ... here, and...
4745 (build/gencfn-macros): ... here, and...
4746 (build/genmatch): ... here.
4748 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4749 Chung-Ju Wu <jasonwucj@gmail.com>
4751 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4752 * config/nds32/nds32-graywolf.md: New file.
4753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4756 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4757 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4758 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4759 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4760 * config/nds32/pipelines.md: Include n15 settings.
4762 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4763 Chung-Ju Wu <jasonwucj@gmail.com>
4765 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4766 * config/nds32/nds32-n13.md: New file.
4767 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4768 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4770 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4771 * config/nds32/nds32.md (pipeline_model): Add n13.
4772 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4773 * config/nds32/pipelines.md: Include n13 settings.
4775 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4776 Chung-Ju Wu <jasonwucj@gmail.com>
4778 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4779 * config/nds32/nds32-n10.md: New file.
4780 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4781 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4783 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4784 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4785 * config/nds32/nds32.md (pipeline_model): Add n10.
4786 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4787 * config/nds32/pipelines.md: Include n10 settings.
4789 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4790 Kito Cheng <kito.cheng@gmail.com>
4791 Chung-Ju Wu <jasonwucj@gmail.com>
4793 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4794 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4795 Add enum values for DSP extension instructions.
4796 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4798 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4799 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4801 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4802 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4803 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4804 * config/nds32/nds32-intrinsic.md: Likewise.
4805 * config/nds32/nds32_intrinsic.h: Likewise.
4806 * config/nds32/nds32-md-auxiliary.c: Likewise.
4807 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4808 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4809 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4810 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4811 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4812 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4813 TYPE_DMAC in switch statement.
4814 * config/nds32/nds32.c: New checking and implementation for DSP
4815 extension instructions.
4816 * config/nds32/nds32.h: Likewise.
4817 * config/nds32/nds32.md: Likewise.
4818 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4819 * config/nds32/predicates.md: Implement new predicates for DSP
4822 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4824 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4825 Reformat alternatives and attributes so it is easier to identify
4826 which constraints/attributes go with which instruction.
4827 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4828 (mov<mode>_softfloat32, FMOVE64): Likewise.
4829 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4830 (mov<mode>_softfloat64, FMOVE64): Likewise.
4832 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4834 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4836 (Basic PowerPC Built-in Functions): The new name of the
4837 subsection previously known as "PowerPC Built-in Functions".
4838 (Basic PowerPC Built-in Functions Available on all Configurations):
4840 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4841 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4842 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4843 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4845 2018-05-11 Martin Jambor <mjambor@suse.cz>
4848 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4851 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4854 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4856 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4858 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4859 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4860 (ix86_handle_option): Handle -mwaitpkg.
4861 * config.gcc: New header.
4862 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4863 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4864 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4866 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4867 OPTION_MASK_ISA_WAITPKG.
4868 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4869 (ix86_option_override_internal): Add PTA_WAITPKG.
4870 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4871 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4872 IX86_BUILTIN_TPAUSE.
4873 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4874 __builtin_ia32_umwait and __builtin_ia32_tpause.
4875 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4876 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4877 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4878 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4879 UNSPECV_TPAUSE): New.
4880 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4881 * config/i386/i386.opt: Add -mwaitpkg.
4882 * config/i386/waitpkgintrin.h: New file.
4883 * config/i386/x86intrin.h: New header.
4884 * doc/invoke.texi: Add -mwaitpkg.
4886 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4889 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4891 (cortex-m0): Use armv6s-m isa.
4892 (cortex-m0plus): Likewise.
4893 (cortex-m1): Likewise.
4894 (cortex-m0.small-multiply): Likewise.
4895 (cortex-m0plus.small-multiply): Likewise.
4896 (cortex-m1.small-multiply): Likewise.
4898 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4899 Jakub Jelinek <jakub@redhat.com>
4901 PR tree-optimization/85692
4902 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4903 source permute as well.
4905 2018-05-11 Martin Liska <mliska@suse.cz>
4908 * doc/extend.texi: Document LLVM style format for no_sanitize
4911 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4913 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4914 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4915 (mode_supports_vsx_dform_quad): Likewise.
4916 (mode_supports_vmx_dform): Move these functions to be next to the
4917 other mode_supports functions.
4918 (mode_supports_dq_form): Likewise.
4919 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4920 mode_supports_dq_form.
4921 (reg_offset_addressing_ok_p): Likewise.
4922 (offsettable_ok_by_alignment): Likewise.
4923 (rs6000_legitimate_offset_address_p): Likewise.
4924 (legitimate_lo_sum_address_p): Likewise.
4925 (rs6000_legitimize_address): Likewise.
4926 (rs6000_legitimize_reload_address): Likewise.
4927 (rs6000_secondary_reload_inner): Likewise.
4928 (rs6000_preferred_reload_class): Likewise.
4929 (rs6000_output_move_128bit): Likewise.
4931 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4933 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4934 Generate SImode target register for null target.
4935 <case IX86_BUILTIN_XGETBV>: Ditto.
4936 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4937 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4939 2018-05-10 Carl Love <cel@us.ibm.com>
4941 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4942 dcbtt and dcbtstt if operands[2] is 0.
4944 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4947 * config/i386/sse.md (usadv64qi): New expander.
4949 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4951 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4952 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4953 -maltivec=be support.
4954 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4955 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4956 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4957 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4958 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4959 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4960 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4961 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4962 altivec_vsumsws): Adjust.
4963 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4964 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4965 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4967 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4968 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4969 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4970 (altivec_lve<VI_char>x): Delete expand.
4971 (*altivec_lve<VI_char>x_internal): Rename to...
4972 (altivec_lve<VI_char>x): ... this.
4973 (altivec_lvxl_<mode>): Delete expand.
4974 (*altivec_lvxl_<mode>_internal): Rename to ...
4975 (altivec_lvxl_<mode>): ... this.
4976 (altivec_stvxl_<mode>): Delete expand.
4977 (*altivec_stvxl_<mode>_internal): Rename to ...
4978 (altivec_stvxl_<mode>): ... this.
4979 (altivec_stve<VI_char>x): Delete expand.
4980 (*altivec_stve<VI_char>x_internal): Rename to ...
4981 (altivec_stve<VI_char>x): ... this.
4982 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4983 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4984 reduc_plus_scal_<mode>): Adjust.
4985 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4987 (rs6000_cpu_cpp_builtins): Adjust.
4988 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4989 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4990 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4991 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4992 -maltivec=be support.
4993 (rs6000_split_vec_extract_var): Adjust.
4994 (rs6000_split_v4si_init): Adjust.
4995 (swap_selector_for_mode): Delete.
4996 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4997 altivec_expand_stvex_be): Delete.
4998 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4999 -maltivec=be support.
5000 (rs6000_gimple_fold_builtin): Ditto.
5001 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
5003 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
5004 (TARGET_DIRECT_MOVE_64BIT): Adjust.
5005 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
5006 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
5007 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
5008 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
5009 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
5010 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
5011 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
5012 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
5013 anonymous split): Adjust.
5014 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
5015 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5017 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5019 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5020 when --with-gxx-include-dir is also specified.
5021 * configure: Regenerate.
5023 2018-05-09 Jim Wilson <jimw@sifive.com>
5026 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5027 * config/riscv/t-withmultilib: New.
5028 * config/riscv/withmultilib.h: New.
5029 * doc/install.texi: Document RISC-V --with-multilib-list support.
5031 2018-05-09 Richard Biener <rguenther@suse.de>
5033 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5035 (vect_bb_vectorization_profitable_p): Adjust. Compute
5036 actual scalar cost using the cost vector and the add_stmt_cost
5039 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5041 PR rtl-optimization/85645
5042 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5043 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5045 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5047 PR rtl-optimization/85645
5048 * shrink-wrap.c (spread_components): Return a boolean saying if
5049 anything was changed.
5050 (try_shrink_wrapping_separate): Iterate spread_components until
5051 nothing changes anymore.
5053 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5055 PR rtl-optimization/85645
5056 * regrename.c (build_def_use): Also kill the chains that include the
5057 destination of a REG_CFA_REGISTER note.
5059 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5061 PR rtl-optimization/85645
5062 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5063 insn that has a REG_CFA_REGISTER note.
5065 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5067 * cfgexpand.c (expand_clobber): New function.
5068 (expand_gimple_stmt_1): Use it.
5069 * tree-vect-stmts.c (vect_clobber_variable): New function,
5071 (vectorizable_simd_clone_call): ...here.
5072 (vectorizable_store): Emit a clobber either side of an
5073 IFN_STORE_LANES sequence.
5074 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5076 2018-05-09 Tom de Vries <tom@codesourcery.com>
5079 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5080 (define_insn "trap_if_false"): Add exit after trap.
5082 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5084 PR rtl-optimization/85638
5085 * bb-reorder.c: Include common/common-target.h.
5086 (create_forwarder_block): New function extracted from...
5087 (fix_up_crossing_landing_pad): ...here. Rename into...
5088 (dw2_fix_up_crossing_landing_pad): ...this.
5089 (sjlj_fix_up_crossing_landing_pad): New function.
5090 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5091 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5092 from both partitions and exit the loop after one iteration.
5094 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5097 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5099 (Basic PowerPC Built-in Functions): The new name of the
5100 subsection previously known as "PowerPC Built-in Functions".
5101 (Basic PowerPC Built-in Functions Available on all Configurations):
5103 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5105 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5107 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5109 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5112 2018-05-08 Jim Wilson <jimw@sifive.com>
5114 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5115 (LD_EMUL_SUFFIX): New.
5116 (LINK_SPEC): Use it.
5118 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5120 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5122 (Basic PowerPC Built-in Functions): The new name of the
5123 subsection previously known as "PowerPC Built-in Functions".
5124 (Basic PowerPC Built-in Functions Available on all Configurations):
5126 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5128 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5130 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5132 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5135 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5138 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5139 after cmpelim optimization.
5141 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5143 * config.gcc: Support "goldmont".
5144 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5145 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5147 * config/i386/i386.c (m_GOLDMONT): Define.
5148 (processor_target_table): Add "goldmont".
5149 (PTA_GOLDMONT): Define.
5150 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5151 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5152 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5153 (fold_builtin_cpu): Add "goldmont".
5154 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5155 (ix86_option_override_internal): Add "goldmont".
5156 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5157 (processor_type): Add PROCESSOR_GOLDMONT.
5158 * config/i386/i386.md: Add CPU "glm".
5159 * config/i386/glm.md: New file.
5160 * config/i386/x86-tune.def: Add m_GOLDMONT.
5161 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5163 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5166 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5168 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5169 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5170 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5173 * config/i386/i386.c (ix86_fold_builtin): Handle
5174 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5177 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5178 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5180 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5183 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5184 (check_arch): Likewise.
5185 (check_fpu): Return the result rather than printing it.
5186 (end arch): Fix operator precedence.
5187 (end cpu): Likewise.
5188 (END): Print the result from check_fpu.
5190 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5191 Alan Hayward <alan.hayward@arm.com>
5192 David Sherwood <david.sherwood@arm.com>
5194 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5195 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5196 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5197 (*fcmuo<mode>_and): New patterns.
5199 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5201 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5202 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5203 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5204 (cmp_op, sve_imm_con): New code attributes.
5205 (SVE_COND_INT_CMP, imm_con): Delete.
5206 (cmp_op): Remove above unspecs from int attribute.
5207 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5209 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5210 comparison-specific unspecs.
5211 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5212 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5213 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5214 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5215 (*vec_fcm<cmp_op><mode>): Rename to...
5216 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5217 (*vec_fcmuo<mode>): Rename to...
5218 (*fcmuo<mode>): ...this and adjust likewise.
5219 (*pred_fcm<cmp_op><mode>): New pattern.
5220 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5221 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5223 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5225 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5226 (aarch64_emit_sve_predicated_cond): New function.
5227 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5228 (aarch64_emit_unspec_cond_or): Replace with...
5229 (aarch64_emit_sve_or_conds): ...this new function. Use
5230 aarch64_emit_sve_ptrue_op for the individual comparisons and
5231 aarch64_emit_binop to OR them together.
5232 (aarch64_emit_inverted_unspec_cond): Replace with...
5233 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5234 aarch64_emit_sve_ptrue_op for the comparison and
5235 aarch64_emit_unop to invert the result.
5236 (aarch64_expand_sve_vec_cmp_float): Update after the above
5237 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5239 2018-05-07 Nathan Sidwell <nathan@acm.org>
5241 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5242 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5243 (Backwards Compatibility): Likewise.
5245 2018-05-07 Luis Machado <luis.machado@linaro.org>
5249 2018-05-07 Luis Machado <luis.machado@linaro.org>
5251 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5252 <prefetch_dynamic_strides>: New const bool field.
5253 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5254 prefetch_dynamic_strides.
5255 (exynosm1_prefetch_tune): Likewise.
5256 (thunderxt88_prefetch_tune): Likewise.
5257 (thunderx_prefetch_tune): Likewise.
5258 (thunderx2t99_prefetch_tune): Likewise.
5259 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5261 (aarch64_override_options_internal): Update to set
5262 PARAM_PREFETCH_DYNAMIC_STRIDES.
5263 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5264 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5265 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5266 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5267 prefetch-dynamic-strides setting.
5269 2018-05-07 Luis Machado <luis.machado@linaro.org>
5271 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5272 <minimum_stride>: New const int field.
5273 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5274 minimum_stride field.
5275 (exynosm1_prefetch_tune): Likewise.
5276 (thunderxt88_prefetch_tune): Likewise.
5277 (thunderx_prefetch_tune): Likewise.
5278 (thunderx2t99_prefetch_tune): Likewise.
5279 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5280 (aarch64_override_options_internal): Update to set
5281 PARAM_PREFETCH_MINIMUM_STRIDE.
5282 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5283 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5284 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5285 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5286 stride is constant and is below the minimum stride threshold.
5288 2018-05-07 Luis Machado <luis.machado@linaro.org>
5290 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5293 2018-05-07 Luis Machado <luis.machado@linaro.org>
5295 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5296 <prefetch_dynamic_strides>: New const bool field.
5297 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5298 prefetch_dynamic_strides.
5299 (exynosm1_prefetch_tune): Likewise.
5300 (thunderxt88_prefetch_tune): Likewise.
5301 (thunderx_prefetch_tune): Likewise.
5302 (thunderx2t99_prefetch_tune): Likewise.
5303 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5305 (aarch64_override_options_internal): Update to set
5306 PARAM_PREFETCH_DYNAMIC_STRIDES.
5307 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5308 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5309 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5310 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5311 prefetch-dynamic-strides setting.
5313 2018-05-07 Luis Machado <luis.machado@linaro.org>
5315 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5316 <minimum_stride>: New const int field.
5317 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5318 minimum_stride field.
5319 (exynosm1_prefetch_tune): Likewise.
5320 (thunderxt88_prefetch_tune): Likewise.
5321 (thunderx_prefetch_tune): Likewise.
5322 (thunderx2t99_prefetch_tune): Likewise.
5323 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5324 (aarch64_override_options_internal): Update to set
5325 PARAM_PREFETCH_MINIMUM_STRIDE.
5326 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5327 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5328 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5329 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5330 stride is constant and is below the minimum stride threshold.
5332 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5335 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5336 the type is addressable. Don't force op into register if it has
5339 2018-05-05 Roland McGrath <mcgrathr@google.com>
5342 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5343 any section for which we don't know a specific type it should have,
5344 regardless of name. Previously this was done only for the exact
5345 names ".init_array", ".fini_array", and ".preinit_array".
5346 (default_elf_asm_named_section): Add comment about
5347 relationship with default_section_type_flags and SECTION_NOTYPE.
5348 (get_section): Don't consider it a type conflict if one side has
5349 SECTION_NOTYPE and the other doesn't, as long as neither has the
5350 SECTION_BSS et al used in the default_section_type_flags logic.
5352 2018-05-05 Tom de Vries <tom@codesourcery.com>
5355 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5356 (workaround_barsyncs): New function.
5357 (nvptx_reorg): Use workaround_barsyncs.
5358 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5359 (define_expand "nvptx_membar_cta"): New define_expand.
5360 (define_insn "*nvptx_membar_cta"): New insn.
5362 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5364 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5365 To improve optimization opportunities.
5366 * builtin-types.def: The new needed builtin types for the above.
5368 2018-05-04 Richard Biener <rguenther@suse.de>
5370 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5371 * gimple-ssa-store-merging.c
5372 (imm_store_chain_info::output_merged_store): Remove redundant create,
5373 release split_store vector contents on failure.
5374 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5375 scalar stmt vector on cache hit.
5377 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5379 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5381 * config.gcc (powerpc-xilinx-eabi*): Remove.
5382 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5384 (fusion_addis_mem_combo_load): Ditto.
5385 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5387 (rs6000_cpu_cpp_builtins): Ditto.
5388 * config/rs6000/rs6000-linux.c
5389 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5390 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5391 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5393 (rs6000_setup_reg_addr_masks): Ditto.
5394 (rs6000_init_hard_regno_mode_ok): Ditto.
5395 (rs6000_option_override_internal): Ditto.
5396 (legitimate_lo_sum_address_p): Ditto.
5397 (rs6000_legitimize_address): Ditto.
5398 (rs6000_legitimize_reload_address): Ditto.
5399 (rs6000_legitimate_address_p): Ditto.
5400 (abi_v4_pass_in_fpr): Ditto.
5401 (setup_incoming_varargs): Ditto.
5402 (rs6000_gimplify_va_arg): Ditto.
5403 (rs6000_split_multireg_move): Ditto.
5404 (rs6000_savres_strategy): Ditto.
5405 (rs6000_emit_prologue_components): Ditto.
5406 (rs6000_emit_epilogue_components): Ditto.
5407 (rs6000_emit_prologue): Ditto.
5408 (rs6000_emit_epilogue): Ditto.
5409 (rs6000_elf_file_end): Ditto.
5410 (rs6000_function_value): Ditto.
5411 (rs6000_libcall_value): Ditto.
5412 * config/rs6000/rs6000.h: Ditto.
5413 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5414 (TARGET_MINMAX): ... this. New.
5415 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5416 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5417 (*movsi_internal1_single): Delete.
5418 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5419 mfpu=, mxilinx-fpu): Delete.
5420 * config/rs6000/singlefp.h: Delete.
5421 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5422 * config/rs6000/t-rs6000: Ditto.
5423 * config/rs6000/t-xilinx: Delete.
5424 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5425 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5426 (VStype_simple): Delete.
5427 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5428 * config/rs6000/xfpu.h: Delete.
5429 * config/rs6000/xfpu.md: Delete.
5430 * config/rs6000/xilinx.h: Delete.
5431 * config/rs6000/xilinx.opt: Delete.
5432 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5433 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5435 2018-05-04 Tom de Vries <tom@codesourcery.com>
5438 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5441 2018-05-04 Richard Biener <rguenther@suse.de>
5444 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5445 (expand_complex_div_wide): Likewise.
5446 (expand_complex_operations_1): Likewise.
5447 (expand_complex_libcall): Preserve EH info of the original stmt.
5448 (tree_lower_complex): Handle removed blocks.
5449 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5450 on complex multiplication and division libcall builtins.
5452 2018-05-04 Richard Biener <rguenther@suse.de>
5455 * fold-const.c (negate_expr_p): Restrict negation of operand
5456 zero of a division to when we know that can happen without
5458 (fold_negate_expr_1): Likewise.
5460 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5463 * real.h (real_nextafter): Declare.
5464 * real.c (real_nextafter): New function.
5465 * fold-const-call.c (fold_const_nextafter): New function.
5466 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5467 CASE_CFN_NEXTTOWARD.
5468 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5469 even when arg1_mode is different from arg0_mode.
5471 2018-05-03 Nathan Sidwell <nathan@acm.org>
5473 * doc/extend.texi (Deprecated Features): Remove
5475 (Backwards Compatibility): Likewise.
5476 * doc/invoke.texi (C++ Language Options): Likewise.
5477 (C++ Dialect Options): Likewise.
5479 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5482 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5483 _mm512_mask_mullox_epi64): New intrinsics.
5485 2018-05-03 Tom de Vries <tom@codesourcery.com>
5488 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5489 dump files): Add offload-tree.
5491 2018-05-03 Richard Biener <rguenther@suse.de>
5493 PR tree-optimization/85615
5494 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5495 to loops not nested in BBs loop father to avoid creating multi-entry
5498 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5500 PR tree-optimization/70291
5501 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5502 arguments. Change return type to tree. Emit libcall as a new
5503 statement rather than replacing existing one when inplace_p is true.
5504 (expand_complex_multiplication_components): New function.
5505 (expand_complex_multiplication): Expand floating-point complex
5506 multiplication using the above.
5507 (expand_complex_division): Rename inner_type parameter to type.
5508 Update expand_complex_libcall call-site.
5509 (expand_complex_operations_1): Update expand_complex_multiplication
5510 and expand_complex_division call-sites.
5512 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5515 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5516 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5517 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5518 the highest significant bit of the shift count mask is clear. In
5519 check whether and[sq]i3 is needed verify that all significant bits
5520 of the shift count other than the highest are set.
5522 2018-05-02 Tom de Vries <tom@codesourcery.com>
5525 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5526 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5527 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5528 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5529 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5530 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5531 __builtin_goacc_parlevel_size.
5533 2018-05-02 Richard Biener <rguenther@suse.de>
5535 PR tree-optimization/85597
5536 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5537 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5540 2018-05-02 Tom de Vries <tom@codesourcery.com>
5543 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5544 dump files): Add ltrans-tree.
5546 2018-05-02 Tom de Vries <tom@codesourcery.com>
5549 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5550 dump files): Add wpa-ipa.
5552 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5554 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5555 powerpc*-*-linux*paired* target.
5556 * config/rs6000/750cl.h: Delete.
5557 * config/rs6000/paired.h: Delete.
5558 * config/rs6000/paired.md: Delete.
5559 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5561 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5562 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5563 comment. Remove paired float support.
5564 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5565 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5567 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5568 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5570 * config/rs6000/rs6000.c: Remove paired float support.
5571 (paired_expand_vector_init, paired_expand_vector_move,
5572 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5573 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5574 paired_expand_builtin, paired_expand_predicate_builtin,
5575 paired_init_builtins): Delete.
5576 * config/rs6000/rs6000.h: Remove paired float support.
5577 * config/rs6000/rs6000.md: Remove paired float support.
5578 (move_from_CR_ov_bit): Delete.
5579 * config/rs6000/rs6000.opt (mpaired): Delete.
5580 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5581 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5583 2018-05-02 Richard Biener <rguenther@suse.de>
5586 * gimplify.c (gimplify_save_expr): When in SSA form allow
5587 SAVE_EXPRs to compute to SSA vars.
5589 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5592 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5593 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5594 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5595 clobber operands[2], instead use a new pseudo. Formatting fixes.
5597 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5599 PR tree-optimization/85586
5600 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5601 exit early for statements in the same group if the accesses are
5604 2018-05-02 Tom de Vries <tom@codesourcery.com>
5607 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5610 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5612 PR tree-optimization/85143
5613 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5615 2018-05-01 Tom de Vries <tom@codesourcery.com>
5618 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5619 not found" error message.
5621 2018-05-01 Tom de Vries <tom@codesourcery.com>
5624 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5625 * vec.c (test_ordered_remove_if): New function.
5626 (vec_c_tests): Call test_ordered_remove_if.
5627 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5628 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5629 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5630 VEC_ORDERED_REMOVE_IF.
5632 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5634 PR tree-optimization/82665
5635 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5636 pointer subtraction where arguments come from a memchr call.
5638 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5640 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5641 --push-state --as-needed and --pop-state instead of --as-needed and
5642 --no-as-needed if ld supports it.
5643 * configure: Regenerated.
5646 * doc/install.texi2html: Replace _002d with - and _002a with * in
5647 generated html files using sed.
5649 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5652 * gcc-rich-location.c (blank_line_before_p): New function.
5653 (use_new_line): New function.
5654 (gcc_rich_location::add_fixit_insert_formatted): New function.
5655 * gcc-rich-location.h
5656 (gcc_rich_location::add_fixit_insert_formatted): New function.
5658 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5660 * selftest.c (assert_streq): Rename "expected" and "actual" to
5661 "val1" and "val2". Extend NULL-handling to cover both inputs
5662 symmetrically, while still requiring both to be non-NULL for a pass.
5663 * selftest.h (assert_streq): Rename "expected" and "actual" to
5665 (ASSERT_EQ): Likewise.
5666 (ASSERT_EQ_AT): Likewise.
5667 (ASSERT_KNOWN_EQ): Likewise.
5668 (ASSERT_KNOWN_EQ_AT): Likewise.
5669 (ASSERT_NE): Likewise.
5670 (ASSERT_MAYBE_NE): Likewise.
5671 (ASSERT_MAYBE_NE_AT): Likewise.
5672 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5673 the assertion to pass.
5674 (ASSERT_STREQ_AT): Likewise.
5676 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5678 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5679 interaction with -pie.
5681 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5683 * selftest.h: Fix alphabetization of per-source-file selftest
5686 2018-04-30 Jason Merrill <jason@redhat.com>
5688 PR c++/61982 - dead stores to destroyed objects.
5689 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5692 2018-04-30 Jason Merrill <jason@redhat.com>
5694 * tree.c (build_clobber): New.
5695 * tree.h: Declare it.
5696 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5698 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5700 * diagnostic-show-locus.c (layout::layout): Update for
5701 location_get_source_line returning a char_span.
5702 (struct char_span): Move to input.h.
5703 (struct correction): Update for fields in char_span becoming
5705 (struct source_line): Update for location_get_source_line
5706 returning a char_span.
5707 (layout::print_line): Likewise.
5708 * edit-context.c (edited_file::print_content): Likewise.
5709 (edited_file::print_diff_hunk): Likewise.
5710 (edited_file::print_run_of_changed_lines): Likewise.
5711 (edited_file::get_num_lines): Likewise.
5712 (edited_line::edited_line): Likewise.
5713 * final.c (asm_show_source): Likewise.
5714 * input.c (location_get_source_line): Convert return type
5715 from const char * to char_span, losing the final "line_len"
5717 (dump_location_info): Update for the above.
5718 (get_substring_ranges_for_loc): Likewise. Use a char_span
5719 when handling the literal within the line.
5720 (test_reading_source_line): Update for location_get_source_line
5721 returning a char_span.
5722 * input.h (class char_span): Move here from
5723 diagnostic-show-locus.c, converting from a struct to a class.
5724 Make data members private.
5725 (char_span::operator bool): New.
5726 (char_span::length): New.
5727 (char_span::get_buffer): New.
5728 (char_span::operator[]): New.
5729 (char_span::subspan): Make const.
5730 (char_span::xstrdup): New.
5731 (location_get_source_line): Convert return type from const char *
5732 to char_span, losing the final "line_size" param.
5734 2018-04-30 Jan Hubicka <jh@suse.cz>
5736 * lto-wrapper.c (ltrans_priorities): New static var.
5737 (cmp_priority): New.
5738 (run_gcc): Read priorities and if doing parallel build order
5739 the Makefile by them.
5741 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5743 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5745 2018-04-30 Richard Biener <rguenther@suse.de>
5747 * tree-cfg.c (verify_address): Remove base argument, add
5748 flag whether to check TREE_ADDRESSABLE and do that.
5749 (verify_expr): Remove.
5750 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5751 (verify_gimple_assign_single): Likewise.
5752 (verify_gimple_switch): Likewise.
5753 (verify_expr_location_1): Dereference tp once. Add (disabled)
5754 piece from verify_expr.
5755 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5757 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5759 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5761 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5763 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5764 (small_data_pattern): Likewise.
5765 (arc_rewrite_small_data): Likewise.
5766 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5767 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5768 (get_symbol_alignment): New function.
5769 (legitimate_small_data_address_p): Likewise.
5770 (legitimate_scaled_address): Update, call
5771 legitimate_small_data_address_p.
5772 (output_sdata): New static variable.
5773 (arc_print_operand): Update how we handle small data operands.
5774 (arc_print_operand_address): Likewise.
5775 (arc_legitimate_address_p): Update, use
5776 legitimate_small_data_address_p.
5777 (arc_rewrite_small_data_p): Remove.
5778 (arc_rewrite_small_data_1): Likewise.
5779 (arc_rewrite_small_data): Likewise.
5780 (small_data_pattern): Likewise.
5781 (compact_sda_memory_operand): Update to use
5782 legitimate_small_data_address_p and get_symbol_alignment.
5783 (prepare_move_operands): Don't rewite sdata pattern.
5784 (prepare_extend_operands): Remove.
5785 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5787 (zero_extendqisi2): Likewise.
5788 (zero_extendhisi2): Likewise.
5789 (extendqihi2): Likewise.
5790 (extendqisi2): Likewise.
5791 (extendhisi2): Likewise.
5795 * config/arc/constraints.md (Usd): Change it to memory constraint.
5797 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5799 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5800 as source of std instructions.
5801 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5802 allow 6-bit constants as source for store instructions.
5803 (movdi_insn): Update instruction pattern to allow 6-bit constants
5804 as source for store instructions.
5806 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5808 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5810 2018-04-30 Nathan Sidwell <nathan@acm.org>
5811 Sandra Loosemore <sandra@codesourcery.com>
5813 * dumpfile.c (dump_open): Allow '-' for stdout.
5814 * doc/invoke.texi (Developer Options): Document dump filename
5815 determination early. Document stdin/stdout selection.
5817 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5819 Microblaze Target: PIC data text relative
5821 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5822 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5824 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5825 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5826 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5827 New addressing mode for data-text relative position indepenedent code.
5828 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5829 'ADDRESS_SYMBOLIC_TXT_REL'.
5830 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5831 (microblaze_legitimate_pic_operand): Exclude function calls from
5832 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5833 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5835 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5836 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5837 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5838 for 'address + offset'.
5839 (microblaze_expand_prologue): Add new function prologue call for
5841 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5842 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5843 table in case of TARGET_PIC_DATA_TEXT_REL.
5844 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5845 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5846 Add new macros 'UNSPEC_TEXT',
5847 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5848 + exclude function calls from 'UNSPEC_PLT' in case of data text
5850 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5851 new target hook for generating address diff vector tables in case of
5853 * doc/tm.texi : Regenerate.
5854 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5855 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5856 of addr diff vector generation.
5857 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5858 target hook definition.
5859 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5860 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5861 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5862 option for data text relative.
5864 2018-04-30 Richard Biener <rguenther@suse.de>
5866 * tree-chrec.h (evolution_function_is_constant_p): Remove
5868 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5870 2018-04-30 Richard Biener <rguenther@suse.de>
5873 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5875 2018-04-30 Richard Biener <rguenther@suse.de>
5877 PR tree-optimization/28364
5878 PR tree-optimization/85275
5879 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5880 copying first exit test.
5882 2018-04-28 Mark Wielaard <mark@klomp.org>
5884 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5886 (dwarf_AT): Handle DW_AT_addr_base.
5887 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5889 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5892 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5893 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5894 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5895 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5897 2018-04-28 Richard Biener <rguenther@suse.de>
5899 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5900 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5901 to reflect use. Only add interesting stmts.
5903 2018-04-27 Martin Jambor <mjambor@suse.cz>
5906 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5907 the jump function allows for passing through aggregate values.
5909 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5911 * input.h (in_system_header_at): Convert from macro to inline
5913 (from_macro_expansion_at): Likewise.
5914 (from_macro_definition_at): Likewise.
5916 2018-04-27 Jeff Law <law@redhat.com>
5918 * config.gcc: Mark tile* targets as deprecated/obsolete.
5920 2018-04-27 Richard Biener <rguenther@suse.de>
5922 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5925 2018-04-27 Richard Biener <rguenther@suse.de>
5927 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5929 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5931 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5932 with Yd constraint. Set "preferred_for_speed" attribute from
5933 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5935 (*movdi_internal): Ditto.
5936 (movti_interunit splitters): Remove
5937 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5938 (movdi_interunit splitters): Ditto.
5939 * config/i386/constraints.md (Ye): Remove.
5940 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5942 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5945 * config/aarch64/constraints.md (Usg): Limit to 31.
5948 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5950 PR tree-optimization/85529
5951 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5952 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5953 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5954 zero extension or masking of the MSB bit.
5955 (optimize_range_tests): Add FIRST_BB argument, pass it through
5956 to optimize_range_tests_var_bound.
5957 (maybe_optimize_range_tests, reassociate_bb): Adjust
5958 optimize_range_tests callers.
5960 2018-04-26 Richard Biener <rguenther@suse.de>
5961 Jakub Jelinek <jakub@redhat.com>
5963 * cgraph.h (symbol_table): Just declare debug method here.
5964 * symtab.c (symbol_table::debug): Define.
5966 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5968 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5970 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5972 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5973 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5974 (*movdi_internal): Substitute Yi and Yj constraint with x
5975 and Ym and Yn constraint with y constraint. Update "isa"
5976 attribute and set "preferred_for_speed" attribute from
5977 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5978 (*movsi_internal): Ditto.
5979 (*movdf_internal): Ditto.
5980 (*movsf_internal): Ditto.
5981 (*zero_extendsidi2): Ditto.
5982 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5983 (sse2_loadld): Ditto.
5984 (*vec_extract<ssevecmodelower>_0): Ditto.
5985 (*vec_extractv4si_0_zext_sse4): Ditto.
5986 (vec_concatv2di): Ditto.
5987 (*vec_dup<mode>): Ditto.
5988 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5989 * config/i386/constraints.md (Yi): Remove.
5994 2018-04-26 Nathan Sidwell <nathan@acm.org>
5996 * dumpfile.c (dump_open): New.
5997 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5998 (dump_finish): Detect stdio/stderr by value not name.
6000 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
6002 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
6004 2018-04-26 Tom de Vries <tom@codesourcery.com>
6007 * config/nvptx/nvptx.c (verify_neutering_jumps)
6008 (verify_neutering_labels): New function
6009 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
6011 2018-04-26 Tom de Vries <tom@codesourcery.com>
6014 * config/nvptx/nvptx.c (needs_neutering_p): New function.
6015 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6018 2018-04-26 Richard Biener <rguenther@suse.de>
6019 Tom de Vries <tom@codesourcery.com>
6022 * lto-streamer-out.c (output_function): Fixup loops if required to match
6023 discovery done in the reader.
6025 2018-04-26 Richard Biener <rguenther@suse.de>
6027 PR tree-optimization/85116
6028 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6029 have a loop exit from the single latch predecessor. Remove
6030 case of header with just condition.
6031 (ch_base::copy_headers): Exclude infinite loops from any
6033 (pass_ch::execute): Record exits.
6035 2018-04-26 Richard Biener <rguenther@suse.de>
6037 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6038 prologue cost vector and pass it to vect_get_load_cost.
6039 (vect_get_peeling_costs_all_drs): Likewise.
6040 (vect_peeling_hash_get_lowest_cost): Likewise.
6041 (vect_enhance_data_refs_alignment): Likewise.
6043 2018-04-26 Richard Biener <rguenther@suse.de>
6046 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6047 checking of integer<->pointer conversions.
6048 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6049 sign-/zero-extending pointer types.
6050 (expand_omp_for_static_chunk): Likewise.
6052 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6053 Jean Lee <xiaoyur347@gmail.com>
6055 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6056 (TARGET_ASAN_SHADOW_OFFSET): Define.
6057 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6058 true for -fsanitize=address.
6060 2018-04-25 Mark Wielaard <mark@klomp.org>
6062 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6065 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6067 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6068 than "alu", remove explicit "memory" and "imm_disp" attributes.
6069 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6072 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6073 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6074 gen_lowpart_no_emit.
6076 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6079 * config/i386/i386.c (ix86_expand_builtin): Change memory
6080 operand to XI, extend p0 to Pmode.
6081 * config/i386/i386.md: Change unspec volatile and operand
6082 1 mode to XI, change operand 0 mode to P.
6084 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6086 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6087 GET_MODE_MASK before any checking.
6088 (nds32_can_use_bset_p): Likewise.
6089 (nds32_can_use_btgl_p): Likewise.
6091 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6093 * config/nds32/nds32-doubleword.md: New define_split pattern for
6094 illegal register number.
6096 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6098 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6100 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6102 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6104 2018-04-25 Richard Biener <rguenther@suse.de>
6106 * lto-streamer.h (LTO_major_version): Bump to 8.
6108 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6110 * BASE-VER: Set to 9.0.0.
6112 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6114 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6115 in __abskf2 and __powikf2.
6117 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6120 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6121 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6122 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6123 Use the above on operand 2. Reindent.
6124 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6126 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6129 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6131 * config/i386/i386.opt (mcet): Removed.
6132 * doc/install.texi: Remove -mcet documentation.
6133 * doc/invoke.texi: Likewise.
6135 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6138 * doc/install.texi: Remove -mcet from bootstrap-cet.
6140 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6143 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6144 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6148 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6149 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6150 containing a CONST_VECTOR.
6152 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6154 * doc/install.texi: Update newlib dependency for nvptx.
6156 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6159 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6160 instead of INTVAL when shifting x left.
6162 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6164 PR tree-optimization/85478
6165 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6166 vect_grouped_store_supported for single element vectors.
6168 2018-04-24 Richard Biener <rguenther@suse.de>
6171 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6172 load cost increase to the case of non-constant step.
6174 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6177 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6178 destination if any_malformed_asm.
6180 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6183 * expr.c (store_field): In the bitfield case, if the value comes from
6184 a function call and is returned in registers by means of a PARALLEL,
6185 do not change the mode of the temporary unless BLKmode and VOIDmode.
6187 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6189 PR rtl-optimization/85423
6190 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6191 dependencies to debug insns when the previous insn is non-debug.
6193 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6195 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6196 enums into a single definition.
6197 (fls): Fix predicates and printing.
6200 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6202 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6203 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6204 and short u6 immediate.
6205 (check_if_valid_sleep_operand): Remove.
6206 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6208 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6210 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6211 flag_always_save_lp condition.
6212 * config/nds32/nds32.opt (malways-save-lp): New option.
6214 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6216 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6217 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6218 * config/nds32/nds32.h
6219 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6220 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6222 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6224 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6225 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6227 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6228 Chung-Ju Wu <jasonwucj@gmail.com>
6230 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6232 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6233 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6235 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6237 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6239 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6241 * config/nds32/nds32-protos.h (nds32_data_alignment,
6242 nds32_local_alignment): Declare.
6243 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6244 nds32_local_alignment): New functions.
6245 (TARGET_CONSTANT_ALIGNMENT): Define.
6246 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6248 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6250 * config/nds32/nds32.c
6251 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6252 (TARGET_MODES_TIEABLE_P): Likewise.
6254 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6256 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6259 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6260 Chung-Ju Wu <jasonwucj@gmail.com>
6262 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6263 for unaligned access.
6264 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6266 * config/nds32/nds32-intrinsic.md: Likewise.
6267 * config/nds32/nds32_intrinsic.h: Likewise.
6268 * config/nds32/nds32.h (nds32_builtins): Likewise.
6269 * config/nds32/nds32.opt (munaligned-access): New option.
6270 * config/nds32/nds32.c (nds32_asm_file_start): Display
6271 flag_unaligned_access status.
6273 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6275 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6276 -mno-relax is present.
6277 * config/riscv/linux.h (LINK_SPEC): Ditto.
6279 2018-04-20 Martin Sebor <msebor@redhat.com>
6282 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6284 (gimple_fold_builtin_stxcpy_chk): Same.
6285 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6287 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6290 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6291 __powikf2 when long double is IEEE 128-bit.
6293 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6295 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6296 step to make sure stack always aligned.
6298 2018-04-20 Carl Love <cel@us.ibm.com>
6301 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6302 size check for arg0.
6304 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6305 Tom de Vries <tom@codesourcery.com>
6308 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6309 Emit insns for calls too.
6310 (nvptx_find_par): Always look for worker-level predecessor insn.
6311 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6313 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6314 (nvptx_process_pars): Propagate frames for calls.
6316 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6319 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6321 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6322 (ix86_handle_option): Don't handle OPT_mibt.
6323 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6325 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6327 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6328 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6329 (ix86_target_macros): Define __CET__ with flag_cf_protection
6330 for -fcf-protection.
6331 * config/i386/i386.c (isa2_opts): Remove -mibt.
6332 * config/i386/i386.h (TARGET_IBT): Removed.
6333 (TARGET_IBT_P): Likewise.
6334 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6335 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6336 * config/i386/i386.opt (mcet): Update help message.
6339 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6340 -mcet as an alias for -mshstk.
6342 2018-04-20 Richard Biener <rguenther@suse.de>
6345 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6346 complexity by forcing a single use of the multiply operand.
6348 2018-04-20 Martin Jambor <mjambor@suse.cz>
6351 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6352 recursion dependency to only apply to non-clones.
6354 2018-04-20 Martin Jambor <mjambor@suse.cz>
6357 * ipa-cp.c (create_specialized_node): Check that clones of
6358 self-recursive edges exist during IPA-CP.
6360 2018-04-19 Toon Moene <toon@moene.org>
6362 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6365 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6367 PR tree-optimization/85467
6368 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6369 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6370 VECTOR_CST element to type.
6372 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6375 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6376 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6377 (builtin_longjmp): Likewise.
6378 (save_stack_nonlocal): New pattern.
6379 (restore_stack_nonlocal): Likewise.
6381 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6384 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6385 Replace ASM_OUTPUT_LABEL with fprintf.
6387 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6390 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6391 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6392 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6393 define __IBT__ and __SHSTK__ for -fcf-protection.
6394 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6396 (ix86_trampoline_init): Likewise.
6397 (x86_output_mi_thunk): Likewise.
6398 (ix86_notrack_prefixed_insn_p): Likewise.
6399 (ix86_option_override_internal): Don't disallow -fcf-protection.
6400 * config/i386/i386.md (rdssp<mode>): Also enable for
6402 (incssp<mode>): Likewise.
6403 (nop_endbr): Likewise.
6404 * config/i386/i386.opt (mcet): Change help message to built-in
6408 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6409 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6410 enable CET built-in functions.
6412 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6414 * common/config/i386/i386-common.c
6415 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6416 OPTION_MASK_ISA_MOVDIRI_UNSET,
6417 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6418 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6419 * config.gcc (movdirintrin.h): New header.
6420 * config/i386/cpuid.h (bit_MOVDIRI,
6421 bit_MOVDIR64B): New bits.
6422 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6424 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6425 (VOID, PVOID, PCVOID)): New function types.
6426 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6427 __builtin_ia32_directstoreu_u64,
6428 __builtin_ia32_movdir64b): New builtins.
6429 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6430 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6432 (ix86_valid_target_attribute_inner_p): Ditto.
6433 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6434 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6435 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6436 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6437 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6438 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6439 (movdiri<mode>, movdir64b_<mode>): New.
6440 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6441 * config/i386/immintrin.h: Include movdirintrin.h.
6442 * config/i386/movdirintrin.h: New file.
6443 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6445 2018-04-19 Richard Biener <rguenther@suse.de>
6448 * cfg.c (clear_bb_flags): When loop state says we have
6449 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6451 2018-04-19 Richard Biener <rguenther@suse.de>
6453 PR tree-optimization/84737
6454 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6455 copying restrict info.
6456 (vect_setup_realignment): Use it.
6457 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6458 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6459 the first DR to all generated stores.
6460 (vectorizable_load): Likewise for loads.
6462 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6464 PR tree-optimization/85446
6465 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6466 the integral and pointer types to have the same precision.
6468 * doc/install.texi: Document --disable-cet being the default and
6471 2018-04-18 Martin Liska <mliska@suse.cz>
6473 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6476 2018-04-18 Martin Liska <mliska@suse.cz>
6479 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6482 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6483 arguments if they are comparable.
6485 2018-04-18 Martin Liska <mliska@suse.cz>
6488 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6491 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6494 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6497 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6498 ENDBR after calling __morestack.
6500 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6503 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6504 by using gcc_base_ver to generate a gcc_driver_version, and use
6505 it when generating GCC_DRIVER_NAME.
6506 * configure: Regenerate.
6508 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6511 * config.gcc: Obsolete powerpc*-*-*spe*.
6513 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6516 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6517 (stabstr_D): Change type of unum from unsigned int to
6518 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6521 2018-04-17 Jim Wilson <jimw@sifive.com>
6524 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6525 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6526 Set arg_pointer_offset after using pretend_args_size.
6528 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6530 PR rtl-optimization/85431
6531 * dse.c (record_store): Ignore zero width stores.
6534 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6535 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6536 __builtin_stack_restore rather than after it.
6537 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6538 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6539 argument instead of virtual_dynamic_stack_rtx.
6541 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6543 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6545 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6546 Add note to error message to explain internal mapping of overloaded
6547 built-in function name to non-overloaded built-in function name.
6548 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6551 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6554 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6555 where the inputs overlap with the output.
6557 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6559 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6560 (=v, v) alternative and explicit "memory" attribute.
6561 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6562 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6564 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6565 "sselog1" type instead of "sselog".
6566 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6567 "sselog". Remove explicit "memory" attribute.
6568 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6569 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6571 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6572 "sselog1" type instead of "sselog". Remove explicit "memory"
6574 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6575 use "sselog1" type instead of "sselog". Remove explicit "memory"
6577 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6578 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6580 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6581 "sselog1" type instead of "sselog". Remove explicit "memory"
6583 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6584 use "sselog1" type instead of "sselog". Remove explicit "memory"
6588 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6591 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6594 2018-04-17 Martin Jambor <mjambor@suse.cz>
6597 * ipa-cp.c (create_specialized_node): Call
6598 expand_all_artificial_thunks if necessary.
6600 2018-04-17 Martin Liska <mliska@suse.cz>
6603 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6604 in message, remote space in between '_G' and '('.
6606 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6609 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6610 avx512f_vmcmp<mode>3<round_saeonly_name>,
6611 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6612 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6613 avx512f_rndscale<mode><round_saeonly_name>,
6614 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6615 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6616 Use %<iptr>2 instead of %2 for -masm=intel.
6617 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6618 avx512f_vcvttss2usi<round_saeonly_name>,
6619 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6621 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6622 avx512f_vcvttsd2usi<round_saeonly_name>,
6623 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6624 Use %q1 instead of %1 for -masm=intel.
6625 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6626 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6627 of %3 for -masm=intel.
6628 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6630 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6632 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6634 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6635 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6637 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6639 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6640 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6641 %g1 and one with %0 and %1.
6642 (avx512er_vmrcp28<mode><round_saeonly_name>,
6643 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6645 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6646 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6647 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6648 of %0 and %{%4%} for -masm=intel.
6649 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6650 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6651 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6652 order of %0 and %{%5%}%{z%} for -masm=intel.
6654 2018-04-17 Jan Hubicka <jh@suse.cz>
6657 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6659 2018-04-17 Martin Liska <mliska@suse.cz>
6662 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6663 for target_clone error message. Make default implementation
6664 clone to be a local declaration.
6665 (separate_attrs): Add new argument and check for an empty
6667 (expand_target_clones): Handle it.
6668 (ipa_target_clone): Make redirection just for target_clones
6671 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6672 Tom de Vries <tom@codesourcery.com>
6675 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6676 tiled basic blocks without omp continue statements.
6678 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6681 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6682 vec_extract expression as having side effects to make sure it gets
6685 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6688 * config/i386/i386.c (get_builtin_code_for_version): Check
6691 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6694 * gcc/config.gcc: Support "skylake".
6695 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6697 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6698 (processor_target_table): Add "skylake".
6699 (ix86_option_override_internal): Add "skylake".
6700 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6701 PROCESSOR_CANNONLAKE.
6702 (get_builtin_code_for_version): Fix priority for
6703 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6704 PROCESSOR_SKYLAKE-AVX512.
6705 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6706 (processor_type): Add PROCESSOR_SKYLAKE.
6708 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6709 Jason Merrill <jason@redhat.com>
6712 * convert.c (convert_to_integer_1): Use direct recursion for
6713 enumeral types and types with a precision less than the number
6714 of bits in their mode.
6716 2018-04-16 Julia Koval <julia.koval@intel.com>
6719 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6720 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6722 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6725 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6726 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6727 and -mno-direct-move.
6729 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6732 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6733 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6734 Ensure negative shifts result in {0}.
6736 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6738 PR rtl-optimization/79916
6739 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6740 regs (if any) to define how to gnerate SD moves when LRA is in
6743 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6745 PR rtl-optimization/85393
6746 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6747 * except.c (expand_dw2_landing_pad_for_region): Make static.
6748 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6749 a label and unconditional jump to old_bb, rather than
6750 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6753 PR rtl-optimization/85376
6754 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6755 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6756 instead of a specific value.
6758 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6759 Bin Cheng <bin.cheng@arm.com>
6761 PR tree-optimization/82965
6762 PR tree-optimization/83991
6763 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6764 by_profile_only parameter.
6765 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6766 information if the loop was predicted to iterate too many times.
6767 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6769 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6772 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6775 2018-04-13 Martin Liska <mliska@suse.cz>
6776 Jakub Jelinek <jakub@redhat.com>
6779 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6780 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6781 * builtins.c (expand_builtin_memory_copy_args): Use
6782 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6783 handle dest_addr == pc_rtx.
6785 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6788 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6789 asked to not generate direct moves.
6790 (fix_trunc<mode>si2_stfiwx): Similar.
6791 (fix_trunc<mode>si2_internal): Similar.
6793 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6796 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6797 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6798 lookup if dest in some wider mode is known to be const0_rtx and
6799 if so, record permanent equivalence for it to be ZERO_EXTEND of
6800 the narrower mode destination.
6802 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6804 * lto-streamer-out.c (output_function): Revert 259346.
6805 * omp-expand.c (expand_oacc_for): Likewise.
6807 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6809 PR rtl-optimization/85354
6810 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6811 * sel-sched.c (sel_global_init): ... here.
6813 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6816 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6817 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6818 mode for PE-COFF targets.
6819 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6820 (i386_pe_asm_lto_end): Likewise.
6821 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6822 (TARGET_ASM_LTO_END): Likewise.
6823 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6824 (i386_pe_asm_lto_start): New function.
6825 (i386_pe_asm_lto_end): Likewise.
6827 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6828 Richard Biener <rguenther@suse.de>
6831 * lto-streamer-out.c (output_function): Fix CFG loop state before
6833 * omp-expand.c (expand_oacc_for): Handle calls to internal
6834 functions like regular functions.
6836 2018-04-12 Richard Biener <rguenther@suse.de>
6839 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6840 for the early LTO debug to properly generate references to it
6841 during DIE emission. Do not re-use that for the skeleton for
6843 (dwarf2out_early_finish): Likewise.
6845 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6848 * config/i386/sse.md
6849 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6850 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6851 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6852 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6853 and output is a reg, avoid creating invalid lowpart subreg, but
6854 instead split into a 512-bit move. Don't split if not AVX512VL,
6855 input is xmm16+ reg and output is a mem.
6856 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6857 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6858 xmm16+ reg and output is a mem.
6860 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6862 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6863 also for flag_dwarf2_cfi_asm.
6865 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6867 PR rtl-optimization/85342
6868 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6869 a bool scalar var inside of the loop instead. Don't try to update
6870 recog_data.operand after failed apply_change_group.
6872 2018-04-12 Tom de Vries <tom@codesourcery.com>
6875 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6876 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6877 array with flexible array member as array without given dimension.
6878 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6879 argument for undefined param to true.
6881 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6884 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6885 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6886 from PowerPC section.
6887 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6888 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6889 help text that is too long.
6890 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6891 help text that is too long.
6892 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6893 help text that is too long.
6895 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6897 * config/alpha/alpha.md (stack_probe_internal): Rename
6898 from "probe_stack". Update all callers.
6900 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6902 PR rtl-optimization/84566
6903 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6904 sched_macro_fuse_insns.
6906 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6909 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6910 (compute_block_dependences): ... from here.
6912 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6914 PR tree-optimization/85331
6915 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6916 from int to HOST_WIDE_INT.
6918 2018-04-11 Martin Jambor <mjambor@suse.cz>
6921 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6922 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6923 not the same as the source val.
6924 (cgraph_edge_brings_value_p): New parameter.
6925 (gather_edges_for_value): Pass destination value to
6926 cgraph_edge_brings_value_p.
6927 (perhaps_add_new_callers): Likewise.
6928 (get_info_about_necessary_edges): Likewise and exclude values brought
6929 only by self-recursive edges.
6930 (create_specialized_node): Redirect only clones of self-calling edges.
6931 (+self_recursive_pass_through_p): New function.
6932 (find_more_scalar_values_for_callers_subset): Use it.
6933 (find_aggregate_values_for_callers_subset): Likewise.
6934 (known_aggs_to_agg_replacement_list): Removed.
6935 (decide_whether_version_node): Re-calculate known constants for all
6936 remaining context clones.
6938 2018-04-11 Richard Biener <rguenther@suse.de>
6941 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6942 from early DWARF output.
6943 (dwarf2out_early_finish): Output line info unconditionally into
6944 early DWARF and add reference to it.
6946 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6949 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6950 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6951 other than V2DFmode using iptr mode attribute.
6952 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6954 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6956 PR rtl-optimization/84659
6957 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6959 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6962 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6964 (output_loc_list): Pass address of a dummy size variable even in the
6965 locview handling loop.
6966 (index_location_lists): Add comment on why skip_loc_list_entry can't
6969 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6972 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6975 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6978 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6979 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6980 and -mstring-compare-inline-limit.
6982 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6985 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6986 for stack clash protection in a register whenever we need it to be in
6989 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6991 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6992 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6994 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6997 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6999 (mlong-double-): Ditto.
7000 * config/rs6000/sysv4.opt (msdata=): Ditto.
7001 (mtls-size=): Ditto.
7003 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7005 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
7006 erroneous entries for
7007 "vector int vec_ldl (int, long int *)", and
7008 "vector unsigned int vec_ldl (int, unsigned long int *)".
7009 Add comments and entries for
7010 "vector bool char vec_ldl (int, bool char *)",
7011 "vector bool short vec_ldl (int, bool short *)",
7012 "vector bool int vec_ldl (int, bool int *)",
7013 "vector bool long long vec_ldl (int, bool long long *)",
7014 "vector pixel vec_ldl (int, pixel *)",
7015 "vector long long vec_ldl (int, long long *)",
7016 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7017 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7018 type tree bool_long_long_type_node and correct definition of
7019 bool_V2DI_type_node to make reference to this new type tree.
7020 (rs6000_mangle_type): Replace erroneous reference to
7021 bool_long_type_node with bool_long_long_type_node.
7022 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7023 comments to emphasize sign distinctions for char and int types and
7024 replace RS6000_BTI_bool_long constant with
7025 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7026 use of RS6000_BTI_pixel.
7027 (bool_long_type_node): Remove this macro definition.
7028 (bool_long_long_type_node): New macro definition
7030 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7032 PR rtl-optimization/85300
7033 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7034 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7035 simplify_unary_operation fails.
7037 2018-04-10 Martin Liska <mliska@suse.cz>
7039 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7040 cgraph_edge and ipa_ref.
7042 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7046 * config/i386/sse.md
7047 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7048 computation of the VEC_MERGE selector from mask.
7049 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7050 Fix decoding of the VEC_MERGE selector into mask.
7052 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7054 PR tree-optimization/85286
7055 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7057 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7059 * final.c (final_1): Set insn_last_address as well as
7060 insn_current_address.
7062 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7065 * explow.c (emit_stack_probe): Call validize_mem on memory location
7066 before passing it to gen_probe_stack. Create address operand and
7067 legitimize it for the probe_stack_address case.
7069 2018-04-09 Jan Hubicka <jh@suse.cz>
7072 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7073 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7074 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7076 (free_lang_data): Rebuild type inheritance graph.
7078 2018-04-09 Martin Sebor <msebor@redhat.com>
7080 * invoke.texi (-finline-small-functions): Mention other optimization
7082 (-findirect-inlining, -fpartial-inlining): Same.
7083 (-finline-functions-called-once): Same.
7084 (-freorder-blocks-and-partition): Same.
7086 2018-04-09 Jan Hubicka <jh@suse.cz>
7089 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7090 jumps; choose last target that matches the criteria (i.e.
7091 no partition changes for non-crossing jumps).
7092 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7093 support for redirecting crossing jumps to non-crossing.
7095 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7097 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7098 also for naked functions.
7100 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7102 * config/arc/arc.md (add_shift): New pattern.
7103 (add_shift2): Likewise.
7104 (sub_shift): Likewise.
7105 (sub_shift_cmp0_noout): Likewise.
7106 (compare_si_ashiftsi): Likewise.
7107 (xbfu_cmp0_noout): New combine pattern.
7108 (xbfu_cmp0"): Likewise.
7109 (movsi_set_cc_insn): Place the predicable variant first.
7110 (commutative_binary_cmp0_noout): Remove clobber.
7111 (commutative_binary_cmp0): New pattern.
7112 (noncommutative_binary_cmp0): Likewise.
7113 (noncommutative_binary_cmp0_noout): Likewise.
7114 (noncommutative_binary_comparison_result_used): Removed.
7115 (rsub_cmp0): New pattern.
7116 (rsub_cmp0_noout): Likewise.
7117 (extzvsi): Changed, keep only meaningful variants.
7118 (SQH, SEZ): New iterators.
7119 (SQH_postfix): New mode attribute.
7120 (SEZ_prefix): New code attribute.
7121 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7122 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7123 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7125 (noncommutative_operator): Check the availability of barrel
7128 2018-04-09 Richard Biener <rguenther@suse.de>
7130 PR tree-optimization/85284
7131 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7132 Only use the niter constraining form of simple_iv when the exit
7135 2018-04-09 Tom de Vries <tom@codesourcery.com>
7138 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7139 (define_expand "*memory_barrier"): New define_expand.
7140 (define_insn "memory_barrier"): New insn.
7142 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7144 PR rtl-optimization/80463
7145 PR rtl-optimization/83972
7146 PR rtl-optimization/83480
7148 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7149 correct producer for the insn.
7150 (tidy_control_flow): Fixup seqnos in case of debug insns.
7152 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7154 PR rtl-optimization/83913
7156 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7157 different sched-times when merging exprs.
7159 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7161 PR rtl-optimization/83962
7163 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7164 tidy_fallthru_edge and tidy_control_flow.
7166 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7168 PR rtl-optimization/83530
7170 * sel-sched.c (force_next_insn): New global variable.
7171 (remove_insn_for_debug): When force_next_insn is true, also leave only
7172 next insn in the ready list.
7173 (sel_sched_region): When the region wasn't scheduled, make another pass
7174 over it with force_next_insn set to 1.
7176 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7178 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7180 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7181 for interrupt control.
7182 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7183 functions for interrupt control.
7184 * config/nds32/nds32-intrinsic.md: Likewise.
7185 * config/nds32/nds32_intrinsic.h: Likewise.
7186 * config/nds32/nds32.h (nds32_builtins): Likewise.
7188 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7190 * config/nds32/nds32.c (nds32_init_machine_status,
7191 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7192 strict_aligned_p field.
7193 (nds32_expand_to_rtl_hook): New function.
7194 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7195 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7197 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7198 Chung-Ju Wu <jasonwucj@gmail.com>
7200 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7201 * config/nds32/nds32-n7.md: New file.
7202 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7203 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7205 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7206 * config/nds32/nds32.md (pipeline_model): Add n7.
7207 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7208 * config/nds32/pipelines.md: Include n7 settings.
7210 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7211 Chung-Ju Wu <jasonwucj@gmail.com>
7213 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7214 * config/nds32/nds32-e8.md: New file.
7215 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7216 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7218 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7219 * config/nds32/nds32.md (pipeline_model): Add e8.
7220 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7221 * config/nds32/pipelines.md: Include e8 settings.
7223 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7224 Chung-Ju Wu <jasonwucj@gmail.com>
7226 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7227 * config/nds32/nds32-n8.md: New file.
7228 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7231 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7232 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7233 * config/nds32/nds32.md (pipeline_model): Add n8.
7234 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7235 * config/nds32/pipelines.md: Include n8 settings.
7237 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7238 Chung-Ju Wu <jasonwucj@gmail.com>
7240 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7241 * config/nds32/nds32-n9-2r1w.md: New file.
7242 * config/nds32/nds32-n9-3r2w.md: New file.
7243 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7244 nds32_register_ports): New or modify for cpu n9.
7245 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7247 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7248 * config/nds32/nds32-utils.c: New file.
7249 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7250 TARGET_MUL_SLOW): Define.
7251 * config/nds32/nds32.md (pipeline_model): New attribute.
7252 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7253 New options that support cpu n9.
7254 * config/nds32/pipelines.md: Include n9 settings.
7255 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7257 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7259 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7260 information if necessary.
7261 (output_cond_branch_compare_zero): Likewise.
7262 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7263 (nds32_target_alignment): Refine for alignment.
7264 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7265 (FUNCTION_BOUNDARY): Modify.
7266 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7268 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7270 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7272 * config/nds32/constants.md (unspec_volatile_element): Add values for
7273 TLB operation and data prefetch.
7274 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7275 functions for TLB operation and data prefetch.
7276 * config/nds32/nds32-intrinsic.md: Likewise.
7277 * config/nds32/nds32_intrinsic.h: Likewise.
7278 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7279 (nds32_print_operand): Likewise.
7280 * config/nds32/nds32.h (nds32_builtins): Likewise.
7282 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7283 Andrew Pinski <pinsika@gcc.gnu.org>
7286 * match.pd: Use constant_boolean_node of correct type instead of
7287 boolean_true_node or boolean_false_node for simplifying
7288 pointer comparisons to zero.
7290 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7292 PR tree-optimization/80021
7293 * tree.c (verify_type_variant): Make error call in verify_variant_match
7294 translatable and remove final full stop.
7296 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7298 * config/nds32/constants.md (unspec_volatile_element): Add
7299 UNSPEC_VOLATILE_EH_RETURN.
7300 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7301 nds32_output_stack_pop): Support dwarf exception handling process.
7302 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7303 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7304 exception handling process.
7305 (nds32_compute_stack_frame): Likewise.
7306 (nds32_return_addr_rtx): Likewise.
7307 (nds32_initial_elimination_offset): Likewise.
7308 (nds32_expand_prologue): Likewise.
7309 (nds32_expand_epilogue): Likewise.
7310 (nds32_dynamic_chain_address): New function.
7311 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7313 (DYNAMIC_CHAIN_ADDRESS): Define.
7314 (EH_RETURN_DATA_REGNO): Define.
7315 (EH_RETURN_STACKADJ_RTX): Define.
7316 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7317 patterns for dwarf exception handling.
7319 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7321 * config/nds32/nds32.h: Clean up obsolete macros.
7323 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7325 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7326 Add enum values for particular instructions.
7327 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7328 particular intrinsic functions.
7329 * config/nds32/nds32-intrinsic.md: Likewise.
7330 * config/nds32/nds32_intrinsic.h: Likewise.
7331 * config/nds32/nds32.h (nds32_builtins): Likewise.
7332 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7333 (btst, ave): New patterns for particular instructions.
7335 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7337 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7338 Add enum values for atomic load/store and memory sync.
7339 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7341 * config/nds32/nds32-intrinsic.md: Likewise.
7342 * config/nds32/nds32_intrinsic.h: Likewise.
7343 * config/nds32/nds32.h (nds32_builtins): Likewise.
7345 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7347 PR tree-optimization/85257
7348 * fold-const.c (native_encode_vector): If not all elts could fit
7349 and off is -1, return 0 rather than offset.
7350 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7351 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7352 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7353 adjust buffer in native_interpret_expr call.
7355 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7357 * config/nds32/constants.md (unspec_volatile_element): Add cache
7358 control enum values.
7359 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7360 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7361 * config/nds32/nds32.c (nds32_cctl_names): New.
7362 (nds32_print_operand): Handle cache control register names.
7363 * config/nds32/nds32.h (nds32_builtins): New enum values.
7364 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7366 * config/nds32/nds32.md (type): Add mmu.
7367 * config/nds32/pipelines.md (simple_insn): Add mmu.
7369 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7371 * config/nds32/nds32.md (type): Remove call.
7372 * config/nds32/pipelines.md (simple_insn): Likewise.
7374 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7376 * config/nds32/constants.md (unspec_volatile_element): Add
7377 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7378 UNSPEC_VOLATILE_FMFCFG.
7379 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7380 description for fmfcfg and fmfcsr.
7381 (bdesc_1arg): Add fmtcsr.
7382 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7383 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7384 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7385 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7386 unspec_fmfcfg): New patterns.
7387 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7388 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7389 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7390 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7391 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7392 __nds32__fmfcfg): Define.
7394 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7396 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7397 intrinsic register names.
7398 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7399 intrinsic register enum values and macros.
7401 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7403 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7404 for load/store addressing form.
7405 (nds32_print_operand_address): Likewise.
7407 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7410 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7411 based on LABEL_REF. Remove useless assertion.
7412 (pic_address_needs_scratch): Fix formatting.
7413 (sparc_legitimize_pic_address): Minor tweaks.
7414 (sparc_delegitimize_address): Adjust assertion accordingly.
7415 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7416 into symbolic_operand.
7417 (movsi_high_pic_label_ref): Likewise.
7418 (movsi_lo_sum_pic_label_ref): Likewise.
7419 (movdi_pic_label_ref): Likewise.
7420 (movdi_high_pic_label_ref): Likewise.
7421 (movdi_lo_sum_pic_label_ref): Likewise.
7423 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7425 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7426 custom LIB_SPEC setup.
7428 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7429 Kito Cheng <kito.cheng@gmail.com>
7431 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7432 * config/riscv/freebsd.h: New.
7434 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7436 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7437 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7440 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7441 Kito Cheng <kito.cheng@gmail.com>
7443 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7444 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7445 * config/nds32/nds32-protos.h (nds32_output_call,
7446 nds32_output_return): Declare.
7447 * config/nds32/nds32.md: Refine all the call and return patterns.
7449 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7452 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7453 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7455 PR rtl-optimization/84872
7456 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7457 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7460 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7462 * expr.c (copy_blkmode_to_reg): Revert 254862.
7463 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7465 2018-04-06 Richard Biener <rguenther@suse.de>
7468 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7469 after seeing a component reference with an adjacent field. Treat
7470 refs to arrays at struct end of external decls similar to
7471 refs to unconstrained commons.
7473 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7476 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7477 look through SAVE_EXPRs with non-side-effects argument. Adjust
7479 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7482 2018-04-06 Richard Biener <rguenther@suse.de>
7485 * alias.c (find_base_term): New wrapper around find_base_term
7486 unwinding CSELIB_VAL_PTR changes.
7487 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7490 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7492 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7494 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7495 constant definitions.
7496 ("nop"): lr 0,0 -> nopr r0
7497 ("nop_lr0", "nop_lr1"): New insn definitions.
7499 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7501 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7502 NDS32_V3PUSH_AVAILABLE_P macro.
7504 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7505 Chung-Ju Wu <jasonwucj@gmail.com>
7507 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7508 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7509 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7510 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7511 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7512 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7513 * config/nds32/constraints.md: New constraints and checking for hard
7514 float configuration.
7515 * config/nds32/iterators.md: New mode iterator and attribute for hard
7516 float configuration.
7517 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7519 * config/nds32/nds32-fpu.md: New file.
7520 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7521 deal with hard float code generation.
7522 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7524 (abi_type, float_reg_number): New enum type.
7525 * config/nds32/nds32-predicates.c: New predicates for hard float.
7526 * config/nds32/nds32-protos.h: Declare functions for hard float.
7527 * config/nds32/nds32.c: Implementation for hard float configuration.
7528 * config/nds32/nds32.h: Definitions for hard float configuration.
7529 * config/nds32/nds32.md: Include hard float machine description and
7530 modify patterns for hard float configuration.
7531 * config/nds32/nds32.opt: New options for hard float configuration.
7532 * config/nds32/predicates.md: New predicates for hard float
7535 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7537 * common/config/nds32/nds32-common.c
7538 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7540 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7543 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7544 CONSTRUCTOR_ELT (ctor, ...)->value.
7546 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7549 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7551 2018-04-05 Tom de Vries <tom@codesourcery.com>
7554 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7557 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7558 Kito Cheng <kito.cheng@gmail.com>
7560 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7561 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7562 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7563 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7565 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7566 Kito Cheng <kito.cheng@gmail.com>
7568 * config/nds32/constraints.md (Ufe): New memory constraint.
7569 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7570 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7571 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7573 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7574 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7576 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7578 * config/nds32/nds32.md: Use optimize_size in the condition for
7579 alu-shift instructions.
7581 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7583 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7585 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7587 * config/nds32/nds32.md (negsi2): Refine pattern.
7589 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7590 Chung-Ju Wu <jasonwucj@gmail.com>
7592 * config/nds32/iterators.md (shift_rotate): New code iterator.
7593 (shift): New code attribute.
7594 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7595 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7596 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7597 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7598 bit-wise operations.
7599 (andsi3, *andsi3): Ditto.
7600 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7601 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7602 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7603 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7604 nds32_ior_operand, nds32_xor_operand): New predicates.
7606 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7608 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7609 (addsi3, subsi3): ... this.
7611 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7613 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7615 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7617 * config/nds32/nds32.md: Adjust indention.
7619 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7621 * config/nds32/nds32.md (feature): New attribute.
7623 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7625 * config/nds32/nds32.md (subtype): New attribute.
7627 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7630 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7631 expansion to perform a bitwise AND of the argument followed by a
7632 boolean negation of the result.
7634 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7636 PR rtl-optimization/84878
7637 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7638 the basic block. Assert the use reference is not artificial and that
7639 it has an associated insn.
7641 2018-04-04 Michael Matz <matz@suse.de>
7643 * builtins.c (compute_objsize): Pass correct operand
7644 to array_at_struct_end_p.
7646 2018-04-04 Richard Biener <rguenther@suse.de>
7649 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7650 from contexts for DINFO_LEVEL_TERSE and below.
7652 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7654 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7655 resiter_operand condition.
7656 * config/nds32/nds32.md (*move<mode>): Ditto.
7658 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7659 Monk Chiang <sh.chiang04@gmail.com>
7661 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7663 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7665 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7667 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7668 Kito Cheng <kito.cheng@gmail.com>
7670 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7671 nds32_cond_code_str, output_cond_branch,
7672 output_cond_branch_compare_zero, nds32_expand_cbranch,
7673 nds32_expand_cstore, nds32_expand_movcc,
7674 nds32_output_cbranchsi4_equality_zero,
7675 nds32_output_cbranchsi4_equality_reg,
7676 nds32_output_cbranchsi4_equality_reg_or_const_int,
7677 nds32_output_cbranchsi4_greater_less_zero: New functions.
7678 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7679 nds32_expand_cstore, nds32_expand_movcc,
7680 nds32_output_cbranchsi4_equality_zero,
7681 nds32_output_cbranchsi4_equality_reg,
7682 nds32_output_cbranchsi4_equality_reg_or_const_int,
7683 nds32_output_cbranchsi4_greater_less_zero): Declare.
7684 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7685 nds32_rimm11s_operand): New predicates.
7686 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7687 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7690 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7692 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7693 * config/nds32/nds32.md: Ditto.
7694 * config/nds32/pipelines.md: Ditto.
7696 2018-04-04 Richard Biener <rguenther@suse.de>
7698 PR tree-optimization/85168
7699 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7700 propagating abnormals.
7702 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7704 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7706 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7707 Kito Cheng <kito.cheng@gmail.com>
7709 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7710 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7711 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7712 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7713 * config/nds32/nds32.md (sibcall_internal): New.
7714 (sibcall_register): Remove.
7715 (sibcall_immediate): Remove.
7716 (sibcall_value_internal): New.
7717 (sibcall_value_register): Remove.
7718 (sibcall_value_immediate): Remove.
7719 * config/nds32/predicates.md (nds32_general_register_operand): New.
7720 (nds32_call_address_operand): New.
7722 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7724 PR rtl-optimization/85167
7725 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7726 bb_defs if *split_p, instead preinitialize it to NULL.
7728 PR tree-optimization/85156
7729 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7730 evaluating the argument multiple times.
7732 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7734 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7736 (_mm_cvtpd_ps): Likewise.
7737 (_mm_cvttpd_epi32): Likewise.
7738 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7739 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7740 vector, pixel, and bool following altivec.h include.
7742 2018-04-03 Martin Sebor <msebor@redhat.com>
7744 * doc/extend.texi (Common Function Attributes): Clarify.
7745 (const attribute): Likewise.
7746 (pure attribute): Likewise.
7748 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7751 * config/i386/i386.c (ix86_expand_vector_set): Use
7752 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7754 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7756 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7757 instructions when changing rounding bits to preserve precision bits
7758 in the x87 control word.
7760 2018-04-03 Martin Liska <mliska@suse.cz>
7762 PR tree-optimization/82491
7763 * rtl.h (strip_offset_and_add): Replace += suboffset with
7764 poly_uint64 () + suboffset.
7766 2018-03-29 Martin Liska <mliska@suse.cz>
7767 Martin Jambor <mjambor@suse.cz>
7770 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7771 param_type is not an integral or pointer type.
7773 2018-04-03 Richard Biener <rguenther@suse.de>
7775 * sese.h (recompute_all_dominators): Remove.
7777 2018-04-02 Martin Sebor <msebor@redhat.com>
7779 * doc/invoke.texi (-Wrestrict): Fix typos.
7781 2018-04-02 Jim Wilson <jimw@sifive.com>
7783 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7784 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7785 (<optab>di3, <optab>si3_extend): Likewise.
7786 (<optab>si3_mask, <optab>si3_mask_1): New.
7787 (<optab>di3_mask, <optab>di3_mask_1): New.
7788 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7789 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7790 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7792 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7794 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7797 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7799 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7800 (nds32_canonicalize_comparison): New function.
7802 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7803 Kito Cheng <kito.cheng@gmail.com>
7804 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7806 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7807 * config/nds32/constants.md (unspec_volatile_element): Add
7808 UNSPEC_VOLATILE_RELAX_GROUP.
7809 * config/nds32/nds32-relax-opt.c: New file.
7810 * config/nds32/nds32-predicates.c
7811 (nds32_symbol_load_store_p): New function.
7812 * config/nds32/nds32-protos.h
7813 (nds32_symbol_load_store_p): Declare function.
7814 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7815 * config/nds32/nds32.c
7816 (nds32_register_pass): New function to register pass.
7817 (nds32_register_passes): New function to register passes.
7818 * config/nds32/nds32.md (relax_group): New pattern.
7819 * config/nds32/nds32.opt (mrelax-hint): New option.
7820 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7822 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7824 * config/nds32/t-nds32: Modify files dependency.
7826 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7828 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7829 (PROFILE_HOOK): Define its implementation.
7831 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7833 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7834 type and 32-bit size.
7836 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7839 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7840 (V_128_256): New mode iterator.
7841 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7842 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7843 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7845 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7846 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7848 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7851 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7852 NaN inputs correctly.
7854 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7857 * config/rs6000/vsx.md (??r): New mode attribute.
7858 (*vsx_mov<mode>_64bit): Use it.
7859 (*vsx_mov<mode>_32bit): Likewise.
7861 2018-03-30 Martin Sebor <msebor@redhat.com>
7863 PR tree-optimization/84818
7864 * builtins.c (check_access): Use warning_n.
7866 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7869 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7871 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7874 2018-03-30 Julia Koval <julia.koval@intel.com>
7877 * x86-tune.def (movx, partial_reg_dependency): Enable for
7880 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7883 * lra-constraints.c (process_alt_operands): Move setting
7884 this_alternative_matches below.
7886 2018-03-29 Martin Liska <mliska@suse.cz>
7889 * doc/invoke.texi: Document how LTO works with debug info.
7890 Describe auto-load support of binutils. Mention 'x86-64'
7891 as valid option value of -march option.
7893 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7895 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7898 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7899 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7900 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7903 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7906 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7907 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7908 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7909 for RS6000_BTM_POWERPC64.
7910 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7911 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7912 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7917 2018-03-28 Carl Love <cel@us.ibm.com>
7920 2017-09-27 Carl Love <cel@us.ibm.com>
7922 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7923 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7924 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7927 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7929 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7930 instead of __vector bool.
7931 (_mm_max_pu8): Likewise.
7932 (_mm_min_pi16): Likewise.
7934 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7937 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7938 (DIVWEUO): Likewise.
7940 (DIVDEUO): Likewise.
7941 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7942 DIVWEUO and DIVDEUO.
7943 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7944 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7945 (div_extend): Likewise.
7946 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7948 (__builtin_divweuo): Likewise.
7949 (__builtin_divdeo): Likewise.
7950 (__builtin_divdeuo): Likewise.
7952 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7955 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7956 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7958 PR tree-optimization/82004
7959 * gimple-match-head.c (optimize_pow_to_exp): New function.
7960 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7961 Don't fold to exp if optimize_pow_to_exp is false.
7963 2018-03-28 Martin Liska <mliska@suse.cz>
7966 * calls.c (initialize_argument_information): Fix trailing space.
7967 * common.opt: Fix typo and provide better explanation for
7968 -fsanitize-coverage option.
7969 * config/i386/i386.opt: Fix typo.
7971 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7972 Martin Liska <mliska@suse.cz>
7975 * gimplify.c (asan_poison_variable): Don't do the check for
7976 gimplify_omp_ctxp here.
7977 (gimplify_decl_expr): Do it here.
7978 (gimplify_target_expr): Likewise.
7980 2018-03-28 Martin Liska <mliska@suse.cz>
7983 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7984 chkp_type_bounds_count if MPX is not enabled.
7986 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7988 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7990 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7993 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7994 function to create the function decl for complex long double
7995 multiply and divide for -mabi=ieeelongdouble.
7996 (init_float128_ieee): Call it.
7998 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
8001 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
8002 -fcf-protection=branch -mibt.
8003 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
8005 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8008 * config/arm/arm.c (arm_valid_symbolic_address): Handle
8009 arm_word_relocations.
8011 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
8014 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
8015 extern array declarations.
8017 2018-03-27 Richard Biener <rguenther@suse.de>
8020 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8021 explicit single_use checks.
8023 2018-03-27 Richard Biener <rguenther@suse.de>
8025 PR tree-optimization/85082
8026 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8029 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8031 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8032 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8033 Turn on fasynchronous-unwind-tables and funwind-tables.
8035 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8038 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8039 (*bmi_blsr_<mode>_ccz): Ditto.
8041 2018-03-26 Tom de Vries <tom@codesourcery.com>
8043 PR tree-optimization/85063
8044 * omp-general.c (offloading_function_p): New function. Factor out
8046 * omp-offload.c (pass_omp_target_link::gate): ... here.
8047 * omp-general.h (offloading_function_p): Declare.
8048 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8049 with attribute omp declare target for offloading functions.
8051 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8053 PR tree-optimization/84005
8054 * tree-data-ref.h (get_base_for_alignment): Declare.
8055 * tree-data-ref.c (get_base_for_alignment_1): New function.
8056 (get_base_for_alignment): Likewise.
8057 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8058 get_base_for_alignment to find a suitable base object, instead
8059 of always using drb->base_address.
8061 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8064 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8065 known size by default.
8067 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8070 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8071 and non BLKmode operands.
8073 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8076 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8077 Clean up attributes.
8079 2018-03-23 Richard Biener <rguenther@suse.de>
8082 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8083 we are going to emit early debug for LTO.
8085 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8088 * function.c (match_asm_constraints_1): Don't optimize if input
8089 doesn't satisfy general_operand predicate for output's mode.
8092 * alias.c (write_dependence_p): Don't require for x_canonicalized
8093 non-VOIDmode if x has VOIDmode.
8096 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8097 just don't try to optimize it rather than assert it never happens.
8099 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8101 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8102 macro expansions for definition of ST_INTERNAL_<mode> and
8103 LD_INTERNAL_<mode> builtins.
8104 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8106 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8108 (altivec_expand_st_builtin): Likewise.
8109 (altivec_expand_builtin): Remove calls to deleted functions.
8110 (rs6000_address_for_altivec): Delete this function.
8111 * config/rs6000/vector.md: Remove expands for
8112 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8114 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8117 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8118 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8119 re-computing once computed.
8120 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8121 (arm_init_machine_status): Initialize
8122 machine->static_chain_stack_bytes.
8124 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8127 * doc/extend.texi: Add four new prototypes for vec_ld.
8128 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8129 definitions for more logical presentation.
8130 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8131 entries for V1TI variants of __builtin_altivec_ld builtin.
8132 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8133 handling of V1TI variant of LVX icode pattern.
8134 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8135 (rs6000_gimple_fold_builtin): Likewise.
8136 (altivec_init_builtins): Add code to define
8137 __builtin_altivec_lvx_v1ti function.
8139 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8142 * function.c (match_asm_constraints_1): Don't do the optimization
8143 if input isn't a REG, SUBREG, MEM or constant.
8145 2018-03-22 Tom de Vries <tom@codesourcery.com>
8147 PR tree-optimization/84956
8148 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8149 bb_has_abnormal_pred.
8151 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8154 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8155 DECL_INITIAL (decl) to decl at the end.
8156 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8159 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8161 * doc/extend.texi (__builtin_tgmath): Document when complex
8162 integer types are treated as _Complex _Float64.
8164 2018-03-21 Tom de Vries <tom@codesourcery.com>
8166 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8168 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8170 PR tree-optimization/84960
8171 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8172 if it is ENTRY block, move them into single succ of ENTRY in that case.
8174 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8176 PR tree-optimization/84811
8177 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8178 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8179 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8180 (known_subrange_p): Update accordingly. Cast each value involved
8181 in the size comparison, rather than casting the result of the
8184 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8186 PR tree-optimization/84982
8187 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8188 by flipping the least significant bit rather than all bits from
8189 bitpos to bitpos + bitsize - 1.
8191 2018-03-21 Nathan Sidwell <nathan@acm.org>
8193 * doc/extend.texi (Deprecated Features): Remove mention of
8194 long-deleted deprecations.
8196 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8199 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8200 * configure: Regenerate.
8202 2018-03-21 Tom de Vries <tom@codesourcery.com>
8204 PR tree-optimization/83126
8205 * tree-parloops.c (num_phis): New function.
8206 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8208 2018-03-21 Nathan Sidwell <nathan@acm.org>
8210 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8211 mention anon-struct/union members and trailing attributes.
8213 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8215 PR tree-optimization/84969
8216 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8217 builtin memset partitions if they set different rhs values.
8219 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8221 PR rtl-optimization/84989
8222 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8223 VEC_DUPLICATE with scalar result mode.
8225 2018-03-21 Martin Liska <mliska@suse.cz>
8228 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8229 not intended return statement.
8231 2018-03-21 Martin Liska <mliska@suse.cz>
8234 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8235 (chkp_find_bound_slots_1): Limit number of iterations.
8237 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8240 * Minor grammar fixes for x86 options.
8242 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8245 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8246 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8248 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8251 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8252 (altivec_lvx_<mode>_1op): Likewise.
8253 (altivec_stvx_<mode>_2op): Likewise.
8254 (altivec_stvx_<mode>_1op): Likewise.
8255 (altivec_lvx_<VM2:mode>): New define_expand.
8256 (altivec_stvx_<VM2:mode>): Likewise.
8257 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8258 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8259 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8260 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8261 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8262 (rs6000_gen_lvx): Likewise.
8263 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8264 (altivec_expand_stv_builtin): Likewise.
8265 (altivec_expand_builtin): Likewise.
8266 * config/rs6000/vector.md: Likewise.
8268 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8271 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8274 2018-03-20 Richard Biener <rguenther@suse.de>
8277 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8278 sign-conversions as zero, fall back to standard scalar_stmt
8281 2018-03-20 Martin Liska <mliska@suse.cz>
8284 * predict.c (rebuild_frequencies): Handle case when we have
8285 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8287 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8290 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8291 flag_section_anchors.
8292 * varasm.c (use_blocks_for_decl_p): Remove hack for
8293 dw2_force_const_mem.
8296 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8298 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8299 be created, use lowpart_subreg of operands[0] rather than operands[0]
8301 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8302 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8303 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8304 and n constraint instead of aarch64_shift_imm_di and Usd.
8305 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8306 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8308 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8311 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8312 to favor GPR over NEON registers.
8313 (<shift>di3_neon): Likewise.
8315 2018-03-20 Tom de Vries <tom@codesourcery.com>
8318 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8319 (nvptx_process_pars): Emit bar.sync asap and alap.
8321 2018-03-20 Tom de Vries <tom@codesourcery.com>
8324 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8325 seen_label if seen_label is already set.
8327 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8330 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8331 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8332 Use 1U instead of 1. Formatting fixes.
8335 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8336 instead of TREE_TYPE (s1) for the return value.
8338 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8340 PR tree-optimization/84946
8341 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8342 bitsize + bitsize in poly_uint64 rather than poly_int64.
8345 * dwarf2asm.c: Include fold-const.c.
8346 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8347 of decl rather than decl itself.
8349 PR rtl-optimization/84643
8350 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8352 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8355 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8356 calling assemble_variable.
8358 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8361 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8362 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8364 2018-03-19 Jim Wilson <jimw@sifive.com>
8367 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8368 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8369 (riscv_first_stack_step): Likewise.
8370 (riscv_option_override): Use STACK_BOUNDARY instead of
8372 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8374 (BIGGEST_ALIGNMENT): Set to 128.
8375 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8376 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8379 2018-03-19 Richard Biener <rguenther@suse.de>
8381 PR tree-optimization/84933
8382 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8383 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8385 2018-03-19 Richard Biener <rguenther@suse.de>
8387 PR tree-optimization/84859
8388 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8389 (cond_if_else_store_replacement): Perform sinking operation on
8390 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8391 Generalize what a BB with a single eligible store is.
8393 2018-03-19 Richard Biener <rguenther@suse.de>
8395 PR tree-optimization/84929
8396 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8397 chrec_is_positive against non-chrec arg.
8399 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8402 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8404 2018-03-18 Martin Liska <mliska@suse.cz>
8406 PR rtl-optimization/84635
8407 * regrename.c (build_def_use): Use matches_mode only when
8410 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8412 PR tree-optimization/84913
8413 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8414 vectorize chains of COND_EXPRs.
8416 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8418 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8420 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8422 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8424 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8426 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8428 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8429 Kito Cheng <kito.cheng@gmail.com>
8431 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8432 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8433 (nds32_adjust_reg_alloc_order): New function.
8434 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8436 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8438 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8439 nds32_print_operand, nds32_print_operand_address): Use
8440 HOST_WIDE_INT_PRINT_DEC instead.
8442 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8444 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8446 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8449 * config/i386/i386.c (initial_ix86_tune_features,
8450 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8452 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8453 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8454 rather than 1u << ix86_tune. Formatting fix.
8455 (ix86_option_override_internal): Change ix86_arch_mask from
8456 unsigned int to unsigned HOST_WIDE_INT, initialize to
8457 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8458 (ix86_function_specific_restore): Likewise.
8460 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8463 * postreload.c (reload_combine_recognize_pattern): Perform
8464 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8465 truncate_int_for_mode the result for the destination's mode.
8468 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8469 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8471 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8474 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8477 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8478 regno_allocno_class_array and sorted_pseudos.
8479 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8480 insns where regno is used.
8482 2018-03-16 Martin Liska <mliska@suse.cz>
8485 * multiple_target.c (create_dispatcher_calls): Redirect
8486 reference in the symbol table.
8488 2018-03-16 Martin Liska <mliska@suse.cz>
8491 * multiple_target.c (create_dispatcher_calls): Redirect also
8494 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8498 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8499 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8500 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8502 2018-03-16 Julia Koval <julia.koval@intel.com>
8504 * doc/invoke.texi (Skylake Server): Add CLWB.
8505 Cannonlake): Remove CLWB.
8507 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8509 PR tree-optimization/84841
8510 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8512 (FLOAT_ONE_CONST_TYPE): Define.
8513 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8514 (sort_by_operand_rank): Put entries with higher constant_type last
8515 rather than first to match comments.
8517 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8519 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8522 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8525 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8526 check and use address of target always.
8528 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8531 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8532 (indirect_thunk_bnd_needed): Likewise.
8533 (indirect_thunks_used): Likewise.
8534 (indirect_thunks_bnd_used): Likewise.
8535 (indirect_return_needed): New.
8536 (indirect_return_bnd_needed): Likewise.
8537 (output_indirect_thunk_function): Add a bool argument for
8539 (output_indirect_thunk_function): Don't generate alias for
8540 function return thunk.
8541 (ix86_code_end): Call output_indirect_thunk_function to generate
8542 function return thunks.
8543 (ix86_output_function_return): Set indirect_return_bnd_needed
8544 and indirect_return_needed instead of indirect_thunk_bnd_needed
8545 and indirect_thunk_needed.
8547 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8549 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8550 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8551 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8553 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8554 Paul Hua <paul.hua.gm@gmail.com>
8557 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8559 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8561 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8562 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8565 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8568 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8569 instead of GET_MODE_SIZE when comparing Units.
8571 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8574 * varasm.c (hash_section): Return an unchangeble hash value
8575 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8576 Return !aarch64_can_use_per_function_literal_pools_p ().
8578 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8581 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8582 rather than address of cmode as last argument to prepare_cmp_insn.
8584 2018-03-15 Julia Koval <julia.koval@intel.com>
8586 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8587 F_AVX512VNNI, F_AVX512BITALG): New.
8589 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8592 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8593 insn for floating-point loads and stores.
8595 2018-03-14 Carl Love <cel@us.ibm.com>
8597 * config/rs6000/rs6000-c.c: Add macro definitions for
8598 ALTIVEC_BUILTIN_VEC_PERMXOR.
8599 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8600 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8601 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8602 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8604 * config/doc/extend.texi: Add prototypes for vec_permxor.
8606 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8609 * diagnostic-show-locus.c (class layout_point): Convert m_line
8610 from int to linenum_type.
8611 (line_span::comparator): Use linenum "compare" function when
8612 comparing line numbers.
8613 (test_line_span): New function.
8614 (layout_range::contains_point): Convert param "row" from int to
8616 (layout_range::intersects_line_p): Likewise.
8617 (layout::will_show_line_p): Likewise.
8618 (layout::print_source_line): Likewise.
8619 (layout::should_print_annotation_line_p): Likewise.
8620 (layout::print_annotation_line): Likewise.
8621 (layout::print_leading_fixits): Likewise.
8622 (layout::annotation_line_showed_range_p): Likewise.
8623 (struct line_corrections): Likewise for field m_row.
8624 (line_corrections::line_corrections): Likewise for param "row".
8625 (layout::print_trailing_fixits): Likewise.
8626 (layout::get_state_at_point): Likewise.
8627 (layout::get_x_bound_for_row): Likewise.
8628 (layout::print_line): Likewise.
8629 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8630 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8631 * input.c (selftest::test_linenum_comparisons): New function.
8632 (selftest::input_c_tests): Call it.
8633 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8634 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8635 * selftest.h (ASSERT_GT): New macro.
8636 (ASSERT_GT_AT): New macro.
8637 (ASSERT_LT): New macro.
8638 (ASSERT_LT_AT): New macro.
8640 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8642 PR rtl-optimization/84780
8643 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8645 2018-03-14 Martin Liska <mliska@suse.cz>
8647 * tree.c (record_node_allocation_statistics): Use
8648 get_stats_node_kind.
8649 (get_stats_node_kind): New function extracted from
8650 record_node_allocation_statistics.
8651 (free_node): Use get_stats_node_kind.
8653 2018-03-14 Richard Biener <rguenther@suse.de>
8655 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8656 that the value-set of ANTIC_IN doesn't grow.
8659 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8661 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8662 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8664 2018-03-14 Julia Koval <julia.koval@intel.com>
8666 * config.gcc (icelake-client, icelake-server): New.
8668 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8669 (initial_ix86_arch_features): Ditto.
8670 (PTA_SKYLAKE): Add SGX.
8671 (PTA_ICELAKE): Remove.
8672 (PTA_ICELAKE_CLIENT): New.
8673 (PTA_ICELAKE_SERVER): New.
8674 (ix86_option_override_internal): Split up icelake on icelake client and
8676 (get_builtin_code_for_version): Ditto.
8677 (fold_builtin_cpu): Ditto.
8678 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8679 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8680 * config/i386/i386.h (processor_type): Ditto.
8681 * doc/invoke.texi: Ditto.
8683 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8686 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8687 INTEGER_CST offset, add it together with bitpos / 8 and
8688 sign extend based on POINTER_SIZE.
8692 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8695 * config/i386/constraints.md (Yc): New register constraint.
8696 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8697 Use Yc constraint for alternative 2 of operand 0. Remove
8698 preferred_for_speed attribute.
8700 2018-03-14 Richard Biener <rguenther@suse.de>
8702 PR tree-optimization/84830
8703 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8704 with the old one to avoid oscillations.
8706 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8709 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8711 (assign_by_spills): Return a flag of reload assignment failure.
8712 Do not process the reload assignment failures. Do not spill other
8713 reload pseudos if they has the same reg class. Update n if
8715 (lra_assign): Add a return arg. Set up from the result of
8716 assign_by_spills call.
8717 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8718 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8719 usage_insns if it is not NULL.
8720 (spill_hard_reg_in_range): New function.
8721 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8722 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8723 function prototypes.
8724 (lra_assign): Change prototype.
8725 * lra.c (lra): Add code to deal with fails by splitting hard reg
8728 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8730 * config/riscv/riscv.opt (mrelax): New option.
8731 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8732 "norelax" when riscv_mrelax is disabled.
8733 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8735 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8738 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8739 reassociation for int modes.
8741 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8743 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8744 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8746 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8747 * config/aarch64/aarch64-sve.md
8748 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8749 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8750 (*extend<mode><Vwide>2): Rename to...
8751 (aarch64_sve_extend<mode><Vwide>2): ...this.
8752 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8753 renaming the old pattern to...
8754 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8756 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8757 define_expand, renaming the old pattern to...
8758 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8759 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8760 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8761 account when deciding which SVE instruction the optab should use.
8762 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8764 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8766 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8767 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8768 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8769 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8770 (tlsdesc_small_<mode>): Turn a define_expand and use
8771 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8772 (tlsdesc_small_advsimd_<mode>): ...this.
8773 (tlsdesc_small_sve_<mode>): New pattern.
8775 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8777 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8778 (UNSPEC_UMUL_HIGHPART): New constants.
8779 (MUL_HIGHPART): New int iteraor.
8780 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8781 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8783 (*<su>mul<mode>3_highpart): New define_insn.
8785 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8788 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8791 2018-03-13 Martin Liska <mliska@suse.cz>
8794 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8796 (sem_item_optimizer::~sem_item_optimizer): Release it.
8797 (sem_item_optimizer::merge_classes): Register variable aliases.
8798 (sem_item_optimizer::fixup_pt_set): New function.
8799 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8800 * ipa-icf.h: Declare new variables and functions.
8802 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8805 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8806 integer_pow2p@2 and test integer_pow2p in condition.
8807 (A < 0 ? C : 0): Similarly for @1.
8810 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8811 characters starting at p contain '\0' character, don't look beyond
8815 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8816 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8819 * reg-stack.c (change_stack): Change update_end var from int to
8820 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8821 also call set_block_for_insn on the newly added insns and rescan.
8824 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8825 on the last operand.
8828 * tree.c (stabilize_reference_1): Return save_expr (e) for
8829 STATEMENT_LIST even if it doesn't have side-effects.
8831 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8833 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8835 2018-03-12 Renlin Li <renlin.li@arm.com>
8837 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8838 aarch64_output_scalar_simd_mov_immediate.
8840 2018-03-12 Martin Sebor <msebor@redhat.com>
8842 PR tree-optimization/83456
8843 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8844 for perfectly overlapping calls to memcpy.
8845 (gimple_fold_builtin_memory_chk): Same.
8846 (gimple_fold_builtin_strcpy): Handle no-warning.
8847 (gimple_fold_builtin_stxcpy_chk): Same.
8848 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8850 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8852 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8853 parameter. Use it for SFmode.
8854 (rs6000_function_arg_advance_1): Adjust.
8855 (rs6000_function_arg): Adjust.
8856 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8858 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8860 PR rtl-optimization/84169
8861 PR rtl-optimization/84780
8862 * combine.c (can_combine_p): Check for a 2-insn combination whether
8863 the destination register is used between the two insns, too.
8865 2018-03-12 Richard Biener <rguenther@suse.de>
8867 PR tree-optimization/84803
8868 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8869 for refs DR analysis didn't process.
8871 2018-03-12 Richard Biener <rguenther@suse.de>
8873 PR tree-optimization/84777
8874 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8875 force-vectorize loops ignore whether we are optimizing for size.
8877 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8879 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8880 (TARGET_MD_ASM_ADJUST): Define.
8882 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8883 Kito Cheng <kito.cheng@gmail.com>
8884 Chung-Ju Wu <jasonwucj@gmail.com>
8886 * config/nds32/nds32.c (nds32_compute_stack_frame,
8887 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8888 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8889 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8890 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8891 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8892 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8893 * config/nds32/nds32.md (prologue, epilogue): Use macro
8894 NDS32_V3PUSH_AVAILABLE_P to do checking.
8896 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8899 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8900 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8901 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8902 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8903 addition of most attributes on !orig_type_die or the attribute not
8904 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8906 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8907 Chung-Ju Wu <jasonwucj@gmail.com>
8909 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8911 * config/nds32/nds32.opt (mvh): New option.
8913 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8914 Chung-Ju Wu <jasonwucj@gmail.com>
8916 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8918 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8919 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8922 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8923 Chung-Ju Wu <jasonwucj@gmail.com>
8925 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8927 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8928 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8930 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8931 Kito Cheng <kito.cheng@gmail.com>
8932 Chung-Ju Wu <jasonwucj@gmail.com>
8934 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8935 UNSPEC_FFMISM and UNSPEC_FLMISM.
8936 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8937 for ffb, ffmism and flmism.
8938 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8939 (unspec_ffmism): Ditto.
8940 (unspec_flmism): Ditto.
8941 (nds32_expand_builtin_impl): Check if string extension is available.
8942 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8943 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8945 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8948 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8951 * lra-assigns.c (assign_by_spills): Return a flag of reload
8952 assignment failure. Do not process the reload assignment
8953 failures. Do not spill other reload pseudos if they has the same
8955 (lra_assign): Add a return arg. Set up from the result of
8956 assign_by_spills call.
8957 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8958 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8959 usage_insns if it is not NULL.
8960 (spill_hard_reg_in_range): New function.
8961 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8962 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8963 function prototypes.
8964 (lra_assign): Change prototype.
8965 * lra.c (lra): Add code to deal with fails by splitting hard reg
8968 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8971 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8973 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8976 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8977 (dw_val_node): Add val_symbolic_view.
8978 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8979 (symview_upper_bound): New.
8980 (new_line_info_table): Initialize symviews_since_reset.
8981 (dwarf2out_source_line): Count symviews_since_reset and set
8982 symview_upper_bound.
8983 (dw_val_equal_p): Handle symview.
8984 (add_AT_symview): New.
8985 (print_dw_val): Handle symview.
8986 (attr_checksum, attr_checksum_ordered): Likewise.
8987 (same_dw_val_p, size_of_die): Likewise.
8988 (value_format, output_die): Likewise.
8989 (add_high_low_attributes): Use add_AT_symview for entry_view.
8990 (dwarf2out_finish): Reset symview_upper_bound, clear
8993 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8996 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8997 Add strict argument and use it.
8998 (rs6000_split_multireg_move): Update for new strict argument.
8999 (mem_operand_gpr): Disallow all non-offsettable addresses.
9000 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
9002 2018-03-09 Jakub Jelinek <jakub@redhat.com>
9005 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
9006 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
9007 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
9010 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
9011 decl, use remap_type if we want to use the type.
9013 2018-03-09 Martin Sebor <msebor@redhat.com>
9015 PR tree-optimization/84526
9016 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9018 (builtin_access::generic_overlap): Be prepared to handle non-array
9021 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9023 PR rtl-optimization/84682
9024 * lra-constraints.c (process_address_1): Check is_address flag
9025 for address constraints.
9026 (process_alt_operands): Likewise.
9027 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9028 preprocess_constraints.
9029 * recog.h (preprocess_constraints): Add oploc parameter.
9031 * recog.c (preprocess_constraints): Test address_operand for
9032 CT_ADDRESS constraints.
9034 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9037 * lra-assigns.c (assign_by_spills): Return a flag of reload
9038 assignment failure. Do not process the reload assignment
9039 failures. Do not spill other reload pseudos if they has the same
9041 (lra_assign): Add a return arg. Set up from the result of
9042 assign_by_spills call.
9043 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9044 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9045 usage_insns if it is not NULL.
9046 (spill_hard_reg_in_range): New function.
9047 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9048 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9049 function prototypes.
9050 (lra_assign): Change prototype.
9051 * lra.c (lra): Add code to deal with fails by splitting hard reg
9054 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9057 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9058 Accept complain bool parameter. Only emit errors if it is true.
9059 (arm_parse_cpu_option_name): Likewise.
9060 (arm_target_thumb_only): Adjust callers of the above.
9061 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9062 prototype to take a default true bool parameter.
9063 (arm_parse_arch_option_name): Likewise.
9065 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9066 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9070 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9071 * configure: Regenerate.
9072 * configure.ac ("linker --version-script option"): New.
9073 ("linker soname option"): New.
9075 2018-03-09 Richard Biener <rguenther@suse.de>
9077 PR tree-optimization/84775
9078 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9079 immediate uses of predicate stmts and mark them modified.
9082 PR tree-optimization/84178
9083 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9085 (version_loop_for_if_conversion): Delay update_ssa call.
9086 (tree_if_conversion): Delay update_ssa until after predicate
9089 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9092 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9093 when the function accesses prior frames.
9095 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9098 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9099 gen_llsym, otherwise call maybe_gen_llsym.
9102 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9103 has ',' character inside of it.
9105 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9108 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9109 as clobbering CC_REGNUM.
9111 2018-03-08 Richard Biener <rguenther@suse.de>
9114 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9115 (follow_copies_to_constant): Do not follow SSA names registered
9118 2018-03-08 Richard Biener <rguenther@suse.de>
9120 PR tree-optimization/84178
9121 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9123 (version_loop_for_if_conversion): Delay update_ssa call.
9124 (tree_if_conversion): Delay update_ssa until after predicate
9127 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9129 PR tree-optimization/84178
9130 * tree-if-conv.c (release_bb_predicate): Remove the
9131 the assertion that the stmts have NULL use_ops.
9132 Discard the statements, asserting that they haven't
9133 yet been added to a BB.
9135 2018-03-08 Richard Biener <rguenther@suse.de>
9137 PR tree-optimization/84746
9138 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9139 (phi_translate): Pass in destination ANTIC_OUT set.
9140 (phi_translate_1): Likewise. For a simplified result lookup
9141 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9142 (phi_translate_set): Adjust.
9143 (do_pre_regular_insertion): Likewise.
9144 (do_pre_partial_partial_insertion): Likewise.
9146 2018-03-08 Martin Liska <mliska@suse.cz>
9148 PR gcov-profile/84735
9149 * doc/gcov.texi: Document usage of profile files.
9150 * gcov-io.h: Document changes in the format.
9152 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9156 * dwarf2out.c (struct dw_line_info_table): Update comments for
9158 (FORCE_RESET_NEXT_VIEW): New.
9159 (FORCE_RESETTING_VIEW_P): New.
9160 (RESETTING_VIEW_P): Check for -1 too.
9161 (ZERO_VIEW_P): Likewise.
9162 (new_line_info_table): Force-reset next view.
9163 (dwarf2out_begin_function): Likewise.
9164 (dwarf2out_source_line): Simplify zero_view_p initialization.
9165 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9166 view directly. Omit view when omitting .loc at line 0.
9168 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9170 PR tree-optimization/84740
9171 * tree-switch-conversion.c (process_switch): Call build_constructors
9172 only if info.phi_count is non-zero.
9174 PR tree-optimization/84739
9175 * tree-tailcall.c (find_tail_calls): Check call arguments against
9176 DECL_ARGUMENTS (current_function_decl) rather than
9177 DECL_ARGUMENTS (func) when checking for tail recursion.
9179 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9181 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9182 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9183 Volker Reichelt's entry and add entries for people that perform
9184 GCC fuzzy testing and report numerous bugs.
9186 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9189 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9190 readonly data in sdata, if that is disabled.
9191 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9192 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9193 -mreadonly-in-sdata option.
9195 2018-03-07 Martin Sebor <msebor@redhat.com>
9197 PR tree-optimization/84468
9198 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9199 basic block when looking for nul assignment.
9201 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9204 * except.h (output_function_exception_table): Adjust prototype.
9205 * except.c (output_function_exception_table): Remove FNNAME parameter
9206 and add SECTION parameter. Ouput one part of the table at a time.
9207 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9208 the first part of the exception table and emit unwind directives.
9209 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9210 (i386_pe_seh_cold_init): Likewise.
9211 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9212 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9213 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9214 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9215 * config/i386/winnt.c: Include except.h.
9216 (struct seh_frame_state): Add reg_offset, after_prologue and
9217 in_cold_section fields.
9218 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9219 (i386_pe_seh_cold_init): New function.
9220 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9221 to seh->in_cold_section.
9222 (seh_emit_push): Record the offset of the push.
9223 (seh_emit_save): Record the offet of the save.
9224 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9225 Test seh->after_prologue to disregard the epilogue.
9226 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9227 (i386_pe_end_cold_function): New function.
9229 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9232 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9233 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9236 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9238 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9239 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9242 * multiple_target.c: Include tree-inline.h and intl.h.
9243 (expand_target_clones): Diagnose and fail if node->definition and
9244 !tree_versionable_function_p (node->decl).
9246 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9248 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9250 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9251 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9252 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9254 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9257 * combine.c (try_combine): Use reg_or_subregno instead of handling
9258 just paradoxical SUBREGs and REGs.
9260 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9262 * config/arc/arc.c (arc_finalize_pic): Remove function.
9263 (arc_must_save_register): We use single base PIC register, remove
9264 checks to save/restore the PIC register.
9265 (arc_expand_prologue): Likewise.
9266 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9268 (arc_verify_short): Likewise.
9269 (arc_attr_type): Likewise.
9270 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9271 (walk_stores): Likewise.
9272 (arc_address_cost): Make it static.
9273 (arc_verify_short): Likewise.
9274 (branch_dest): Likewise.
9275 (arc_attr_type): Likewise.
9276 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9277 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9278 (arc_final_prescan_insn): Remove inserting the nops due to
9279 hardware hazards. It is done in reorg step.
9280 (insn_length_variant_t): Remove.
9281 (insn_length_parameters_t): Likewise.
9282 (arc_insn_length_parameters): Likewise.
9283 (arc_get_insn_variants): Likewise.
9284 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9286 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9289 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9292 PR tree-optimization/84687
9293 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9295 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9297 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9299 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9300 Rename to ppc_speculation_barrier.
9301 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9302 __builtin_ppc_speculation_barrier.
9304 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9307 * combine.c (combine_simplify_rtx): Don't try to simplify if
9308 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9311 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9313 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9314 to 32 bytes when compiling for POWER9.
9316 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9319 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9320 regparm >= 3 with no arg reg available also for calls with
9321 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9324 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9326 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9328 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9331 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9333 2018-03-05 Richard Biener <rguenther@suse.de>
9335 PR tree-optimization/84486
9336 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9337 When inserting a __builtin_assume_aligned call set the LHS
9338 SSA name alignment info accordingly.
9340 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9342 PR tree-optimization/84114
9343 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9344 Avoid reassociation of FLOAT_MODE addition.
9346 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9348 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9349 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9350 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9351 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9352 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9353 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9354 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9356 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9357 __builtin_ia32_wbinvd): New builtins.
9358 (SPECIAL_ARGS2): New.
9359 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9360 (SPECIAL_ARGS2): New.
9361 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9362 (ix86_valid_target_attribute_inner_p): Ditto.
9363 (ix86_init_mmx_sse_builtins): Add special_args2.
9364 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9365 TARGET_WBNOINVD_P): New.
9366 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9367 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9368 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9369 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9370 * config/i386/pconfigintrin.h: New file.
9371 * config/i386/wbnoinvdintrin.h: Ditto.
9372 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9374 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9376 2018-03-05 Richard Biener <rguenther@suse.de>
9378 PR tree-optimization/84670
9379 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9381 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9382 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9383 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9384 doesn't grow if all successors (recursively) were visited at least
9387 2018-03-05 Richard Biener <rguenther@suse.de>
9389 PR tree-optimization/84650
9390 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9391 if executed in the loop pipeline.
9393 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9395 * doc/configfiles.texi (Configuration Files): Move info about
9396 conditionalizing $target-protos.h to...
9397 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9398 differs from $target-protos.h.
9400 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9401 Chung-Ju Wu <jasonwucj@gmail.com>
9403 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9404 * config/nds32/nds32-multiple.md (setmemsi): Define.
9405 * config/nds32/nds32-memory-manipulation.c
9406 (nds32_gen_dup_4_byte_to_word_value): New.
9407 (emit_setmem_word_loop): New.
9408 (emit_setmem_byte_loop): New.
9409 (nds32_expand_setmem_loop): New.
9410 (nds32_expand_setmem_loop_v3m): New.
9411 (nds32_expand_setmem_unroll): New.
9412 (nds32_expand_setmem): New.
9414 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9415 Chung-Ju Wu <jasonwucj@gmail.com>
9417 * config/nds32/nds32-memory-manipulation.c
9418 (nds32_emit_load_store): New.
9419 (nds32_emit_post_inc_load_store): New.
9420 (nds32_emit_mem_move): New.
9421 (nds32_emit_mem_move_block): New.
9422 (nds32_expand_movmemsi_loop_unknown_size): New.
9423 (nds32_expand_movmemsi_loop_known_size): New.
9424 (nds32_expand_movmemsi_loop): New.
9425 (nds32_expand_movmemsi_unroll): New.
9426 (nds32_expand_movmemqi): Rename ...
9427 (nds32_expand_movmemsi): ... to this.
9428 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9429 (movmemsi): ... to this.
9430 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9431 (nds32_expand_movmemsi): ... to this.
9433 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9434 Monk Chiang <sh.chiang04@gmail.com>
9435 Chung-Ju Wu <jasonwucj@gmail.com>
9437 * config/nds32/nds32-protos.h
9438 (nds32_expand_load_multiple): New arguments.
9439 (nds32_expand_store_multiple): Ditto.
9440 (nds32_valid_multiple_load_store): Rename ...
9441 (nds32_valid_multiple_load_store_p): ... to this.
9442 * config/nds32/nds32-memory-manipulation.c
9443 (nds32_expand_load_multiple): Refine implementation.
9444 (nds32_expand_store_multiple): Ditto.
9445 * config/nds32/nds32-multiple.md
9446 (load_multiple): Update nds32_expand_load_multiple interface.
9447 (store_multiple): Update nds32_expand_store_multiple interface.
9448 * config/nds32/nds32-predicates.c
9449 (nds32_valid_multiple_load_store): Rename ...
9450 (nds32_valid_multiple_load_store_p): ... to this and refine
9452 * config/nds32/predicates.md
9453 (nds32_load_multiple_and_update_address_operation): New predicate.
9454 (nds32_store_multiple_and_update_address_operation): New predicate.
9456 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9457 Chung-Ju Wu <jasonwucj@gmail.com>
9459 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9460 (combo): New attribute.
9461 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9463 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9465 * config/nds32/nds32.opt: Change -mcmodel= default value.
9467 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9468 Monk Chiang <sh.chiang04@gmail.com>
9469 Chung-Ju Wu <jasonwucj@gmail.com>
9471 * config/nds32/constants.md (unspec_element): New enum.
9472 * config/nds32/constraints.md (Umw): New constraint.
9473 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9474 * config/nds32/nds32-intrinsic.md: Likewise.
9475 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9476 (nds32_valid_smw_lwm_base_p): New.
9477 (nds32_output_smw_single_word): New.
9478 (nds32_output_lmw_single_word): New.
9479 (nds32_expand_unaligned_load): New.
9480 (nds32_expand_unaligned_store): New.
9481 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9482 (nds32_output_smw_single_word): Declare.
9483 (nds32_output_lmw_single_word): Declare.
9484 (nds32_expand_unaligned_load): Declare.
9485 (nds32_expand_unaligned_store): Declare.
9486 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9487 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9488 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9489 NDS32_BUILTIN_UASTORE_DW.
9490 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9493 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9494 Kito Cheng <kito.cheng@gmail.com>
9495 Chung-Ju Wu <jasonwucj@gmail.com>
9497 * config/nds32/nds32-intrinsic.c
9498 (nds32_expand_builtin_null_ftype_reg): Delete.
9499 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9500 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9501 (nds32_read_argument): New.
9502 (nds32_legitimize_target): Ditto.
9503 (nds32_legitimize_argument): Ditto.
9504 (nds32_check_constant_argument): Ditto.
9505 (nds32_expand_unop_builtin): Ditto.
9506 (nds32_expand_unopimm_builtin): Ditto.
9507 (nds32_expand_binop_builtin): Ditto.
9508 (nds32_builtin_decl_impl): Ditto.
9509 (builtin_description): Ditto.
9510 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9511 (nds32_init_builtins_impl): Ditto.
9512 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9513 (nds32_builtin_decl): New.
9514 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9515 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9517 2018-03-02 Jeff Law <law@redhat.com>
9519 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9520 (redundant_insn, fill_simple_delay_slots): Likewise.
9521 (fill_slots_from_thread): Likewise.
9522 * resource.c (mark_referenced_resources): Likewise.
9523 (mark_set_resources, find_dead_or_set_registers): Likewise.
9525 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9527 * substring-locations.h (format_warning_va): Formatting fix for
9529 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9531 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9532 * substring-locations.c: Include intl.h.
9533 (format_warning_va): Turned into small wrapper around
9534 format_warning_n_va, renamed to ...
9535 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9536 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9538 (format_warning_at_substring_n): New function.
9539 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9540 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9541 format_warning_at_substring with just a shorter name instead of
9542 const function pointer.
9543 (fmtwarn_n): New function.
9544 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9545 appropriate, get rid of all the fmtstr temporaries, move conditionals
9546 with G_() wrapped string literals directly into fmtwarn arguments,
9547 cast dir.len to (int), formatting fixes.
9549 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9551 * doc/invoke.texi: Remove "Cilk Plus" references.
9553 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9554 Richard Biener <rguenther@suse.de>
9557 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9558 for error or warning attributes if CALL_FROM_THUNK_P is set.
9561 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9564 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9565 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9568 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9569 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9571 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9572 instead of -1U in last predictors element's probability member.
9574 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9577 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9578 arguments if they are comparable.
9580 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9582 PR tree-optimization/84634
9583 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9584 masks and masked_loop_p with a single loop_masks, making sure it's
9585 null for bb vectorization.
9587 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9589 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9590 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9591 loop->force_vectorize to check whether there is no alias.
9593 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9596 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9598 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9600 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9601 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9602 instead of a loop around prev_real_insn.
9603 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9607 * config/i386/i386.c (ix86_print_operand): Use conditional
9608 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9611 2018-03-02 Richard Biener <rguenther@suse.de>
9613 PR tree-optimization/84427
9614 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9615 (bitmap_set_subtract_values): Rewrite to handle multiple
9618 (prune_clobbered_mems): Likewise.
9619 (phi_translate): Take edge instead of pred/phiblock.
9620 (phi_translate_1): Likewise.
9621 (phi_translate_set): Likewise. Insert all translated
9622 exprs for a value into the set, keeping possibly multiple
9623 expressions per value.
9624 (compute_antic_aux): Adjust for phi_translate changes.
9625 When intersecting union the expressions and prune those
9626 not in the final value set, keeping possibly multiple
9627 expressions per value. Do not use value-insertion
9628 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9629 all expressions. Add verification that the value-sets
9630 only shrink during iteration.
9631 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9632 (do_pre_regular_insertion): Likewise.
9633 (do_pre_partial_partial_insertion): Likewise.
9635 2018-03-02 Richard Biener <rguenther@suse.de>
9638 * config/darwin.c (saved_debug_info_level): New static global.
9639 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9640 (darwin_asm_lto_end): Restore debug info generation settings.
9642 2018-03-01 Martin Liska <mliska@suse.cz>
9645 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9648 2018-03-01 Richard Biener <rguenther@suse.de>
9651 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9652 type annotation with LTO when debug was disabled at compile-time.
9654 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9656 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9658 (mips_final_postscan_insn): Likewise.
9660 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9662 PR rtl-optimization/84528
9663 * alias.c (init_alias_target): Add commentary.
9664 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9665 a unique base value if the frame pointer is not eliminated
9666 to the stack pointer.
9668 2018-03-01 Tom de Vries <tom@codesourcery.com>
9670 PR rtl-optimization/83327
9671 * lra-int.h (hard_regs_spilled_into): Declare.
9672 * lra.c (hard_regs_spilled_into): Define.
9673 (init_reg_info): Init hard_regs_spilled_into.
9674 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9675 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9676 (process_bb_lives): Handle hard_regs_spilled_into.
9677 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9678 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9680 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9682 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9683 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9684 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9685 * config/rs6000/aix72.h: New file.
9687 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9689 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9690 instead of warning_at with conditional singular and plural messages
9694 * stor-layout.c (update_alignment_for_field): For
9695 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9696 && !DECL_PACKED (field), do the alignment update, just use
9697 only desired_align instead of MAX (type_align, desired_align)
9699 (place_field): Don't do known_align < desired_align handling
9700 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9701 is non-NULL, instead do it after rli->prev_field handling and
9702 only if not within a bitfield word. For DECL_PACKED (field)
9703 use type_align of BITS_PER_UNIT.
9705 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9707 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9708 superfluous parentheses and trailing spaces.
9710 2018-02-28 Richard Biener <rguenther@suse.de>
9712 PR tree-optimization/84584
9713 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9714 SCoPs with fake exit edge.
9716 2018-02-28 Martin Liska <mliska@suse.cz>
9719 * timevar.c (timer::print): Fix format to properly print 100%
9722 2018-02-28 Richard Biener <rguenther@suse.de>
9725 * genmatch.c (capture_info::walk_match): Do not mark
9726 captured expressions without operands as expr_p given
9727 they act more like predicates and should be subject to
9728 "lost tail" side-effect preserving.
9730 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9732 PR rtl-optimization/81611
9733 * auto-inc-dec.c (attempt_change): Move dead note from
9734 mem_insn if it's the next use of regno
9735 (find_address): Take address use of reg holding
9736 non-incremented value. Add parm to limit search to the named
9738 (merge_in_block): Attempt to use a mem insn that is the next
9739 use of the original regno.
9741 2018-02-27 Martin Sebor <msebor@redhat.com>
9744 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9745 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9747 2018-02-27 Martin Sebor <msebor@redhat.com>
9749 PR translation/84207
9750 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9751 n argument to unsigned HOST_WIDE_INT.
9752 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9753 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9754 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9755 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9757 2018-02-27 Richard Biener <rguenther@suse.de>
9759 PR tree-optimization/84512
9760 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9761 Do not use the estimate returned from record_stmt_cost for
9762 the scalar iteration cost but sum properly using add_stmt_cost.
9764 2018-02-27 Richard Biener <rguenther@suse.de>
9766 PR tree-optimization/84466
9767 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9768 Adjust last change to less strictly validate use operands.
9770 2018-02-27 Martin Liska <mliska@suse.cz>
9772 PR gcov-profile/84548
9773 * gcov.c (process_file): Allow partial overlap and consider it
9774 also as group functions.
9775 (output_lines): Properly calculate range of lines for a group.
9777 2018-02-27 Martin Liska <mliska@suse.cz>
9779 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9780 'ggc' suffixes. Change first column width.
9781 (timer::print): Fix formatting of the column.
9783 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9785 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9786 preserve inline entry blocks for the sake of debug inline
9787 entry point markers alone.
9788 (remove_unused_locals): Suggest in comments a better place to
9789 force the preservation of inline entry blocks that are
9790 otherwise unused, but do not preserve them.
9792 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9794 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9796 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9799 * config/i386/constraints.md (Bs): Replace
9800 ix86_indirect_branch_register with
9801 TARGET_INDIRECT_BRANCH_REGISTER.
9803 * config/i386/i386.md (indirect_jump): Likewise.
9804 (tablejump): Likewise.
9805 (*sibcall_memory): Likewise.
9806 (*sibcall_value_memory): Likewise.
9807 Peepholes of indirect call and jump via memory: Likewise.
9808 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9809 (*sibcall_value_GOT_32): Likewise.
9810 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9811 (GOT_memory_operand): Likewise.
9812 (call_insn_operand): Likewise.
9813 (sibcall_insn_operand): Likewise.
9814 (GOT32_symbol_operand): Likewise.
9815 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9817 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9819 PR rtl-optimization/83496
9820 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9821 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9822 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9823 redundant insn, if any.
9824 (relax_delay_slots): Likewise.
9825 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9827 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9829 PR tree-optimization/83965
9830 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9831 that grouped statements are part of a reduction chain. Return
9832 true if the statement is not marked as a reduction itself but
9834 (vect_recog_dot_prod_pattern): Don't check whether the statement
9835 is part of a group here.
9836 (vect_recog_sad_pattern): Likewise.
9837 (vect_recog_widen_sum_pattern): Likewise.
9839 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9842 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9844 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9847 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9849 (ix86_output_indirect_function_return): New prototype.
9850 (ix86_split_simple_return_pop_internal): Likewise.
9851 * config/i386/i386.c (indirect_return_via_cx): New.
9852 (indirect_return_via_cx_bnd): Likewise.
9853 (indirect_thunk_name): Handle return va CX_REG.
9854 (output_indirect_thunk_function): Create alias for
9855 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9856 (ix86_output_indirect_jmp): Remove the bool argument.
9857 (ix86_output_indirect_function_return): New function.
9858 (ix86_split_simple_return_pop_internal): Likewise.
9859 * config/i386/i386.md (*indirect_jump): Don't pass false
9860 to ix86_output_indirect_jmp.
9861 (*tablejump_1): Likewise.
9862 (simple_return_pop_internal): Change it to define_insn_and_split.
9863 Call ix86_split_simple_return_pop_internal to split it for
9865 (simple_return_indirect_internal): Call
9866 ix86_output_indirect_function_return instead of
9867 ix86_output_indirect_jmp.
9869 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9872 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9873 memset and value initialization afterwards.
9875 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9877 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9879 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9882 * common/config/aarch64/aarch64-common.c
9883 (aarch_option_optimization_table[]): Switch
9884 off fomit-frame-pointer
9886 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9887 Chung-Ju Wu <jasonwucj@gmail.com>
9889 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9891 (store_multiple): Ditto.
9893 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9895 * config.gcc: Add --with-cpu support for nds32 target.
9896 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9897 * config/nds32/nds32.opt: Add -mcpu= option.
9899 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9901 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9902 isel=yes): Warn for these deprecated options.
9904 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9906 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9907 ISA_2_5_MASKS_EMBEDDED.
9909 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9911 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9912 p->max as pointers rather than using iterative_hash_expr.
9914 2018-02-23 Carl Love <cel@us.ibm.com>
9916 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9917 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9919 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9920 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9921 P8V_BUILTIN_VEC_VUNSIGNED2.
9923 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9926 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9927 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9929 * lra-constraints.c (curr_insn_transform): Set up
9930 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9932 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9933 * lra-eliminations.c (spill_pseudos): Ditto.
9934 (process_insn_for_elimination): Ditto.
9935 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9936 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9937 LRA_NON_CLOBBERED_ALT.
9939 2018-02-22 Martin Sebor <msebor@redhat.com>
9941 PR tree-optimization/84480
9942 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9943 to maybe_diag_stxncpy_trunc. Call it.
9944 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9945 from gimple_fold_builtin_strcpy. Print inlining stack.
9946 (handle_builtin_stxncpy): Print inlining stack.
9947 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9949 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9952 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9953 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9954 and -fcheck-pointer-bounds are used together.
9955 (indirect_thunk_prefix): New enum.
9956 (indirect_thunk_need_prefix): New function.
9957 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9958 "_nt" instead of "_bnd" for NOTRACK prefix.
9959 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9960 (output_indirect_thunk_function): Likewise.
9962 (ix86_code_end): Update output_indirect_thunk_function calls.
9963 (ix86_output_indirect_branch_via_reg): Replace
9964 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9965 (ix86_output_indirect_branch_via_push): Likewise.
9966 (ix86_output_function_return): Likewise.
9967 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9968 incompatible with -fcf-protection=branch and
9969 -fcheck-pointer-bounds.
9971 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9974 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9975 Change gcc_assert call to output_operand_lossage.
9977 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9979 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9981 2018-02-22 DJ Delorie <dj@redhat.com>
9982 Sebastian Perta <sebastian.perta@renesas.com>
9983 Oleg Endo <olegendo@gcc.gnu.org>
9985 * config/rx/rx.c (rx_rtx_costs): New function.
9986 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9988 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9990 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9992 2018-02-22 Martin Liska <mliska@suse.cz>
9995 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9996 Add "native" as a possible value.
9998 2018-02-22 Martin Liska <mliska@suse.cz>
10001 * config/i386/i386.c (ix86_option_override_internal):
10002 Add "native" as a possible value for -march and -mtune.
10004 2018-02-22 Jakub Jelinek <jakub@redhat.com>
10007 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
10008 to all type variants.
10010 PR tree-optimization/84503
10011 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
10012 width as info->bitpos + info->bitsize - start.
10013 (merged_store_group::merge_overlapping): Simplify width computation.
10014 (check_no_overlap): New function.
10015 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10016 start + width and last_order of the group, fail if check_no_overlap
10018 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10019 to group if check_no_overlap fails.
10021 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10023 * config/rs6000/altivec.md: Delete contraint arguments to
10024 define_expand, define_split, and define_peephole2, and in
10025 define_insn_and_split if always unused.
10026 * config/rs6000/darwin.md: Ditto.
10027 * config/rs6000/dfp.md: Ditto.
10028 * config/rs6000/rs6000.md: Ditto.
10029 * config/rs6000/sync.md: Ditto.
10030 * config/rs6000/vector.md: Ditto.
10031 * config/rs6000/vsx.md: Ditto.
10033 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10035 * config/rs6000/altivec.md: Write output control strings as braced
10036 blocks instead of double-quoted strings.
10037 * config/rs6000/darwin.md: Ditto.
10038 * config/rs6000/rs6000.md: Ditto.
10039 * config/rs6000/vector.md: Ditto.
10040 * config/rs6000/vsx.md: Ditto.
10042 2018-02-21 Jason Merrill <jason@redhat.com>
10044 PR c++/84314 - ICE with templates and fastcall attribute.
10045 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10047 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10049 * ipa-cp.c (determine_versionability): Fix comment typos.
10051 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10054 * ipa-cp.c (determine_versionability): Do not version functions caling
10057 2018-02-21 Martin Liska <mliska@suse.cz>
10060 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10061 Add "native" as a possible value.
10062 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10063 the macro when native cpu detection is available.
10065 2018-02-21 Martin Liska <mliska@suse.cz>
10068 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10069 Add "native" as a possible value.
10070 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10071 when native cpu detection is available.
10073 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10074 Martin Sebor <msebor@redhat.com>
10076 PR tree-optimization/84478
10077 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10079 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10080 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10081 support which is conservatively correct, for 2 only stay conservative
10082 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10083 argument to the 2 argument get_range_strlen, adjust 6 arg
10084 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10086 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10087 (gimple_fold_builtin_strlen): Pass true as last argument to
10090 2018-02-20 Martin Sebor <msebor@redhat.com>
10092 PR middle-end/84095
10093 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10094 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10095 (builtin_memref::builtin_memref): Factor out parts into
10096 set_base_and_offset and call it.
10098 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10100 PR middle-end/84406
10101 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10102 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10103 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10104 search at the associated MODE_INT.
10106 2018-02-20 Jeff Law <law@redhat.com>
10108 PR middle-end/82123
10109 PR tree-optimization/81592
10110 PR middle-end/79257
10111 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10112 for range data rather than using global data.
10113 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10114 range data rather than using global data.
10115 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10116 pass it to children as needed.
10117 (struct directive::fmtresult): Similarly.
10118 (struct directive::set_width): Similarly.
10119 (struct directive::set_precision): Similarly.
10120 (format_integer, format_directive, parse_directive): Similarly.
10121 (format_none): Accept unnamed vr_values parameter.
10122 (format_percent, format_floating, format_character): Similarly.
10123 (format_string, format_plain): Similarly.
10124 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10125 the EVRP range analyzer for range data rather than using global data.
10126 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10127 gimple-ssa-evrp-analyze.h
10128 (class sprintf_dom_walker): Add after_dom_children member function.
10129 Add evrp_range_analyzer member.
10130 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10131 range analyzer as needed.
10132 (sprintf_dom_walker::after_dom_children): New member function.
10133 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10135 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10136 (evrp_range_analyzer::pop_to_marker): Likewise.
10138 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10140 PR tree-optimization/84419
10141 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10142 with the required type if its current type is compatible but
10145 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10147 PR middle-end/82004
10148 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10149 after vectorization.
10151 2018-02-20 Martin Liska <mliska@suse.cz>
10154 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10155 possible values if we don't have a hint.
10157 2018-02-20 Martin Liska <mliska@suse.cz>
10161 * final.c (shorten_branches): Build align_tab array with one
10163 * opts.c (finish_options): Add alignment option limit check.
10164 (MAX_CODE_ALIGN): Likewise.
10165 (MAX_CODE_ALIGN_VALUE): Likewise.
10166 * doc/invoke.texi: Document maximum allowed option value for
10167 all -falign-* options.
10169 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10172 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10173 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10174 * var-tracking.c (emit_note_insn_var_location): Remove all references
10175 to NOTE_INSN_CALL_ARG_LOCATION.
10176 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10177 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10178 Use copy_rtx_if_shared.
10179 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10180 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10181 (dwarf2out_var_location): Remove handling of
10182 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10184 * final.c (final_scan_insn): Remove all references to
10185 NOTE_INSN_CALL_ARG_LOCATION.
10186 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10187 before dumping final insns.
10188 * except.c (emit_note_eh_region_end): Remove all references to
10189 NOTE_INSN_CALL_ARG_LOCATION.
10190 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10191 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10192 * config/arc/arc.c (hwloop_optimize): Likewise.
10193 * config/arm/arm.c (create_fix_barrier): Likewise.
10194 * config/s390/s390.c (s390_chunkify_start): Likewise.
10195 * config/sh/sh.c (find_barrier): Likewise.
10196 * config/i386/i386.c (rest_of_insert_endbranch,
10197 ix86_seh_fixup_eh_fallthru): Likewise.
10198 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10199 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10200 * config/frv/frv.c (frv_function_prologue): Likewise.
10201 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10203 (note_outside_basic_block_p): Remove all references to
10204 NOTE_INSN_CALL_ARG_LOCATION.
10205 * gengtype.c (adjust_field_rtx_def): Likewise.
10206 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10208 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10209 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10212 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10215 PR tree-optimization/84452
10216 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10217 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10220 2018-02-19 Martin Liska <mliska@suse.cz>
10223 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10225 2018-02-19 Martin Liska <mliska@suse.cz>
10226 Richard Sandiford <richard.sandiford@linaro.org>
10228 PR tree-optimization/82491
10229 * gimple-fold.c (get_base_constructor): Make earlier bail out
10232 2018-02-19 Carl Love <cel@us.ibm.com>
10234 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10235 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10237 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10238 P8V_BUILTIN_VEC_NEG.
10240 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10242 * config/rl78/rl78.md (movdf): New define expand.
10244 2018-02-19 Martin Liska <mliska@suse.cz>
10247 * doc/invoke.texi: Fix typo.
10248 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10250 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10252 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10253 handle rs6000_single_float and rs6000_double_float specially for
10256 2018-02-16 Jeff Law <law@redhat.com>
10258 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10259 (pop_regs): Use it.
10261 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10264 * ipa-inline.c (inline_small_functions): Fix a typo.
10266 2018-02-16 Nathan Sidwell <nathan@acm.org>
10268 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10270 2018-02-16 Carl Love <cel@us.ibm.com>
10272 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10273 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10274 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10275 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10276 expansion to P8V_BUILTIN_VEC_FLOAT2.
10278 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10280 PR rtl-optimization/70023
10281 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10282 src_regno into account.
10284 2018-02-16 Carl Love <cel@us.ibm.com>
10286 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10287 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10288 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10289 * config/rs6000/rs6000.c: Remove case statements for
10290 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10291 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10292 and P9V_BUILTIN_VEC_VINSERT4B.
10293 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10294 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10295 * config/rs6000/vsx.md:
10296 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10299 2018-02-16 Carl Love <cel@us.ibm.com>
10301 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10303 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10305 * config/rs6000/rs6000-c.c: Add the definitions for
10306 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10307 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10308 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10309 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10310 definition for insert4b and define insn *insert3b_internal.
10311 * doc/extend.texi: Add documentation for vec_extract4b.
10313 2018-02-16 Nathan Sidwell <nathan@acm.org>
10315 * doc/extend.texi (Backwards Compatibility): Mention friend
10316 injection. Note for-scope is deprecated.
10317 * doc/invoke.texi (-ffriend-injection): Deprecate.
10319 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10321 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10322 that moved to I2, also allow destinations that are a paradoxical
10323 subreg (instead of a normal reg).
10325 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10328 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10331 2018-02-16 Richard Biener <rguenther@suse.de>
10333 PR tree-optimization/84037
10334 PR tree-optimization/84016
10336 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10337 Adjust vec_construct for the fact we need additional higher latency
10338 128bit inserts for AVX256 and AVX512 vector builds.
10339 (ix86_add_stmt_cost): Scale vector construction cost for
10342 2018-02-16 Richard Biener <rguenther@suse.de>
10344 PR tree-optimization/84417
10345 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10346 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10347 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10349 2018-02-16 Martin Liska <mliska@suse.cz>
10352 * internal-fn.def (ASAN_CHECK): Set proper flags.
10353 (ASAN_MARK): Likewise.
10355 2018-02-16 Julia Koval <julia.koval@intel.com>
10357 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10358 from PTA_CANNONLAKE.
10360 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10363 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10364 Use ++iter rather than iter++ for std::list iterators.
10365 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10366 defer deleting them until all nodes in the forest are processed. Do
10367 free even leaf nodes. Change to_process into auto_vec.
10370 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10371 * vec.h (vec_default_construct): Use memset instead of placement new
10372 if BROKEN_VALUE_INITIALIZATION is defined.
10373 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10374 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10377 PR rtl-optimization/83723
10378 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10379 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10380 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10382 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10384 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10386 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10388 PR rtl-optimization/81443
10389 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10390 from inner REGs to paradoxical SUBREGs.
10392 2018-02-16 Richard Biener <rguenther@suse.de>
10394 PR tree-optimization/84399
10395 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10396 For operands we can analyze at their definition make sure we can
10397 analyze them at each use as well.
10399 2018-02-16 Richard Biener <rguenther@suse.de>
10401 PR tree-optimization/84190
10402 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10403 volatile accesses if the decl isn't volatile.
10405 2018-02-15 Jason Merrill <jason@redhat.com>
10407 PR c++/84314 - ICE with templates and fastcall attribute.
10408 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10409 TYPE_CANONICAL on an existing type.
10411 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10413 PR tree-optimization/84383
10414 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10415 dstoff nor call operand_equal_p if dstbase is NULL.
10417 PR tree-optimization/84334
10418 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10419 also a CONSTANT_CLASS_P, punt.
10421 2018-02-14 Jim Wilson <jimw@sifive.com>
10423 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10424 first SMALL_OPERAND check. New local min_second_step. Move assert
10425 to where locals are set. Add TARGET_RVC support.
10426 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10428 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10430 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10432 2018-02-14 Martin Sebor <msebor@redhat.com>
10434 PR tree-optimization/83698
10435 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10436 arrays constrain the offset range to their bounds.
10437 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10438 (builtin_access::overlap): Avoid setting the size of overlap if it's
10440 (maybe_diag_overlap): Also consider arrays when deciding what values
10441 of offsets to include in diagnostics.
10443 2018-02-14 Martin Sebor <msebor@redhat.com>
10446 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10447 that correspond to the kind of a declaration.
10449 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10452 * config/pa/pa.md: Load address of PIC label using the linkage table
10453 if the label is nonlocal.
10455 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10457 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10458 warning message if user requests -maltivec=be.
10459 * doc/invoke.texi: Document deprecation of -maltivec=be.
10461 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10464 * config/rs6000/rs6000-c.c: Update definitions for
10465 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10466 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10468 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10471 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10472 add _get_ssp intrinsics. Remove argument from
10473 __builtin_ia32_rdssp[d|q].
10474 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10475 * config/i386/i386-builtin.def: Remove argument from
10476 __builtin_ia32_rdssp[d|q].
10477 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10478 ix86_expand_special_args_builtin for _rdssp[d|q].
10479 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10480 Clear register before usage.
10481 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10482 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10484 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10486 PR tree-optimization/84357
10487 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10488 operand 1 of an ARRAY_REF too.
10490 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10493 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10494 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10496 (set_of_reg): New struct.
10497 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10498 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10499 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10501 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10502 Split into bitclr, bitset, bitinvert patterns if appropriate.
10503 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10504 use rx_fuse_in_memory_bitop.
10505 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10506 to named insn, correct maximum insn length.
10508 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10511 * machmode.def: Define a complex mode for PARTIAL_INT.
10512 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10514 * doc/rtl.texi: Document CSPImode.
10515 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10517 (msp430_hard_regno_nregs_with_padding): Likewise.
10519 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10522 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10524 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10526 PR rtl-optimization/84169
10527 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10528 we generated a parallel as new i3 and we split that to new i2 and i3
10529 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10530 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10531 those to i2, not i1. Partially rewrite this scan code.
10533 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10536 * stor-layout.c (place_field): For variable length fields, adjust
10537 offset_align afterwards not just based on the field's alignment,
10538 but also on the size.
10540 PR middle-end/84309
10541 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10542 of exps and logs in the use_exp2 case.
10544 2018-02-13 Jeff Law <law@redhat.com>
10546 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10547 entry for "vector".
10549 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10552 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10556 * common.opt (gas-loc-support, gas-locview-support): New.
10557 (ginline-points, ginternal-reset-location-views): New.
10558 * doc/invoke.texi: Document them. Use @itemx where intended.
10559 (gvariable-location-views): Adjust.
10560 * target.def (reset_location_view): New.
10561 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10562 (TARGET_RESET_LOCATION_VIEW): New.
10563 * doc/tm.texi: Rebuilt.
10564 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10565 (dwarf2out_default_as_locview_support): New.
10566 (output_asm_line_debug_info): Use option variables.
10567 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10568 (output_loc_list): Likewise.
10569 (add_high_low_attributes): Check option variables.
10570 Don't output entry view attribute in strict mode.
10571 (gen_inlined_subroutine_die): Check option variables.
10572 (dwarf2out_inline_entry): Likewise.
10573 (init_sections_and_labels): Likewise.
10574 (dwarf2out_early_finish): Likewise.
10575 (maybe_reset_location_view): New, from...
10576 (dwarf2out_var_location): ... here. Call it.
10577 * debug.h (dwarf2out_default_as_loc_support): Declare.
10578 (dwarf2out_default_as_locview_support): Declare.
10579 * hooks.c (hook_int_rtx_insn_0): New.
10580 * hooks.h (hook_int_rtx_insn_0): Declare.
10581 * toplev.c (process_options): Take -gas-loc-support and
10582 -gas-locview-support from dwarf2out. Enable
10583 -gvariable-location-views by default only with locview
10584 assembler support. Enable -ginternal-reset-location-views by
10585 default only if the target defines the corresponding hook.
10586 Enable -ginline-points by default if location views are
10587 enabled; force it disabled if statement frontiers are
10589 * tree-inline.c (expand_call_inline): Check option variables.
10590 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10592 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10594 PR tree-optimization/84321
10595 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10596 handling. Also check whether the anti-range contains any values
10597 that satisfy the mask; switch to a VR_RANGE if not.
10599 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10602 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10604 2018-02-13 Martin Jambor <mjambor@suse.cz>
10607 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10608 of call statements, also set location of a load to a temporary.
10610 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10612 * config/rl78/rl78.c (add_vector_labels): New function.
10613 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10614 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10615 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10616 which checks that no arguments are passed.
10617 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10618 * doc/extend.texi: Documentation for the new attribute.
10620 2018-02-13 Andreas Schwab <schwab@suse.de>
10622 * config/riscv/linux.h (CPP_SPEC): Define.
10624 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10627 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10628 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10629 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10630 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10631 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10632 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10633 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10634 temporarily for AES and PCLMUL builtins.
10636 PR tree-optimization/84339
10637 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10638 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10641 PR middle-end/84309
10642 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10643 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10644 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10646 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10648 * omp-simd-clone.h: New file.
10649 * omp-simd-clone.c: Include omp-simd-clone.h.
10650 (expand_simd_clones): No longer static.
10651 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10652 cgraph.h and omp-simd-clone.h.
10653 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10654 (vect_recog_widen_shift_pattern): Formatting fix.
10655 (vect_pattern_recog_1): Don't check optab for calls.
10658 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10659 operands[2] into a REG before using gen_lowpart on it.
10661 2018-02-12 Jeff Law <law@redhat.com>
10664 * config/sh/sh.c (find_barrier): Consider a sibling call
10667 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10668 successfully back substituting a reg.
10670 2018-02-12 Richard Biener <rguenther@suse.de>
10672 PR tree-optimization/84037
10673 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10674 parameter, move visited init to caller.
10675 (vect_slp_analyze_operations): Separate cost from validity
10676 check, initialize visited once for all instances.
10677 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10679 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10681 (vect_model_promotion_demotion_cost): Likewise.
10682 (vectorizable_bswap): Guard cost modeling with !slp_node
10683 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10685 (vectorizable_call): Likewise.
10686 (vectorizable_conversion): Likewise.
10687 (vectorizable_assignment): Likewise.
10688 (vectorizable_shift): Likewise.
10689 (vectorizable_operation): Likewise.
10690 (vectorizable_store): Likewise.
10691 (vectorizable_load): Likewise.
10692 (vectorizable_condition): Likewise.
10693 (vectorizable_comparison): Likewise.
10695 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10698 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10699 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10700 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10702 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10704 PR middle-end/83665
10705 * params.def (inline-min-speedup): Increase from 8 to 15.
10706 (max-inline-insns-auto): Decrease from 40 to 30.
10707 * ipa-split.c (consider_split): Add some buffer for function to
10708 be considered inlining candidate.
10709 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10712 2018-02-12 Richard Biener <rguenther@suse.de>
10714 PR tree-optimization/84037
10715 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10716 matched stmts if we cannot swap the non-matched ones.
10718 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10720 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10721 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10722 _mm_maskz_scalef_round_ss): New intrinsics.
10723 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10724 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10725 __builtin_ia32_scalefss_round): Remove.
10726 (__builtin_ia32_scalefsd_mask_round,
10727 __builtin_ia32_scalefss_mask_round): New intrinsics.
10728 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10729 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10730 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10731 "<round_constraint>")): Changed to ...
10732 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10733 "<round_scalar_constraint>")): ... this.
10734 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10735 %0, %1, %2<round_op3>}"): Changed to ...
10736 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10737 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10738 %2<round_scalar_mask_op3>}"): ... this.
10739 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10741 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10743 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10744 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10745 (_mm_maskz_sqrt_round_ss): New intrinsics.
10746 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10747 (__builtin_ia32_sqrtsd_mask_round)
10748 (__builtin_ia32_sqrtss_mask_round): New builtins.
10749 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10750 (__builtin_ia32_sqrtss_round): Remove.
10751 (__builtin_ia32_sqrtsd_mask_round)
10752 (__builtin_ia32_sqrtss_mask_round): New builtins.
10753 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10754 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10755 ((match_operand:VF_128 1 "vector_operand"
10756 "xBm,<round_constraint>")): Changed to ...
10757 ((match_operand:VF_128 1 "vector_operand"
10758 "xBm,<round_scalar_constraint>")): ... this.
10759 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10760 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10761 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10762 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10763 %<iptr>1<round_scalar_mask_op3>}): ... this.
10764 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10765 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10767 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10770 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10771 Cast vec_cmpeq result to correct type.
10772 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10773 Cast vec_cmpgt result to correct type.
10775 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10777 * final.c (final_scan_insn_1): Renamed from...
10778 (final_scan_insn): ... this. New wrapper, to recover
10779 seen from the outermost call in recursive ones.
10780 * config/sparc/sparc.c (output_return): Drop seen from call.
10781 (output_sibcall): Likewise.
10782 * config/visium/visium.c (output_branch): Likewise.
10784 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10786 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10789 2018-02-10 Alan Modra <amodra@gmail.com>
10792 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10793 Specify LR as an input.
10795 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10798 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10799 remove_member_access_dummy_vars): New functions.
10800 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10801 lower_omp_1, execute_lower_omp): Use them.
10803 PR rtl-optimization/84308
10804 * shrink-wrap.c (spread_components): Release todo vector.
10806 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10808 PR rtl-optimization/57193
10809 * ira-color.c (struct allocno_color_data): Add member
10810 conflict_allocno_hard_prefs.
10811 (update_conflict_allocno_hard_prefs): New.
10812 (bucket_allocno_compare_func): Add a preference based on
10813 conflict_allocno_hard_prefs.
10814 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10815 (color_allocnos): Remove a dead code. Initiate
10816 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10818 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10821 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10822 constraint from =wa to wa. Avoid a subreg on the output operand,
10823 instead use a pseudo and subreg it in a move.
10824 (p9_xxbrd_<mode>): Changed to ...
10825 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10826 (p9_xxbrd_v2df): New expander.
10827 (p9_xxbrw_<mode>): Changed to ...
10828 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10829 (p9_xxbrw_v4sf): New expander.
10831 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10833 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10835 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10838 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10839 multiply in 32-bit mode.
10840 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10841 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10844 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10846 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10847 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10848 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10849 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10851 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10854 * dwarf2out.c (is_trivial_indirect_ref): New function.
10855 (dwarf2out_late_global_decl): Do not generate a location
10856 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10857 and that are not defined in the current unit.
10859 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10861 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10862 instead of a libcall for UNORDERED.
10864 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10867 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10868 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10870 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10873 * config/s390/s390.c (s390_set_current_function): Invoke
10874 s390_indirect_branch_settings also if fndecl didn't change.
10876 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10878 * config/rs6000/rs6000.md (blockage): Set length to zero.
10880 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10882 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10884 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10887 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10888 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10892 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10893 PARALLEL incoming that failed vt_get_decl_and_offset check.
10895 PR middle-end/84237
10896 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10897 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10899 (get_variable_section): For decls in named .bss* sections pass true as
10900 second argument to bss_initializer_p.
10902 2018-02-09 Marek Polacek <polacek@redhat.com>
10903 Jakub Jelinek <jakub@redhat.com>
10906 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10907 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10908 Sync some changes from cxx_fold_indirect_ref.
10910 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10912 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10914 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10915 (BLOCK_INLINE_ENTRY_LABEL): New.
10916 (dwarf2out_var_location): Disregard inline entry markers.
10917 (inline_entry_data): New struct.
10918 (inline_entry_data_hasher): New hashtable type.
10919 (inline_entry_data_hasher::hash): New.
10920 (inline_entry_data_hasher::equal): New.
10921 (inline_entry_data_table): New variable.
10922 (add_high_low_attributes): Add DW_AT_entry_pc and
10923 DW_AT_GNU_entry_view attributes if a pending entry is found
10924 in inline_entry_data_table. Add old entry_pc attribute only
10925 if debug nonbinding markers are disabled.
10926 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10927 markers are enabled.
10928 (block_within_block_p, dwarf2out_inline_entry): New.
10929 (dwarf2out_finish): Check that no entries remained in
10930 inline_entry_data_table.
10931 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10932 (final_scan_insn, notice_source_line): Likewise.
10933 (rest_of_clean_state): Skip inline entry markers.
10934 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10936 * gimple.c (gimple_build_debug_inline_entry): New.
10937 * gimple.h (enum gimple_debug_subcode): Add
10938 GIMPLE_DEBUG_INLINE_ENTRY.
10939 (gimple_build_debug_inline_entry): Declare.
10940 (gimple_debug_inline_entry_p): New.
10941 (gimple_debug_nonbind_marker_p): Adjust.
10942 * insn-notes.def (INLINE_ENTRY): New.
10943 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10944 inline entry marker notes.
10945 (print_insn): Likewise.
10946 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10947 (INSN_DEBUG_MARKER_KIND): Likewise.
10948 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10949 * tree-inline.c (expand_call_inline): Build and insert
10950 debug_inline_entry stmt.
10951 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10952 inline entry blocks early, if nonbind markers are enabled.
10953 (dump_scope_block): Dump fragment info.
10954 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10955 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10956 (gimple_build_debug_inline_entry): New.
10957 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10958 Enable/disable inline entry points too.
10959 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10960 (DEBUG_INSN): Describe inline entry markers.
10962 * common.opt (gvariable-location-views): New.
10963 (gvariable-location-views=incompat5): New.
10964 * config.in: Rebuilt.
10965 * configure: Rebuilt.
10966 * configure.ac: Test assembler for view support.
10967 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10968 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10969 * dwarf2out.c (var_loc_view): New typedef.
10970 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10971 (dwarf2out_locviews_in_attribute): New.
10972 (dwarf2out_locviews_in_loclist): New.
10973 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10974 (enum dw_line_info_opcode): Add LI_adv_address.
10975 (struct dw_line_info_table): Add view.
10976 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10977 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10978 (zero_view_p): New variable.
10979 (ZERO_VIEW_P): New macro.
10980 (output_asm_line_debug_info): New.
10981 (struct var_loc_node): Add view.
10982 (add_AT_view_list, AT_loc_list): New.
10983 (add_var_loc_to_decl): Add view param. Test it against last.
10984 (new_loc_list): Add view params. Record them.
10985 (AT_loc_list_ptr): Handle loc and view lists.
10986 (view_list_to_loc_list_val_node): New.
10987 (print_dw_val): Handle dw_val_class_view_list.
10988 (size_of_die): Likewise.
10989 (value_format): Likewise.
10990 (loc_list_has_views): New.
10991 (gen_llsym): Set vl_symbol too.
10992 (maybe_gen_llsym, skip_loc_list_entry): New.
10993 (dwarf2out_maybe_output_loclist_view_pair): New.
10994 (output_loc_list): Output view list or entries too.
10995 (output_view_list_offset): New.
10996 (output_die): Handle dw_val_class_view_list.
10997 (output_dwarf_version): New.
10998 (output_compilation_unit_header): Use it.
10999 (output_skeleton_debug_sections): Likewise.
11000 (output_rnglists, output_line_info): Likewise.
11001 (output_pubnames, output_aranges): Update version comments.
11002 (output_one_line_info_table): Output view numbers in asm comments.
11003 (dw_loc_list): Determine current endview, pass it to new_loc_list.
11004 Call maybe_gen_llsym.
11005 (loc_list_from_tree_1): Adjust.
11006 (add_AT_location_description): Create view list attribute if
11007 needed, check it's absent otherwise.
11008 (convert_cfa_to_fb_loc_list): Adjust.
11009 (maybe_emit_file): Call output_asm_line_debug_info for test.
11010 (dwarf2out_var_location): Reset views as needed. Precompute
11011 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
11012 attribute. Set view.
11013 (new_line_info_table): Reset next view.
11014 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
11015 (dwarf2out_source_line): Likewise. Output view resets and labels to
11016 the assembler, or select appropriate line info opcodes.
11017 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11018 (optimize_string_length): Catch it. Adjust.
11019 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11020 dw_val_class_view_list, and remove it if no longer needed.
11021 (hash_loc_list): Hash view numbers.
11022 (loc_list_hasher::equal): Compare them.
11023 (optimize_location_lists): Check whether a view list symbol is
11024 needed, and whether the locview attribute is present, and
11025 whether they match. Remove the locview attribute if no longer
11027 (index_location_lists): Call skip_loc_list_entry for test.
11028 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11029 Use output_dwarf_version.
11030 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11031 (struct dw_val_node): Add val_view_list.
11032 * final.c (SEEN_NEXT_VIEW): New.
11033 (set_next_view_needed): New.
11034 (clear_next_view_needed): New.
11035 (maybe_output_next_view): New.
11036 (final_start_function): Rename to...
11037 (final_start_function_1): ... this. Take pointer to FIRST,
11038 add SEEN parameter. Emit param bindings in the initial view.
11039 (final_start_function): Reintroduce SEEN-less interface.
11040 (final): Rename to...
11041 (final_1): ... this. Take SEEN parameter. Output final pending
11042 next view at the end.
11043 (final): Reintroduce seen-less interface.
11044 (final_scan_insn): Output pending next view before switching
11045 sections or ending a block. Mark the next view as needed when
11046 outputting variable locations. Notify debug backend of section
11047 changes, and of location view changes.
11048 (rest_of_handle_final): Adjust.
11049 * toplev.c (process_options): Autodetect value for debug variable
11050 location views option. Warn on incompat5 without -gdwarf-5.
11051 * doc/invoke.texi (gvariable-location-views): New.
11052 (gvariable-location-views=incompat5): New.
11053 (gno-variable-location-views): New.
11055 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11057 PR tree-optimization/84136
11058 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11059 that the result of find_edge is non-NULL.
11061 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11064 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11065 storing integer register in SImode. Fix cost of 256 and 512
11066 byte aligned SSE register store.
11068 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11070 * config/i386/i386.c (ix86_multiplication_cost): Fix
11071 multiplication cost for TARGET_AVX512DQ.
11073 2018-02-08 Marek Polacek <polacek@redhat.com>
11075 PR tree-optimization/84238
11076 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11079 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11081 PR tree-optimization/84265
11082 * tree-vect-stmts.c (vectorizable_store): Don't treat
11083 VMAT_CONTIGUOUS accesses as grouped.
11084 (vectorizable_load): Likewise.
11086 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11088 PR tree-optimization/81635
11089 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11090 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11091 (test_round_for_mask): New functions.
11092 (wide_int_cc_tests): Call test_round_for_mask.
11093 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11094 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11095 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11096 range returned by get_range_info.
11098 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11101 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11102 * symtab.c: Include builtins.h
11103 (symtab_node::output_to_lto_symbol_table_p): Move here
11104 from lto-streamer-out.c:output_symbol_p.
11105 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11106 (output_symbol_p): Move all logic to symtab.c
11107 (produce_symtab): Update.
11109 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11111 * config/s390/s390-opts.h (enum indirect_branch): Define.
11112 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11113 (s390_indirect_branch_via_thunk)
11114 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11115 (enum s390_indirect_branch_type): Define.
11116 * config/s390/s390.c (struct s390_frame_layout, struct
11117 machine_function): Remove.
11118 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11119 (indirect_branch_table_label_no, indirect_branch_table_name):
11121 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11122 (enum s390_indirect_branch_option): Define.
11123 (s390_return_addr_from_memory): New function.
11124 (s390_handle_string_attribute): New function.
11125 (s390_attribute_table): Add new attribute handler.
11126 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11127 (s390_indirect_branch_via_thunk): New function.
11128 (s390_indirect_branch_via_inline_thunk): New function.
11129 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11130 sibling call optimization for non z10 compiles.
11131 (s390_emit_call): Force indirect branch target to be a single
11132 register. Add r1 clobber for non-z10 compiles.
11133 (s390_emit_epilogue): Emit return jump via return_use expander.
11134 (s390_reorg): Handle JUMP_INSNs as execute targets.
11135 (s390_option_override_internal): Perform validity checks for the
11136 new command line options.
11137 (s390_indirect_branch_attrvalue): New function.
11138 (s390_indirect_branch_settings): New function.
11139 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11140 (s390_output_indirect_thunk_function): New function.
11141 (s390_code_end): Implement target hook.
11142 (s390_case_values_threshold): Implement target hook.
11143 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11145 * config/s390/s390.h (struct s390_frame_layout)
11146 (struct machine_function): Move here from s390.c.
11147 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11148 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11149 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11150 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11151 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11152 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11153 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11154 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11155 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11156 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11157 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11158 (mnemonic attribute): Add values which aren't recognized
11160 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11161 pattern for branch conversion. Fix mnemonic attribute.
11162 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11163 indirect branch via thunk if requested.
11164 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11165 ("*indirect_jump"): Disable for branch conversion using out of
11167 ("indirect_jump_via_thunk<mode>_z10")
11168 ("indirect_jump_via_thunk<mode>")
11169 ("indirect_jump_via_inlinethunk<mode>_z10")
11170 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11171 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11172 ("casesi_jump_via_inlinethunk<mode>_z10")
11173 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11174 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11175 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11176 ("*indirect2_jump"): Disable for branch conversion.
11177 ("casesi_jump"): Turn into expander and expand patterns for branch
11179 ("return_use"): New expander.
11180 ("*return"): Emit return via thunk and rename it to ...
11181 ("*return<mode>"): ... this one.
11182 * config/s390/s390.opt: Add new options and and enum for the
11185 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11187 * lra-constraints.c (match_reload): Unconditionally use
11188 gen_lowpart_SUBREG, rather than selecting between that
11189 and equivalent gen_rtx_SUBREG code.
11191 2018-02-08 Richard Biener <rguenther@suse.de>
11193 PR tree-optimization/84233
11194 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11195 changed flag instead of boguously re-using phi_inserted.
11197 2018-02-08 Martin Jambor <mjambor@suse.cz>
11199 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11200 static local variables.
11202 2018-02-08 Richard Biener <rguenther@suse.de>
11204 PR tree-optimization/84278
11205 * tree-vect-stmts.c (vectorizable_store): When looking for
11206 smaller vector types to perform grouped strided loads/stores
11207 make sure the mode is supported by the target.
11208 (vectorizable_load): Likewise.
11210 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11212 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11213 Increase LDP/STP opportunities by adding adjacent callee-saves.
11215 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11217 PR rtl-optimization/84068
11218 PR rtl-optimization/83459
11219 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11221 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11223 PR tree-optimization/84224
11224 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11225 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11226 non-zero arguments.
11228 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11231 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11232 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11233 count, remove one USE.
11235 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11237 * doc/install.texi (Configuration): Document the
11238 --with-long-double-format={ibm,ieee} PowerPC configuration
11242 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11243 Convert from define_expand to be define_insn_and_split. Rework
11244 float/double/_Float128 conversions to QI/HI/SImode to work with
11245 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11246 conversions to QI/HImode types did a store and then a load to
11247 truncate the value. For conversions to VSX registers, don't split
11248 the insn, instead emit the code directly. Use the code iterator
11249 any_fix to combine signed and unsigned conversions.
11250 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11251 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11252 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11253 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11254 (fix_<mode>di2_hw): Likewise.
11255 (fixuns_<mode>di2_hw): Likewise.
11256 (fix_<mode>si2_hw): Likewise.
11257 (fixuns_<mode>si2_hw): Likewise.
11258 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11259 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11260 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11261 fix<uns>_trunc<SFDF:mode>si2_p8.
11262 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11263 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11264 (fix<uns>_<mode>_mem): Likewise.
11265 (fctiw<u>z_<mode>_mem): Likewise.
11266 (fix<uns>_<mode>_mem): Likewise.
11267 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11268 the register allocator from doing a direct move to the GPRs to do
11269 a store, and instead use the ISA 3.0 store byte/half-word from
11270 vector register instruction. For IEEE 128-bit floating point,
11271 also optimize stores of 32-bit ints.
11272 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11274 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11276 * genextract.c (push_pathstr_operand): New function to support
11278 (walk_rtx): Call push_pathstr_operand.
11279 (print_path): Support [a-zA-Z].
11281 2018-02-07 Richard Biener <rguenther@suse.de>
11283 PR tree-optimization/84037
11284 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11285 (cse_and_gimplify_to_preheader): Declare.
11286 (vect_get_place_in_interleaving_chain): Likewise.
11287 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11289 (_loop_vec_info::~_loop_vec_info): Delete it.
11290 (cse_and_gimplify_to_preheader): New function.
11291 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11292 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11293 (vectorizable_load): Likewise. For grouped stores always base
11294 the IV on the first element.
11295 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11296 condition before gimplifying.
11298 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11300 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11301 *DIV_EXPR and *MOD_EXPR.
11303 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11306 * config/i386/i386.c (ix86_option_override_internal): Mask out
11307 the CF_SET bit when checking -fcf-protection.
11309 2018-02-07 Tom de Vries <tom@codesourcery.com>
11312 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11315 2018-02-07 Richard Biener <rguenther@suse.de>
11317 PR tree-optimization/84204
11318 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11321 PR tree-optimization/84205
11322 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11323 special-case isl_ast_op_zdiv_r.
11325 PR tree-optimization/84223
11326 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11327 Only add conditions from within the region.
11328 (gather_bbs::after_dom_children): Adjust.
11330 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11333 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11334 * config/avr/avr.md: Only post-reload split REG-REG moves if
11335 either register is GENERAL_REG_P.
11337 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11339 PR tree-optimization/84235
11340 * tree-ssa-scopedtables.c
11341 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11342 if the subtraction is performed in floating point type where NaNs are
11343 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11344 build 1. Formatting fix.
11346 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11349 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11350 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11351 and skip it regardless of bb boundaries. Use CALL_P macro,
11352 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11353 unnecessarily, formatting fix.
11355 2018-02-06 Michael Collison <michael.collison@arm.com>
11357 * config/arm/thumb2.md:
11358 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11359 (*thumb_mov_notscc): Ditto.
11361 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11364 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11365 unsigned_fix, not "s".
11367 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11369 * configure.ac (gcc_fn_eh_frame_ro): New function.
11370 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11371 correct .eh_frame permissions.
11372 * configure: Regenerate.
11374 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11376 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11377 irrelevant options.
11379 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11381 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11382 Display warning message for -mno-speculate-indirect-jumps.
11384 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11386 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11388 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11390 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11392 PR tree-optimization/84225
11393 * tree-eh.c (find_trapping_overflow): Only call
11394 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11396 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11399 * config/i386/i386.c: Reimplement the check of possible options
11400 -mibt/-mshstk conbination. Change error messages.
11401 * doc/invoke.texi: Fix a typo: remove extra '='.
11403 2018-02-06 Marek Polacek <polacek@redhat.com>
11405 PR tree-optimization/84228
11406 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11408 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11411 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11412 emitted arch directives.
11413 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11414 __ARM_FEATURE_COPROC before changing architectures.
11416 2018-02-06 Richard Biener <rguenther@suse.de>
11418 * config/i386/i386.c (print_reg): Fix typo.
11419 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11421 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11423 * configure: Regenerate.
11425 2018-02-05 Martin Sebor <msebor@redhat.com>
11427 PR tree-optimization/83369
11428 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11431 2018-02-05 Martin Liska <mliska@suse.cz>
11433 * doc/invoke.texi: Cherry-pick upstream r323995.
11435 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11437 * ira.c (ira_init_register_move_cost): Adjust comment.
11439 2018-02-05 Martin Liska <mliska@suse.cz>
11441 PR gcov-profile/84137
11442 * doc/gcov.texi: Fix typo in documentation.
11444 2018-02-05 Martin Liska <mliska@suse.cz>
11446 PR gcov-profile/83879
11447 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11448 using dlopen functionality.
11450 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11452 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11453 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11454 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11455 _mm_maskz_range_round_ss): New intrinsics.
11456 (__builtin_ia32_rangesd128_round)
11457 (__builtin_ia32_rangess128_round): Remove.
11458 (__builtin_ia32_rangesd128_mask_round,
11459 __builtin_ia32_rangess128_mask_round): New builtins.
11460 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11461 __builtin_ia32_rangess128_round): Remove.
11462 (__builtin_ia32_rangesd128_mask_round,
11463 __builtin_ia32_rangess128_mask_round): New builtins.
11464 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11465 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11466 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11467 "<round_saeonly_constraint>")): Changed to ...
11468 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11469 "<round_saeonly_scalar_constraint>")): ... this.
11470 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11471 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11472 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11473 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11474 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11476 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11478 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11480 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11481 Remove all values except native, 8540 and 8548.
11483 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11485 * config/i386/i386.c (ix86_output_function_return): Pass
11486 INVALID_REGNUM, instead of -1, as invalid register number to
11487 indirect_thunk_name and output_indirect_thunk.
11489 2018-02-02 Julia Koval <julia.koval@intel.com>
11491 * config.gcc: Add -march=icelake.
11492 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11493 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11494 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11495 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11496 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11497 (processor_target_table): Add icelake.
11498 (ix86_option_override_internal): Handle new PTAs.
11499 (get_builtin_code_for_version): Handle icelake.
11500 (M_INTEL_COREI7_ICELAKE): New.
11501 (fold_builtin_cpu): Handle icelake.
11502 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11503 * doc/invoke.texi: Add -march=icelake.
11505 2018-02-02 Julia Koval <julia.koval@intel.com>
11507 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11508 to wide_int_bitmask.
11509 * wide-int-bitmask.h: New.
11511 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11514 * config/i386/i386.md: Replace Pmode with word_mode in
11515 builtin_setjmp_setup and builtin_longjmp to support x32.
11517 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11521 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11523 (rs6000_supported_cpu_names): New static variable.
11524 (linux_cpu_translation_table): Likewise.
11525 (elf_platform) <cpu>: Define new static variable and use it.
11526 Translate kernel AT_PLATFORM name to canonical name if needed.
11527 Error if platform name is unknown.
11529 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11532 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11534 2018-02-01 Jeff Law <law@redhat.com>
11537 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11538 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11539 the scratch if RELEASE_VIA_POP is false.
11540 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11541 If we have to save a temporary register, decrement SIZE appropriately.
11542 Pass new arguments to release_scratch_register_on_entry.
11543 (ix86_adjust_stack_and_probe): Likewise.
11544 (ix86_emit_probe_stack_range): Pass new arguments to
11545 release_scratch_register_on_entry.
11547 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11549 PR rtl-optimization/84157
11550 * combine.c (change_zero_ext): Use REG_P predicate in
11551 front of HARD_REGISTER_P predicate.
11553 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11555 * config/avr/avr.c (avr_option_override): Move disabling of
11556 -fdelete-null-pointer-checks to...
11557 * common/config/avr/avr-common.c (avr_option_optimization_table):
11560 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11562 PR tree-optimization/81635
11563 * tree-data-ref.c (split_constant_offset_1): For types that
11564 wrap on overflow, try to use range info to prove that wrapping
11567 2018-02-01 Renlin Li <renlin.li@arm.com>
11570 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11571 TAILCALL_ADDR_REGS.
11572 (aarch64_register_move_cost): Likewise.
11573 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11574 TAILCALL_ADDR_REGS.
11575 (REG_CLASS_NAMES): Likewise.
11576 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11577 TAILCALL_ADDR_REGS. Remove IP registers.
11578 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11580 2018-02-01 Richard Biener <rguenther@suse.de>
11582 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11583 for specifying RPO order and allow NULL for that.
11584 * domwalk.c (dom_walker::dom_walker): Likewise.
11585 (dom_walker::walk): Handle NULL RPO order.
11586 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11588 (rewrite_update_dom_walker): Likewise.
11589 (mark_def_dom_walker): Likewise.
11591 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11593 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11594 (aarch64_maybe_expand_sve_subreg_move): Declare.
11595 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11596 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11598 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11599 that are semantically a reverse operation.
11600 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11601 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11602 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11604 (aarch64_can_change_mode_class): For big-endian, forbid changes
11605 between two SVE modes if they have different element sizes.
11607 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11609 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11610 the TImode handling for big-endian targets.
11612 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11614 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11615 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11617 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11618 Remove BSWAP handing for big-endian targets and use the form of
11619 LD1RQ appropariate for the mode.
11621 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11623 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11624 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11625 duplicated element.
11627 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11630 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11631 check for operands that need to go through aarch64_sve_reload_be.
11633 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11635 PR tree-optimization/81661
11636 PR tree-optimization/84117
11637 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11638 * tree-eh.c: Include gimplify.h.
11639 (find_trapping_overflow, replace_trapping_overflow,
11640 rewrite_to_non_trapping_overflow): New functions.
11641 * tree-vect-loop.c: Include tree-eh.h.
11642 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11643 * tree-data-ref.c: Include tree-eh.h.
11644 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11646 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11648 PR rtl-optimization/84123
11649 * combine.c (change_zero_ext): Check if hard register satisfies
11650 can_change_dest_mode before calling gen_lowpart_SUBREG.
11652 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11655 * ira.c (ira_init_register_move_cost): Remove assert.
11657 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11659 PR rtl-optimization/84071
11660 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11661 * doc/tm.texi: Regenerate.
11663 2018-01-31 Richard Biener <rguenther@suse.de>
11665 PR tree-optimization/84132
11666 * tree-data-ref.c (analyze_miv_subscript): Properly
11667 check whether evolution_function_is_affine_multivariate_p
11668 before calling gcd_of_steps_may_divide_p.
11670 2018-01-31 Julia Koval <julia.koval@intel.com>
11673 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11674 * config/i386/i386.md (rdpid_rex64) New.
11675 (rdpid): Make 32bit only.
11677 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11680 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11681 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11683 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11686 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11688 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11690 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11692 PR rtl-optimization/84071
11693 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11694 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11696 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11698 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11699 (arc_attribute_table): Add 'aux' attribute.
11700 (arc_in_small_data_p): Consider aux like variables.
11701 (arc_is_aux_reg_p): New function.
11702 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11703 (arc_get_aux_arg): New function.
11704 (prepare_move_operands): Handle aux-register access.
11705 (arc_handle_aux_attribute): New function.
11706 * doc/extend.texi (ARC Variable attributes): Add subsection.
11708 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11710 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11711 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11712 (arc_attribute_table): Add 'uncached' attribute.
11713 (arc_print_operand): Print '.di' flag for uncached memory
11715 (arc_in_small_data_p): Do not consider for small data the uncached
11717 (arc_is_uncached_mem_p): New function.
11718 * config/arc/predicates.md (compact_store_memory_operand): Check
11719 for uncached memory accesses.
11720 (nonvol_nonimm_operand): Likewise.
11721 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11723 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11726 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11727 falign-loops=): Add Optimization flag.
11729 2018-01-30 Jeff Law <law@redhat.com>
11732 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11733 INT_REGISTERS_SAVED. Check it prior to calling
11734 get_scratch_register_on_entry.
11735 (ix86_adjust_stack_and_probe): Similarly.
11736 (ix86_emit_probe_stack_range): Similarly.
11737 (ix86_expand_prologue): Corresponding changes.
11739 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11742 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11743 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11745 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11748 * lra-constraints.c (curr_insn_transform): Process AND in the
11751 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11753 PR rtl-optimization/83986
11754 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11755 dependence against last_pending_memory_flush in addition to
11756 pending_jump_insns.
11758 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11760 PR tree-optimization/81611
11761 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11764 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11767 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11770 2018-01-30 Richard Biener <rguenther@suse.de>
11771 Jakub Jelinek <jakub@redhat.com>
11773 PR tree-optimization/84111
11774 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11775 inner loops added during recursion, as they don't have up-to-date
11778 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11781 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11782 (can_inline_edge_by_limits_p): ... here.
11783 (can_early_inline_edge_p, check_callers,
11784 update_caller_keys, update_callee_keys, recursive_inlining,
11785 add_new_edges_to_heap, speculation_useful_p,
11786 inline_small_functions,
11787 inline_small_functions, flatten_function,
11788 inline_to_all_callers_1): Update.
11790 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11792 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11795 2018-01-30 Richard Biener <rguenther@suse.de>
11797 PR tree-optimization/83008
11798 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11799 invariant and constant vector uses in stmts when they need
11800 more than one stmt.
11802 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11805 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11806 * configure: Regenerate.
11808 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11810 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11812 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11813 Use gen_rtx_REG rather than gen_lowpart.
11815 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11817 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11818 rather than 0 when creating partial subregs.
11820 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11822 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11825 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11828 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11829 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11830 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11831 flags. This restores the settings used before the 2017-07-24.
11832 Turning off pre increment/decrement/modify allows IVOPTS to
11833 optimize DF/SF loops where the index is an int.
11835 2018-01-29 Richard Biener <rguenther@suse.de>
11836 Kelvin Nilsen <kelvin@gcc.gnu.org>
11839 * tree-vect-stmts.c (vectorizable_call): Don't call
11840 targetm.vectorize_builtin_md_vectorized_function if callee is
11843 2018-01-22 Carl Love <cel@us.ibm.com>
11845 * doc/extend.tex: Fix typo in second arg in
11846 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11848 2018-01-29 Richard Biener <rguenther@suse.de>
11850 PR tree-optimization/84086
11851 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11852 (flush_ssaname_freelist): When SSA names were released reset
11853 the SCEV hash table.
11855 2018-01-29 Richard Biener <rguenther@suse.de>
11857 PR tree-optimization/84057
11858 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11859 removed paths when removing edges.
11861 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11863 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11864 -mfunction-return=@var{choice}.
11866 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11868 PR diagnostic/84034
11869 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11870 Handle CR like TAB.
11871 (layout::print_source_line): Likewise.
11872 (test_get_line_width_without_trailing_whitespace): Add test cases.
11874 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11876 PR middle-end/84040
11877 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11880 2018-01-26 Jim Wilson <jimw@sifive.com>
11882 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11884 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11887 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11889 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11890 and CMP + SUB-immediate -> SUBS.
11892 2018-01-26 Martin Sebor <msebor@redhat.com>
11894 PR tree-optimization/83896
11895 * tree-ssa-strlen.c (get_string_len): Rename...
11896 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11897 Avoid assuming length is constant.
11898 (handle_char_store): Use HOST_WIDE_INT for string length.
11900 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11903 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11904 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11906 2018-01-26 Richard Biener <rguenther@suse.de>
11908 PR rtl-optimization/84003
11909 * dse.c (record_store): Only record redundant stores when
11910 the earlier store aliases at least all accesses the later one does.
11912 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11914 PR rtl-optimization/83985
11915 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11916 REG_CFA_RESTORE insns.
11917 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11918 REG_CFA_RESTORE insns here.
11921 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11922 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11924 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11926 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11927 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11928 (arc_init): Likewise.
11929 (arc_override_options): Likewise.
11930 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11932 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11934 * config/arc/arc.h (TARGET_DBNZ): Define.
11935 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11936 properly set the tune attribute.
11937 (dbnz): Use TARGET_DBNZ guard.
11938 * config/arc/arc.opt (mtune): Add core3 option.
11940 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11942 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11943 recognize new pic like addresses.
11944 (arc_delegitimize_address): Clean up.
11946 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11948 * config/arc/arc-arches.def: Option mrf16 valid for all
11950 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11951 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11952 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11953 * config/arc/arc-tables.opt: Regenerate.
11954 * config/arc/arc.c (arc_conditional_register_usage): Handle
11955 reduced register file case.
11956 (arc_file_start): Set must have build attributes.
11957 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11958 mrf16 option value.
11959 * config/arc/arc.opt (mrf16): Add new option.
11960 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11961 * config/arc/genmultilib.awk: Handle new mrf16 option.
11962 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11963 * config/arc/t-multilib: Regenerate.
11964 * doc/invoke.texi (ARC Options): Document mrf16 option.
11966 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11968 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11969 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11970 (arc_attribute_table): Add 'secure_call' attribute.
11971 (arc_print_operand): Print secure call operand.
11972 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11974 (arc_is_secure_call_p): New function. * config/arc/arc.md
11975 (call_i): Add support for sjli instruction.
11976 (call_value_i): Likewise.
11977 * config/arc/constraints.md (Csc): New constraint.
11979 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11980 John Eric Martin <John.Martin@emmicro-us.com>
11982 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11983 * config/arc/arc.c (_arc_jli_section): New struct.
11984 (arc_jli_section): New type.
11985 (rc_jli_sections): New static variable.
11986 (arc_handle_jli_attribute): New function.
11987 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11988 (arc_file_end): New function.
11989 (TARGET_ASM_FILE_END): Define.
11990 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11991 (arc_add_jli_section): New function.
11992 (jli_call_scan): Likewise.
11993 (arc_reorg): Call jli_call_scan.
11994 (arc_output_addsi): Remove 'S' from printing asm operand.
11995 (arc_is_jli_call_p): New function.
11996 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11998 (movhi_insn): Likewise.
11999 (movsi_insn): Likewise.
12000 (movsi_set_cc_insn): Likewise.
12001 (loadqi_update): Likewise.
12002 (load_zeroextendqisi_update): Likewise.
12003 (load_signextendqisi_update): Likewise.
12004 (loadhi_update): Likewise.
12005 (load_zeroextendhisi_update): Likewise.
12006 (load_signextendhisi_update): Likewise.
12007 (loadsi_update): Likewise.
12008 (loadsf_update): Likewise.
12009 (movsicc_insn): Likewise.
12010 (bset_insn): Likewise.
12011 (bxor_insn): Likewise.
12012 (bclr_insn): Likewise.
12013 (bmsk_insn): Likewise.
12014 (bicsi3_insn): Likewise.
12015 (cmpsi_cc_c_insn): Likewise.
12016 (movsi_ne): Likewise.
12017 (movsi_cond_exec): Likewise.
12018 (clrsbsi2): Likewise.
12019 (norm_f): Likewise.
12029 (call_i): Remove 'S' asm letter, add jli instruction.
12030 (call_value_i): Likewise.
12031 * config/arc/arc.op (mjli-always): New option.
12032 * config/arc/constraints.md (Cji): New constraint.
12033 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12035 (subsf3_fpx): Likewise.
12036 (mulsf3_fpx): Likewise.
12037 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12039 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12040 function attrbutes.
12041 * doc/invoke.texi (ARC): Document mjli-always option.
12043 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12045 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12046 avoid addition with 0 and use incw and decw where possible.
12048 2018-01-26 Richard Biener <rguenther@suse.de>
12050 PR tree-optimization/81082
12051 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12052 association if it requires casting to unsigned.
12053 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12054 from fold_plusminus_mult_expr to catch important cases late when
12055 range info is available.
12057 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12059 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12060 * configure.ac (hidden_linkonce): New test.
12061 * configure: Regenerate.
12062 * config.in: Regenerate.
12064 2018-01-26 Julia Koval <julia.koval@intel.com>
12066 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12067 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12068 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12069 _mm_mask_bitshuffle_epi64_mask): Fix type.
12070 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12071 USI_FTYPE_V4DI_V4DI_USI): Remove.
12072 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12073 __builtin_ia32_vpshufbitqmb256_mask,
12074 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12075 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12076 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12078 2018-01-26 Alan Modra <amodra@gmail.com>
12081 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12082 UNSPEC_VBPERMQ. Sort other unspecs.
12084 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12086 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12088 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12090 PR middle-end/83055
12091 * predict.c (drop_profile): Do not push/pop cfun; update also
12093 (handle_missing_profiles): Fix logic looking for zero profiles.
12095 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12097 PR middle-end/83977
12098 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12099 on functions with #pragma omp declare simd or functions with simd
12101 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12102 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12103 Remove trailing \n from warning_at calls.
12105 2018-01-25 Tom de Vries <tom@codesourcery.com>
12108 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12109 for neutered workers.
12111 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12114 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12115 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12117 2018-01-24 Jeff Law <law@redhat.com>
12120 * i386.c (get_probe_interval): Move to earlier point.
12121 (ix86_compute_frame_layout): If -fstack-clash-protection and
12122 the frame is larger than the probe interval, then use pushes
12123 to save registers rather than reg->mem moves.
12124 (ix86_expand_prologue): Remove conditional for int_registers_saved
12127 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12130 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12131 min/max for never referenced object.
12133 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12135 PR middle-end/83977
12136 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12138 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12139 attributes from DECL_ATTRIBUTES (decl) without affecting
12140 DECL_ATTRIBUTES (current_function_decl).
12141 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12142 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12144 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12146 PR tree-optimization/83979
12147 * fold-const.c (fold_comparison): Use constant_boolean_node
12148 instead of boolean_{true,false}_node.
12150 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12152 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12155 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12157 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12158 Simplify the clause that sets the length attribute.
12159 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12160 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12161 clause that sets the length attribute.
12162 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12164 2018-01-24 Tom de Vries <tom@codesourcery.com>
12167 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12168 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12169 Add strict parameter.
12170 (prevent_branch_around_nothing): Insert dummy insn between branch to
12171 label and label with no ptx insn inbetween.
12172 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12174 2018-01-24 Tom de Vries <tom@codesourcery.com>
12177 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12178 for neutered threads in warp.
12179 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12181 2018-01-24 Richard Biener <rguenther@suse.de>
12183 PR tree-optimization/83176
12184 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12187 2018-01-24 Richard Biener <rguenther@suse.de>
12189 PR tree-optimization/82819
12190 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12191 code generating pluses that are no-ops in the target precision.
12193 2018-01-24 Richard Biener <rguenther@suse.de>
12195 PR middle-end/84000
12196 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12198 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12200 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12201 to merge probabilities.
12202 * predict.c (probably_never_executed): Also mark as cold functions
12203 with global 0 profile and guessed local profile.
12204 * profile-count.c (profile_probability::combine_with_count): New
12206 * profile-count.h (profile_probability::operator*,
12207 profile_probability::operator*=, profile_probability::operator/,
12208 profile_probability::operator/=): Reduce precision to adjusted
12209 and set value to guessed on contradictory divisions.
12210 (profile_probability::combine_with_freq): Remove.
12211 (profile_probability::combine_wiht_count): Declare.
12212 (profile_count::force_nonzero):: Set to adjusted.
12213 (profile_count::probability_in):: Set quality to adjusted.
12214 * tree-ssa-tail-merge.c (replace_block_by): Use
12215 combine_with_count.
12217 2018-01-23 Andrew Waterman <andrew@sifive.com>
12218 Jim Wilson <jimw@sifive.com>
12220 * config/riscv/riscv.c (riscv_stack_boundary): New.
12221 (riscv_option_override): Set riscv_stack_boundary. Handle
12222 riscv_preferred_stack_boundary_arg.
12223 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12224 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12225 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12226 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12227 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12228 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12230 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12233 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12234 of struct ix86_frame.
12235 (ix86_expand_epilogue): Likewise. Add a local variable for
12236 the reg_save_offset field in struct ix86_frame.
12238 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12240 PR tree-optimization/82604
12241 * tree-loop-distribution.c (enum partition_kind): New enum item
12242 PKIND_PARTIAL_MEMSET.
12243 (partition_builtin_p): Support above new enum item.
12244 (generate_code_for_partition): Ditto.
12245 (compute_access_range): Differentiate cases that equality can be
12246 proven at all loops, the innermost loops or no loops.
12247 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12248 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12249 (finalize_partitions, distribute_loop): Don't fuse partition of
12250 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12251 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12252 parloop is enabled.
12254 2018-01-23 Martin Liska <mliska@suse.cz>
12256 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12257 order to ignore the predictor.
12258 (PRED_POLYMORPHIC_CALL): Likewise.
12259 (PRED_RECURSIVE_CALL): Likewise.
12261 2018-01-23 Martin Liska <mliska@suse.cz>
12263 * tree-profile.c (tree_profiling): Print function header to
12264 aware reader which function we are working on.
12265 * value-prof.c (gimple_find_values_to_profile): Do not print
12266 not interesting value histograms.
12268 2018-01-23 Martin Liska <mliska@suse.cz>
12270 * profile-count.h (enum profile_quality): Add
12271 profile_uninitialized as the first value. Do not number values
12272 as they are zero based.
12273 (profile_count::verify): Update sanity check.
12274 (profile_probability::verify): Likewise.
12276 2018-01-23 Nathan Sidwell <nathan@acm.org>
12278 * doc/invoke.texi (ffor-scope): Deprecate.
12280 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12282 PR tree-optimization/83510
12283 * domwalk.c (set_all_edges_as_executable): New function.
12284 (dom_walker::dom_walker): Convert bool param
12285 "skip_unreachable_blocks" to enum reachability. Move setup of
12286 edge flags to set_all_edges_as_executable and only do it when
12287 reachability is REACHABLE_BLOCKS.
12288 * domwalk.h (enum dom_walker::reachability): New enum.
12289 (dom_walker::dom_walker): Convert bool param
12290 "skip_unreachable_blocks" to enum reachability.
12291 (set_all_edges_as_executable): New decl.
12292 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12293 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12295 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12296 but converting true to REACHABLE_BLOCKS.
12297 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12299 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12300 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12301 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12303 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12304 if check_all_array_refs will be called.
12306 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12308 * tree.c (selftest::test_location_wrappers): Add more test
12311 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12313 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12314 (selftest::test_bit_in_range): Likewise.
12316 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12319 * doc/sourcebuild.texi (vect_float): Say that the selector
12320 only describes the situation when -funsafe-math-optimizations is on.
12321 (vect_float_strict): Document.
12323 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12325 PR tree-optimization/83965
12326 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12327 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12328 instead of checking only for a reduction.
12329 (vect_recog_widen_sum_pattern): Likewise.
12331 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12333 * predict.c (probably_never_executed): Only use precise profile info.
12334 (compute_function_frequency): Skip after inlining hack since we now
12335 have quality checking.
12337 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12339 * profile-count.h (profile_probability::very_unlikely,
12340 profile_probability::unlikely, profile_probability::even): Set
12341 precision to guessed.
12343 2018-01-23 Richard Biener <rguenther@suse.de>
12345 PR tree-optimization/83963
12346 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12347 Properly terminate dominator walk when crossing the exit edge not
12348 when visiting its source block.
12350 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12353 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12354 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12356 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12358 PR tree-optimization/83957
12359 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12360 semicolon after for body surrounded by braces.
12362 PR tree-optimization/83081
12363 * profile-count.h (profile_probability::split): New method.
12364 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12365 Use profile_probability::split.
12366 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12367 when splitting a single conditional jump into 2.
12369 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12371 PR tree-optimization/69452
12372 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12375 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12377 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12378 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12379 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12381 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12383 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12385 * config/rl78/rl78.md (movdi): New define_expand.
12386 * config/rl78/rl78.c (rl78_split_movdi): New function.
12388 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12391 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12393 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12394 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12395 128-bit to produce an UNSPEC move to get the double word with the
12396 signbit and then a shift directly to do signbit.
12397 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12398 implementation with a new version that just does either a direct
12399 move or a regular move. Move memory interface to separate insns.
12400 Move insns so they are next to the expander.
12401 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12402 with signbit move. Split big and little endian case.
12403 (signbit<mode>2_dm_mem_le): Likewise.
12404 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12405 (signbit<mode>2_dm2): Likewise.
12407 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12409 * config/rl78/rl78.md (anddi3): New define_expand.
12411 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12413 * config/rl78/rl78.md (umindi3): New define_expand.
12415 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12417 * config/rl78/rl78.md (smindi3): New define_expand.
12419 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12421 * config/rl78/rl78.md (smaxdi3): New define_expand.
12423 2018-01-22 Carl Love <cel@us.ibm.com>
12425 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12426 LVX_V1TI): Add macro expansion.
12427 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12428 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12429 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12430 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12431 Change check to determine if the instruction is a byte reversing
12432 entry. Fix typo in comment.
12433 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12434 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12435 Add def_builtin calls for new builtins.
12436 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12437 Add define_insn expansion.
12439 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12441 * config/rl78/rl78.md (umaxdi3): New define_expand.
12443 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12445 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12446 for non-QImode registers.
12448 2018-01-22 Richard Biener <rguenther@suse.de>
12450 PR tree-optimization/83963
12451 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12452 including the loop exit block.
12453 (scop_detection::merge_sese): Likewise.
12454 (scop_detection::add_scop): Do it here instead.
12456 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12458 * doc/sourcebuild.texi (arm_softfloat): Document.
12460 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12463 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12464 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12465 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12467 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12468 David Edelsohn <dje.gcc@gmail.com>
12471 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12472 Change "crset eq" to "crset 2".
12473 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12474 (*call_indirect_aix<mode>_nospec): Likewise.
12475 (*call_value_indirect_aix<mode>_nospec): Likewise.
12476 (*call_indirect_elfv2<mode>_nospec): Likewise.
12477 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12478 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12479 change assembly output from . to $.
12480 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12481 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12482 (*tablejump<mode>_internal1_nospec): Likewise.
12484 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12487 * config/sh/sh_optimize_sett_clrt.cc:
12488 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12490 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12492 PR tree-optimization/83940
12493 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12494 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12495 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12496 use it to pass back the definition type.
12497 (vect_check_store_rhs): Likewise rhs_dt_out.
12498 (vect_build_gather_load_calls): Add a mask_dt argument and use
12499 it instead of a call to vect_is_simple_use.
12500 (vectorizable_store): Update calls to vect_check_load_store_mask
12501 and vect_check_store_rhs. Use the dt returned by the latter instead
12502 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12503 instead of calls to vect_is_simple_use. Pass the scalar rather
12504 than the vector operand to vect_is_simple_use when handling
12505 second and subsequent copies of an rhs value.
12506 (vectorizable_load): Update calls to vect_check_load_store_mask
12507 and vect_build_gather_load_calls. Use the cached mask_dt and
12508 gs_info.offset_dt instead of calls to vect_is_simple_use.
12510 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12512 PR middle-end/83945
12513 * tree-emutls.c: Include gimplify.h.
12514 (lower_emutls_2): New function.
12515 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12516 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12517 it before further processing.
12520 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12521 UINTVAL (trueop1) instead of INTVAL (op1).
12523 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12527 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12528 INCOMING_FRAME_SP_OFFSET if not defined.
12529 (scan_trace): Add ENTRY argument. If true and
12530 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12531 emit a note to adjust the CFA offset.
12532 (create_cfi_notes): Adjust scan_trace callers.
12533 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12534 INCOMING_FRAME_SP_OFFSET in the CIE.
12535 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12536 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12538 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12539 * doc/tm.texi: Regenerated.
12541 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12543 PR rtl-optimization/83147
12544 * lra-constraints.c (remove_inheritance_pseudos): Use
12545 lra_substitute_pseudo_within_insn.
12547 2018-01-19 Tom de Vries <tom@codesourcery.com>
12548 Cesar Philippidis <cesar@codesourcery.com>
12551 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12553 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12556 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12557 spaces for function labels.
12559 2018-01-19 Martin Liska <mliska@suse.cz>
12561 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12562 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12563 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12564 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12565 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12566 (PRED_CONST_RETURN): Change from 69 to 65.
12567 (PRED_NULL_RETURN): Change from 91 to 71.
12568 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12569 (PRED_LOOP_GUARD): Change from 66 to 73.
12571 2018-01-19 Martin Liska <mliska@suse.cz>
12573 * predict.c (predict_insn_def): Add new assert.
12574 (struct branch_predictor): Change type to signed integer.
12575 (test_prediction_value_range): Amend test to cover
12576 PROB_UNINITIALIZED.
12577 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12578 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12579 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12580 (PRED_LOOP_IV_COMPARE): Likewise.
12581 * predict.h (PROB_UNINITIALIZED): Define new constant.
12583 2018-01-19 Martin Liska <mliska@suse.cz>
12585 * predict.c (dump_prediction): Add new format for
12586 analyze_brprob.py script which is enabled with -details
12588 * profile-count.h (precise_p): New function.
12590 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12592 PR tree-optimization/83922
12593 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12594 there are no statements that need masking.
12595 (vect_active_double_reduction_p): New function.
12596 (vect_analyze_loop_operations): Use it when handling phis that
12597 are not in the loop header.
12599 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12601 PR tree-optimization/83914
12602 * tree-vect-loop.c (vectorizable_induction): Don't convert
12603 init_expr or apply the peeling adjustment for inductions
12604 that are nested within the vectorized loop.
12606 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12608 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12611 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12615 * function.h (gimplify_parameters): Add gimple_seq * argument.
12616 * function.c: Include gimple.h and options.h.
12617 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12618 for the added local temporaries if needed.
12619 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12620 if there are any parameter cleanups, wrap whole body into a
12621 try/finally with the cleanups.
12623 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12626 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12627 Use GET_MODE_CLASS for scalar floating point.
12629 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12633 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12634 Fix call of call_cgraph_insertion_hooks.
12636 2018-01-18 Martin Sebor <msebor@redhat.com>
12638 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12640 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12643 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12646 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12649 * common.opt: (-ffile-prefix-map): New option.
12650 * opts.c (common_handle_option): Defer it.
12651 * opts-global.c (handle_common_deferred_options): Handle it.
12652 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12653 * file-prefix-map.h: New file.
12654 (remap_debug_filename, add_debug_prefix_map): ...here.
12655 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12656 * final.c (debug_prefix_map, add_debug_prefix_map
12657 remap_debug_filename): Move to...
12658 * file-prefix-map.c: New file.
12659 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12660 generalize, get rid of alloca(), use strrchr() instead of strchr().
12661 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12662 Implement in terms of add_prefix_map().
12663 (remap_macro_filename, remap_debug_filename): Implement in term of
12665 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12666 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12667 * dbxout.c: Include file-prefix-map.h.
12668 * varasm.c: Likewise.
12669 * vmsdbgout.c: Likewise.
12670 * xcoffout.c: Likewise.
12671 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12672 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12673 * doc/invoke.texi (-ffile-prefix-map): Document.
12674 (-fdebug-prefix-map): Update description.
12676 2018-01-18 Martin Liska <mliska@suse.cz>
12678 * config/i386/i386.c (indirect_thunk_name): Document that also
12680 (output_indirect_thunk): Document why both instructions
12681 (pause and lfence) are generated.
12683 2018-01-18 Richard Biener <rguenther@suse.de>
12685 PR tree-optimization/83887
12686 * graphite-scop-detection.c
12687 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12688 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12689 (scop_detection::merge_sese): Re-implement with a flood-fill
12690 algorithm that properly finds a SESE region if it exists.
12692 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12695 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12696 pointer_diff optimizations use view_convert instead of convert.
12698 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12700 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12701 Generate different code for -mno-speculate-indirect-jumps.
12702 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12703 (*call_indirect_aix<mode>): Disable for
12704 -mno-speculate-indirect-jumps.
12705 (*call_indirect_aix<mode>_nospec): New define_insn.
12706 (*call_value_indirect_aix<mode>): Disable for
12707 -mno-speculate-indirect-jumps.
12708 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12709 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12710 -mno-speculate-indirect-jumps.
12711 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12713 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12715 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12716 long double type, set the flags for noting the default long double
12717 type, even if we don't pass or return a long double type.
12719 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12722 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12725 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12727 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12728 support for merge[hl].
12729 (fold_mergehl_helper): New helper function.
12730 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12731 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12732 (altivec_vmrglw_direct): Add xxmrglw insn.
12734 2018-01-17 Andrew Waterman <andrew@sifive.com>
12736 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12737 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12739 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12742 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12743 call the lto_location_cache before reading the
12744 DECL_SOURCE_LOCATION of the types.
12746 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12747 Richard Sandiford <richard.sandiford@linaro.org>
12749 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12750 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12751 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12752 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12753 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12755 * config/aarch64/constraints.md (aarch64_movti_operand):
12757 * config/aarch64/predicates.md (Uti): Add new constraint.
12759 2018-01-17 Carl Love <cel@us.ibm.com>
12761 * config/rs6000/vsx.md (define_expand xl_len_r,
12762 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12763 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12765 (define_expand, define_insn): Move the shift left from the
12766 define_insn to the define_expand for lxvl and stxvl instructions.
12767 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12768 and XL_LEN_R definitions to PURE.
12770 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12772 * config/i386/i386.c (indirect_thunk_name): Declare regno
12773 as unsigned int. Compare regno with INVALID_REGNUM.
12774 (output_indirect_thunk): Ditto.
12775 (output_indirect_thunk_function): Ditto.
12776 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12777 in the call to output_indirect_thunk_function.
12779 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12781 PR middle-end/83884
12782 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12783 rather than the size of inner_type to determine the stack slot size
12784 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12786 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12789 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12792 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12794 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12795 endian Linux systems to optionally enable multilibs for selecting
12796 the long double type if the user configured an explicit type.
12797 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12798 have no long double multilibs if not defined.
12799 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12800 warn if the user used -mabi={ieee,ibm}longdouble and we built
12801 multilibs for long double.
12802 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12803 appropriate multilib option.
12804 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12806 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12807 for building long double multilibs.
12808 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12810 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12812 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12815 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12817 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12820 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12823 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12826 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12828 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12829 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12831 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12833 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12834 different rtl trees depending on TARGET_64BIT.
12835 (rs6000_gen_lvx): Likewise.
12837 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12839 * config/visium/visium.md (nop): Tweak comment.
12840 (hazard_nop): Likewise.
12842 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12844 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12845 -mspeculate-indirect-jumps.
12846 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12847 for -mno-speculate-indirect-jumps.
12848 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12849 (*call_value_indirect_elfv2<mode>): Disable for
12850 -mno-speculate-indirect-jumps.
12851 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12852 (indirect_jump): Emit different RTL for
12853 -mno-speculate-indirect-jumps.
12854 (*indirect_jump<mode>): Disable for
12855 -mno-speculate-indirect-jumps.
12856 (*indirect_jump<mode>_nospec): New define_insn.
12857 (tablejump): Emit different RTL for
12858 -mno-speculate-indirect-jumps.
12859 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12860 (tablejumpsi_nospec): New define_expand.
12861 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12862 (tablejumpdi_nospec): New define_expand.
12863 (*tablejump<mode>_internal1): Disable for
12864 -mno-speculate-indirect-jumps.
12865 (*tablejump<mode>_internal1_nospec): New define_insn.
12866 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12869 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12871 * caller-save.c (insert_save): Drop unnecessary parameter. All
12874 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12875 Richard Biener <rguenth@suse.de>
12878 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12879 return early, inline manually is_gimple_sizepos. Make sure if we
12880 call gimplify_expr we don't end up with a gimple constant.
12881 * tree.c (variably_modified_type_p): Don't return true for
12882 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12883 * gimplify.h (is_gimple_sizepos): Remove.
12885 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12887 PR tree-optimization/83857
12888 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12889 vectorizable_live_operation for pure SLP statements.
12890 (vectorizable_live_operation): Handle PHIs.
12892 2018-01-16 Richard Biener <rguenther@suse.de>
12894 PR tree-optimization/83867
12895 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12896 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12898 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12901 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12902 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12903 If off is not INTEGER_CST, issue a may not be aligned warning
12904 rather than isn't aligned. Use isn%'t rather than isn't.
12905 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12907 <case MULT_EXPR>: Improve the case when bottom and one of the
12908 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12909 operand, in that case check if the other operand is multiple of
12910 bottom divided by the INTEGER_CST operand.
12912 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12915 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12916 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12917 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12918 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12919 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12920 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12921 (pa_function_arg_size): New function.
12923 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12925 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12926 in a separate statement.
12928 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12930 PR tree-optimization/83847
12931 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12932 group gathers and scatters.
12934 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12936 PR rtl-optimization/86620
12937 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12939 PR rtl-optimization/83213
12940 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12941 to last if both are JUMP_INSNs.
12943 PR tree-optimization/83843
12944 * gimple-ssa-store-merging.c
12945 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12946 store_immediate_info for bswap/nop orig_stores.
12948 2018-01-15 Andrew Waterman <andrew@sifive.com>
12950 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12952 <UDIV>: Increase cost if !TARGET_DIV.
12954 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12956 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12957 (define_attr "cr_logical_3op"): New.
12958 (cceq_ior_compare): Adjust.
12959 (cceq_ior_compare_complement): Adjust.
12960 (*cceq_rev_compare): Adjust.
12961 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12962 (is_cracked_insn): Adjust.
12963 (insn_must_be_first_in_group): Adjust.
12964 * config/rs6000/40x.md: Adjust.
12965 * config/rs6000/440.md: Adjust.
12966 * config/rs6000/476.md: Adjust.
12967 * config/rs6000/601.md: Adjust.
12968 * config/rs6000/603.md: Adjust.
12969 * config/rs6000/6xx.md: Adjust.
12970 * config/rs6000/7450.md: Adjust.
12971 * config/rs6000/7xx.md: Adjust.
12972 * config/rs6000/8540.md: Adjust.
12973 * config/rs6000/cell.md: Adjust.
12974 * config/rs6000/e300c2c3.md: Adjust.
12975 * config/rs6000/e500mc.md: Adjust.
12976 * config/rs6000/e500mc64.md: Adjust.
12977 * config/rs6000/e5500.md: Adjust.
12978 * config/rs6000/e6500.md: Adjust.
12979 * config/rs6000/mpc.md: Adjust.
12980 * config/rs6000/power4.md: Adjust.
12981 * config/rs6000/power5.md: Adjust.
12982 * config/rs6000/power6.md: Adjust.
12983 * config/rs6000/power7.md: Adjust.
12984 * config/rs6000/power8.md: Adjust.
12985 * config/rs6000/power9.md: Adjust.
12986 * config/rs6000/rs64.md: Adjust.
12987 * config/rs6000/titan.md: Adjust.
12989 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12991 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12992 ix86_indirect_branch_register logic.
12994 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12996 * config/i386/constraints.md (Bs): Update
12997 ix86_indirect_branch_register check. Don't check
12998 ix86_indirect_branch_register with GOT_memory_operand.
13000 * config/i386/predicates.md (GOT_memory_operand): Don't check
13001 ix86_indirect_branch_register here.
13002 (GOT32_symbol_operand): Likewise.
13004 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13006 * config/i386/predicates.md (constant_call_address_operand):
13007 Rewrite ix86_indirect_branch_register logic.
13008 (sibcall_insn_operand): Likewise.
13010 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13012 * config/i386/constraints.md (Bs): Replace
13013 ix86_indirect_branch_thunk_register with
13014 ix86_indirect_branch_register.
13016 * config/i386/i386.md (indirect_jump): Likewise.
13017 (tablejump): Likewise.
13018 (*sibcall_memory): Likewise.
13019 (*sibcall_value_memory): Likewise.
13020 Peepholes of indirect call and jump via memory: Likewise.
13021 * config/i386/i386.opt: Likewise.
13022 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13023 (GOT_memory_operand): Likewise.
13024 (call_insn_operand): Likewise.
13025 (sibcall_insn_operand): Likewise.
13026 (GOT32_symbol_operand): Likewise.
13028 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13030 PR middle-end/83837
13031 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13032 type rather than type addr's type points to.
13033 (expand_omp_atomic_mutex): Likewise.
13034 (expand_omp_atomic): Likewise.
13036 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13039 * config/i386/i386.c (output_indirect_thunk_function): Use
13040 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13041 for __x86_return_thunk.
13043 2018-01-15 Richard Biener <rguenther@suse.de>
13045 PR middle-end/83850
13046 * expmed.c (extract_bit_field_1): Fix typo.
13048 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13051 * config/arm/iterators.md (VF): New mode iterator.
13052 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13053 Remove integer-related logic from pattern.
13054 (neon_vabd<mode>_3): Likewise.
13056 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13058 PR middle-end/82694
13059 * common.opt (fstrict-overflow): No longer an alias.
13060 (fwrapv-pointer): New option.
13061 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13062 also for pointer types based on flag_wrapv_pointer.
13063 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13064 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13065 opts->x_flag_wrapv got set.
13066 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13067 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13068 POINTER_TYPE_OVERFLOW_UNDEFINED.
13069 * match.pd: Likewise in address comparison pattern.
13070 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13072 2018-01-15 Richard Biener <rguenther@suse.de>
13075 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13076 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13077 Reset type names to their identifier if their TYPE_DECL doesn't
13078 have linkage (and thus is used for ODR and devirt).
13079 (save_debug_info_for_decl): Remove.
13080 (save_debug_info_for_type): Likewise.
13081 (add_tree_to_fld_list): Adjust.
13082 * tree-pretty-print.c (dump_generic_node): Make dumping of
13083 type names more robust.
13085 2018-01-15 Richard Biener <rguenther@suse.de>
13087 * BASE-VER: Bump to 8.0.1.
13089 2018-01-14 Martin Sebor <msebor@redhat.com>
13092 * builtins.c (check_access): Avoid warning when the no-warning bit
13095 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13097 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13098 * ira-color (allocno_hard_regs_compare): Likewise.
13100 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13103 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13104 Use .pushsection/.popsection.
13106 2018-01-14 Martin Sebor <msebor@redhat.com>
13109 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13111 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13113 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13114 entry from extra_headers.
13115 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13116 extra_headers, make the list bitwise identical to the i?86-*-* one.
13118 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13120 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13121 -mcmodel=large with -mindirect-branch=thunk,
13122 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13123 -mfunction-return=thunk-extern.
13124 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13125 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13126 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13128 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13130 * config/i386/i386.c (print_reg): Print the name of the full
13131 integer register without '%'.
13132 (ix86_print_operand): Handle 'V'.
13133 * doc/extend.texi: Document 'V' modifier.
13135 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13137 * config/i386/constraints.md (Bs): Disallow memory operand for
13138 -mindirect-branch-register.
13140 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13141 (GOT_memory_operand): Likewise.
13142 (call_insn_operand): Likewise.
13143 (sibcall_insn_operand): Likewise.
13144 (GOT32_symbol_operand): Likewise.
13145 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13146 for -mindirect-branch-register.
13147 (tablejump): Likewise.
13148 (*sibcall_memory): Likewise.
13149 (*sibcall_value_memory): Likewise.
13150 Disallow peepholes of indirect call and jump via memory for
13151 -mindirect-branch-register.
13152 (*call_pop): Replace m with Bw.
13153 (*call_value_pop): Likewise.
13154 (*sibcall_pop_memory): Replace m with Bs.
13155 * config/i386/i386.opt (mindirect-branch-register): New option.
13156 * doc/invoke.texi: Document -mindirect-branch-register option.
13158 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13160 * config/i386/i386-protos.h (ix86_output_function_return): New.
13161 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13162 set function_return_type.
13163 (indirect_thunk_name): Add ret_p to indicate thunk for function
13165 (output_indirect_thunk_function): Pass false to
13166 indirect_thunk_name.
13167 (ix86_output_indirect_branch_via_reg): Likewise.
13168 (ix86_output_indirect_branch_via_push): Likewise.
13169 (output_indirect_thunk_function): Create alias for function
13170 return thunk if regno < 0.
13171 (ix86_output_function_return): New function.
13172 (ix86_handle_fndecl_attribute): Handle function_return.
13173 (ix86_attribute_table): Add function_return.
13174 * config/i386/i386.h (machine_function): Add
13175 function_return_type.
13176 * config/i386/i386.md (simple_return_internal): Use
13177 ix86_output_function_return.
13178 (simple_return_internal_long): Likewise.
13179 * config/i386/i386.opt (mfunction-return=): New option.
13180 (indirect_branch): Mention -mfunction-return=.
13181 * doc/extend.texi: Document function_return function attribute.
13182 * doc/invoke.texi: Document -mfunction-return= option.
13184 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13186 * config/i386/i386-opts.h (indirect_branch): New.
13187 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13188 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13189 with local indirect jump when converting indirect call and jump.
13190 (ix86_set_indirect_branch_type): New.
13191 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13192 (indirectlabelno): New.
13193 (indirect_thunk_needed): Likewise.
13194 (indirect_thunk_bnd_needed): Likewise.
13195 (indirect_thunks_used): Likewise.
13196 (indirect_thunks_bnd_used): Likewise.
13197 (INDIRECT_LABEL): Likewise.
13198 (indirect_thunk_name): Likewise.
13199 (output_indirect_thunk): Likewise.
13200 (output_indirect_thunk_function): Likewise.
13201 (ix86_output_indirect_branch_via_reg): Likewise.
13202 (ix86_output_indirect_branch_via_push): Likewise.
13203 (ix86_output_indirect_branch): Likewise.
13204 (ix86_output_indirect_jmp): Likewise.
13205 (ix86_code_end): Call output_indirect_thunk_function if needed.
13206 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13208 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13209 (ix86_attribute_table): Add indirect_branch.
13210 * config/i386/i386.h (machine_function): Add indirect_branch_type
13211 and has_local_indirect_jump.
13212 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13214 (tablejump): Likewise.
13215 (*indirect_jump): Use ix86_output_indirect_jmp.
13216 (*tablejump_1): Likewise.
13217 (simple_return_indirect_internal): Likewise.
13218 * config/i386/i386.opt (mindirect-branch=): New option.
13219 (indirect_branch): New.
13222 (thunk-inline): Likewise.
13223 (thunk-extern): Likewise.
13224 * doc/extend.texi: Document indirect_branch function attribute.
13225 * doc/invoke.texi: Document -mindirect-branch= option.
13227 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13230 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13232 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13234 * ipa-inline.c (want_inline_small_function_p): Return false if
13235 inlining has already failed with CIF_FINAL_ERROR.
13236 (update_caller_keys): Call want_inline_small_function_p before
13238 (update_callee_keys): Likewise.
13240 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13242 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13244 (rs6000_quadword_masked_address_p): Likewise.
13245 (quad_aligned_load_p): Likewise.
13246 (quad_aligned_store_p): Likewise.
13247 (const_load_sequence_p): Add comment to describe the outer-most loop.
13248 (mimic_memory_attributes_and_flags): New function.
13249 (rs6000_gen_stvx): Likewise.
13250 (replace_swapped_aligned_store): Likewise.
13251 (rs6000_gen_lvx): Likewise.
13252 (replace_swapped_aligned_load): Likewise.
13253 (replace_swapped_load_constant): Capitalize argument name in
13254 comment describing this function.
13255 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13256 and stores that access quad-word aligned addresses and replace
13257 with stvx or lvx instructions when appropriate.
13258 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13259 New function prototype.
13260 (rs6000_quadword_masked_address_p): Likewise.
13261 (rs6000_gen_lvx): Likewise.
13262 (rs6000_gen_stvx): Likewise.
13263 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13264 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13265 when memory address is aligned.
13266 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13267 this split to select lvx instruction when memory address is aligned.
13268 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13269 instruction when memory address is aligned.
13270 (*vsx_le_perm_load_v16qi): Likewise.
13271 (four unnamed splitters): Modify to select the stvx instruction
13272 when memory is aligned.
13274 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13276 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13277 which appears in the queue multiple times.
13279 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13280 Alan Hayward <alan.hayward@arm.com>
13281 David Sherwood <david.sherwood@arm.com>
13283 * tree-vectorizer.h (vec_lower_bound): New structure.
13284 (_loop_vec_info): Add check_nonzero and lower_bounds.
13285 (LOOP_VINFO_CHECK_NONZERO): New macro.
13286 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13287 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13288 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13289 fields. Make seg_len the distance travelled, not including the
13291 (dr_direction_indicator): Declare.
13292 (dr_zero_step_indicator): Likewise.
13293 (dr_known_forward_stride_p): Likewise.
13294 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13296 (runtime_alias_check_p): Allow runtime alias checks with
13298 (operator ==): Compare access_size and align.
13299 (prune_runtime_alias_test_list): Rework for new distinction between
13300 the access_size and seg_len.
13301 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13303 (get_segment_min_max): New function.
13304 (create_intersect_range_checks): Use it.
13305 (dr_step_indicator): New function.
13306 (dr_direction_indicator): Likewise.
13307 (dr_zero_step_indicator): Likewise.
13308 (dr_known_forward_stride_p): Likewise.
13309 * tree-loop-distribution.c (data_ref_segment_size): Return
13310 DR_STEP * (niters - 1).
13311 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13313 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13314 (vect_preserves_scalar_order_p): New function, split out from...
13315 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13316 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13317 (vect_vfa_access_size): New function.
13318 (vect_vfa_align): Likewise.
13319 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13320 (dump_lower_bound): New function.
13321 (vect_check_lower_bound): Likewise.
13322 (vect_small_gap_p): Likewise.
13323 (vectorizable_with_step_bound_p): Likewise.
13324 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13325 depencies if the vectorization factor is 1. Convert the checks
13326 for nonzero steps into checks on the bounds of DR_STEP. Try using
13327 a bunds check for variable steps if the minimum required step is
13328 relatively small. Update calls to the dr_with_seg_len
13329 constructor and to vect_compile_time_alias.
13330 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13332 (vect_loop_versioning): Call it.
13333 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13335 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13338 Alan Hayward <alan.hayward@arm.com>
13339 David Sherwood <david.sherwood@arm.com>
13341 * doc/sourcebuild.texi (vect_scatter_store): Document.
13342 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13344 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13346 * genopinit.c (main): Add supports_vec_scatter_store and
13347 supports_vec_scatter_store_cached to target_optabs.
13348 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13349 IFN_MASK_SCATTER_STORE.
13350 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13352 * internal-fn.h (internal_store_fn_p): Declare.
13353 (internal_fn_stored_value_index): Likewise.
13354 * internal-fn.c (scatter_store_direct): New macro.
13355 (expand_scatter_store_optab_fn): New function.
13356 (direct_scatter_store_optab_supported_p): New macro.
13357 (internal_store_fn_p): New function.
13358 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13359 IFN_MASK_SCATTER_STORE.
13360 (internal_fn_mask_index): Likewise.
13361 (internal_fn_stored_value_index): New function.
13362 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13363 for scatter stores.
13364 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13365 * optabs-query.c (supports_vec_scatter_store_p): New function.
13366 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13367 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13368 true for scatter stores.
13369 (vect_gather_scatter_fn_p): Handle scatter stores too.
13370 (vect_check_gather_scatter): Consider using scatter stores if
13371 supports_vec_scatter_store_p.
13372 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13373 scatter stores too.
13374 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13375 internal_fn_stored_value_index.
13376 (check_load_store_masking): Handle scatter stores too.
13377 (vect_get_store_rhs): Make public.
13378 (vectorizable_call): Use internal_store_fn_p.
13379 (vectorizable_store): Handle scatter store internal functions.
13380 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13381 when deciding whether the end of the group has been reached.
13382 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13383 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13384 (mask_scatter_store<mode>): New insns.
13386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13387 Alan Hayward <alan.hayward@arm.com>
13388 David Sherwood <david.sherwood@arm.com>
13390 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13391 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13392 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13394 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13395 Use vect_truncate_gather_scatter_offset if we can't treat the
13396 operation as a normal gather load or scatter store.
13397 (get_group_load_store_type): Take the gather_scatter_info
13398 as argument. Try using a gather load or scatter store for
13399 single-element groups.
13400 (get_load_store_type): Update calls to get_group_load_store_type
13401 and vect_use_strided_gather_scatters_p.
13403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13404 Alan Hayward <alan.hayward@arm.com>
13405 David Sherwood <david.sherwood@arm.com>
13407 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13408 optional tree argument.
13409 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13411 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13412 but continue to use the current value as a fallback.
13413 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13414 to compare the updates.
13415 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13416 (get_load_store_type): Use it when handling a strided access.
13417 (vect_get_strided_load_store_ops): New function.
13418 (vect_get_data_ptr_increment): Likewise.
13419 (vectorizable_load): Handle strided gather loads. Always pass
13420 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13423 Alan Hayward <alan.hayward@arm.com>
13424 David Sherwood <david.sherwood@arm.com>
13426 * doc/md.texi (gather_load@var{m}): Document.
13427 (mask_gather_load@var{m}): Likewise.
13428 * genopinit.c (main): Add supports_vec_gather_load and
13429 supports_vec_gather_load_cached to target_optabs.
13430 * optabs-tree.c (init_tree_optimization_optabs): Use
13431 ggc_cleared_alloc to allocate target_optabs.
13432 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13433 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13435 * internal-fn.h (internal_load_fn_p): Declare.
13436 (internal_gather_scatter_fn_p): Likewise.
13437 (internal_fn_mask_index): Likewise.
13438 (internal_gather_scatter_fn_supported_p): Likewise.
13439 * internal-fn.c (gather_load_direct): New macro.
13440 (expand_gather_load_optab_fn): New function.
13441 (direct_gather_load_optab_supported_p): New macro.
13442 (direct_internal_fn_optab): New function.
13443 (internal_load_fn_p): Likewise.
13444 (internal_gather_scatter_fn_p): Likewise.
13445 (internal_fn_mask_index): Likewise.
13446 (internal_gather_scatter_fn_supported_p): Likewise.
13447 * optabs-query.c (supports_at_least_one_mode_p): New function.
13448 (supports_vec_gather_load_p): Likewise.
13449 * optabs-query.h (supports_vec_gather_load_p): Declare.
13450 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13451 and memory_type field.
13452 (NUM_PATTERNS): Bump to 15.
13453 * tree-vect-data-refs.c: Include internal-fn.h.
13454 (vect_gather_scatter_fn_p): New function.
13455 (vect_describe_gather_scatter_call): Likewise.
13456 (vect_check_gather_scatter): Try using internal functions for
13457 gather loads. Recognize existing calls to a gather load function.
13458 (vect_analyze_data_refs): Consider using gather loads if
13459 supports_vec_gather_load_p.
13460 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13461 (vect_get_gather_scatter_offset_type): Likewise.
13462 (vect_convert_mask_for_vectype): Likewise.
13463 (vect_add_conversion_to_patterm): Likewise.
13464 (vect_try_gather_scatter_pattern): Likewise.
13465 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13466 (vect_vect_recog_func_ptrs): Add it.
13467 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13468 internal_fn_mask_index and internal_gather_scatter_fn_p.
13469 (check_load_store_masking): Take the gather_scatter_info as an
13470 argument and handle gather loads.
13471 (vect_get_gather_scatter_ops): New function.
13472 (vectorizable_call): Check internal_load_fn_p.
13473 (vectorizable_load): Likewise. Handle gather load internal
13475 (vectorizable_store): Update call to check_load_store_masking.
13476 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13477 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13478 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13479 (aarch64_gather_scale_operand_d): New predicates.
13480 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13481 (mask_gather_load<mode>): New insns.
13483 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13484 Alan Hayward <alan.hayward@arm.com>
13485 David Sherwood <david.sherwood@arm.com>
13487 * optabs.def (fold_left_plus_optab): New optab.
13488 * doc/md.texi (fold_left_plus_@var{m}): Document.
13489 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13490 * internal-fn.c (fold_left_direct): Define.
13491 (expand_fold_left_optab_fn): Likewise.
13492 (direct_fold_left_optab_supported_p): Likewise.
13493 * fold-const-call.c (fold_const_fold_left): New function.
13494 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13495 * tree-parloops.c (valid_reduction_p): New function.
13496 (gather_scalar_reductions): Use it.
13497 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13498 (vect_finish_replace_stmt): Declare.
13499 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13500 (needs_fold_left_reduction_p): New function, split out from...
13501 (vect_is_simple_reduction): ...here. Accept reductions that
13502 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13503 (vect_force_simple_reduction): Also store the reduction type in
13504 the assignment's STMT_VINFO_REDUC_TYPE.
13505 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13506 (merge_with_identity): New function.
13507 (vect_expand_fold_left): Likewise.
13508 (vectorize_fold_left_reduction): Likewise.
13509 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13510 scalar phi in place for it. Check for target support and reject
13511 cases that would reassociate the operation. Defer the transform
13512 phase to vectorize_fold_left_reduction.
13513 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13514 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13515 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13519 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13520 call to ifc_temp_var.
13522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13523 Alan Hayward <alan.hayward@arm.com>
13524 David Sherwood <david.sherwood@arm.com>
13526 * target.def (legitimize_address_displacement): Take the original
13527 offset as a poly_int.
13528 * targhooks.h (default_legitimize_address_displacement): Update
13530 * targhooks.c (default_legitimize_address_displacement): Likewise.
13531 * doc/tm.texi: Regenerate.
13532 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13533 as an argument, moving assert of ad->disp == ad->disp_term to...
13534 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13535 Try calling targetm.legitimize_address_displacement before expanding
13536 the address rather than afterwards, and adjust for the new interface.
13537 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13538 Match the new hook interface. Handle SVE addresses.
13539 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13540 new hook interface.
13542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13544 * Makefile.in (OBJS): Add early-remat.o.
13545 * target.def (select_early_remat_modes): New hook.
13546 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13547 * doc/tm.texi: Regenerate.
13548 * targhooks.h (default_select_early_remat_modes): Declare.
13549 * targhooks.c (default_select_early_remat_modes): New function.
13550 * timevar.def (TV_EARLY_REMAT): New timevar.
13551 * passes.def (pass_early_remat): New pass.
13552 * tree-pass.h (make_pass_early_remat): Declare.
13553 * early-remat.c: New file.
13554 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13556 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13559 Alan Hayward <alan.hayward@arm.com>
13560 David Sherwood <david.sherwood@arm.com>
13562 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13563 vfm1 with a bound_epilog parameter.
13564 (vect_do_peeling): Update calls accordingly, and move the prologue
13565 call earlier in the function. Treat the base bound_epilog as 0 for
13566 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13567 this base when peeling for gaps.
13568 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13569 with fully-masked loops.
13570 (vect_estimate_min_profitable_iters): Handle the single peeled
13571 iteration in that case.
13573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13574 Alan Hayward <alan.hayward@arm.com>
13575 David Sherwood <david.sherwood@arm.com>
13577 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13578 single-element interleaving even if the size is not a power of 2.
13579 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13580 accesses for single-element interleaving if the group size is
13583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13584 Alan Hayward <alan.hayward@arm.com>
13585 David Sherwood <david.sherwood@arm.com>
13587 * doc/md.texi (fold_extract_last_@var{m}): Document.
13588 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13589 * optabs.def (fold_extract_last_optab): New optab.
13590 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13591 * internal-fn.c (fold_extract_direct): New macro.
13592 (expand_fold_extract_optab_fn): Likewise.
13593 (direct_fold_extract_optab_supported_p): Likewise.
13594 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13595 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13596 EXTRACT_LAST_REDUCTION.
13597 (get_initial_def_for_reduction): Do not create an initial vector
13598 for EXTRACT_LAST_REDUCTION reductions.
13599 (vectorizable_reduction): Leave the scalar phi in place for
13600 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13601 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13602 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13603 transform phase to vectorizable_condition.
13604 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13606 (vect_finish_stmt_generation): ...here.
13607 (vect_finish_replace_stmt): New function.
13608 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13609 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13611 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13614 Alan Hayward <alan.hayward@arm.com>
13615 David Sherwood <david.sherwood@arm.com>
13617 * doc/md.texi (extract_last_@var{m}): Document.
13618 * optabs.def (extract_last_optab): New optab.
13619 * internal-fn.def (EXTRACT_LAST): New internal function.
13620 * internal-fn.c (cond_unary_direct): New macro.
13621 (expand_cond_unary_optab_fn): Likewise.
13622 (direct_cond_unary_optab_supported_p): Likewise.
13623 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13624 loops using EXTRACT_LAST.
13625 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13626 (extract_last_<mode>): ...this optab.
13627 (vec_extract<mode><Vel>): Update accordingly.
13629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13630 Alan Hayward <alan.hayward@arm.com>
13631 David Sherwood <david.sherwood@arm.com>
13633 * target.def (empty_mask_is_expensive): New hook.
13634 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13635 * doc/tm.texi: Regenerate.
13636 * targhooks.h (default_empty_mask_is_expensive): Declare.
13637 * targhooks.c (default_empty_mask_is_expensive): New function.
13638 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13639 if the target says that empty masks are expensive.
13640 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13642 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13645 Alan Hayward <alan.hayward@arm.com>
13646 David Sherwood <david.sherwood@arm.com>
13648 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13649 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13650 (vect_use_loop_mask_for_alignment_p): New function.
13651 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13652 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13653 niters_skip argument. Make sure that the first niters_skip elements
13654 of the first iteration are inactive.
13655 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13656 Update call to vect_set_loop_masks_directly.
13657 (get_misalign_in_elems): New function, split out from...
13658 (vect_gen_prolog_loop_niters): ...here.
13659 (vect_update_init_of_dr): Take a code argument that specifies whether
13660 the adjustment should be added or subtracted.
13661 (vect_update_init_of_drs): Likewise.
13662 (vect_prepare_for_masked_peels): New function.
13663 (vect_do_peeling): Skip prologue peeling if we're using a mask
13664 instead. Update call to vect_update_inits_of_drs.
13665 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13667 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13668 alignment. Do not include the number of peeled iterations in
13669 the minimum threshold in that case.
13670 (vectorizable_induction): Adjust the start value down by
13671 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13672 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13673 Take the number of skipped iterations into account when calculating
13675 * tree-vect-stmts.c (vect_gen_while_not): New function.
13677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13678 Alan Hayward <alan.hayward@arm.com>
13679 David Sherwood <david.sherwood@arm.com>
13681 * doc/sourcebuild.texi (vect_fully_masked): Document.
13682 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13683 default value to 0.
13684 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13686 (vect_analyze_loop_2): ...here. Don't check the vectorization
13687 factor against the number of loop iterations if the loop is
13690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13691 Alan Hayward <alan.hayward@arm.com>
13692 David Sherwood <david.sherwood@arm.com>
13694 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13695 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13696 (dump_groups): Update accordingly.
13697 (iv_use::mem_type): New member variable.
13698 (address_p): New function.
13699 (record_use): Add a mem_type argument and initialize the new
13701 (record_group_use): Add a mem_type argument. Use address_p.
13702 Remove obsolete null checks of base_object. Update call to record_use.
13703 (find_interesting_uses_op): Update call to record_group_use.
13704 (find_interesting_uses_cond): Likewise.
13705 (find_interesting_uses_address): Likewise.
13706 (get_mem_type_for_internal_fn): New function.
13707 (find_address_like_use): Likewise.
13708 (find_interesting_uses_stmt): Try find_address_like_use before
13709 calling find_interesting_uses_op.
13710 (addr_offset_valid_p): Use the iv mem_type field as the type
13711 of the addressed memory.
13712 (add_autoinc_candidates): Likewise.
13713 (get_address_cost): Likewise.
13714 (split_small_address_groups_p): Use address_p.
13715 (split_address_groups): Likewise.
13716 (add_iv_candidate_for_use): Likewise.
13717 (autoinc_possible_for_pair): Likewise.
13718 (rewrite_groups): Likewise.
13719 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13720 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13721 (get_alias_ptr_type_for_ptr_address): New function.
13722 (rewrite_use_address): Rewrite address uses in calls that were
13723 identified by find_address_like_use.
13725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13726 Alan Hayward <alan.hayward@arm.com>
13727 David Sherwood <david.sherwood@arm.com>
13729 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13731 * gimple-expr.h (is_gimple_addressable: Likewise.
13732 * gimple-expr.c (is_gimple_address): Likewise.
13733 * internal-fn.c (expand_call_mem_ref): New function.
13734 (expand_mask_load_optab_fn): Use it.
13735 (expand_mask_store_optab_fn): Likewise.
13737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13738 Alan Hayward <alan.hayward@arm.com>
13739 David Sherwood <david.sherwood@arm.com>
13741 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13742 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13743 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13744 (cond_umax@var{mode}): Document.
13745 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13746 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13747 (cond_umin_optab, cond_umax_optab): New optabs.
13748 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13749 (COND_IOR, COND_XOR): New internal functions.
13750 * internal-fn.h (get_conditional_internal_fn): Declare.
13751 * internal-fn.c (cond_binary_direct): New macro.
13752 (expand_cond_binary_optab_fn): Likewise.
13753 (direct_cond_binary_optab_supported_p): Likewise.
13754 (get_conditional_internal_fn): New function.
13755 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13756 Cope with reduction statements that are vectorized as calls rather
13758 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13759 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13760 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13761 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13762 (UNSPEC_COND_EOR): New unspecs.
13763 (optab): Add mappings for them.
13764 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13765 (sve_int_op, sve_fp_op): New int attributes.
13767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13768 Alan Hayward <alan.hayward@arm.com>
13769 David Sherwood <david.sherwood@arm.com>
13771 * optabs.def (while_ult_optab): New optab.
13772 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13773 * internal-fn.def (WHILE_ULT): New internal function.
13774 * internal-fn.h (direct_internal_fn_supported_p): New override
13775 that takes two types as argument.
13776 * internal-fn.c (while_direct): New macro.
13777 (expand_while_optab_fn): New function.
13778 (convert_optab_supported_p): Likewise.
13779 (direct_while_optab_supported_p): New macro.
13780 * wide-int.h (wi::udiv_ceil): New function.
13781 * tree-vectorizer.h (rgroup_masks): New structure.
13782 (vec_loop_masks): New typedef.
13783 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13784 and fully_masked_p.
13785 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13786 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13787 (vect_max_vf): New function.
13788 (slpeel_make_loop_iterate_ntimes): Delete.
13789 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13790 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13791 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13792 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13793 internal-fn.h, stor-layout.h and optabs-query.h.
13794 (vect_set_loop_mask): New function.
13795 (add_preheader_seq): Likewise.
13796 (add_header_seq): Likewise.
13797 (interleave_supported_p): Likewise.
13798 (vect_maybe_permute_loop_masks): Likewise.
13799 (vect_set_loop_masks_directly): Likewise.
13800 (vect_set_loop_condition_masked): Likewise.
13801 (vect_set_loop_condition_unmasked): New function, split out from
13802 slpeel_make_loop_iterate_ntimes.
13803 (slpeel_make_loop_iterate_ntimes): Rename to..
13804 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13805 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13806 (vect_do_peeling): Update call accordingly.
13807 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13809 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13810 mask_compare_type, can_fully_mask_p and fully_masked_p.
13811 (release_vec_loop_masks): New function.
13812 (_loop_vec_info): Use it to free the loop masks.
13813 (can_produce_all_loop_masks_p): New function.
13814 (vect_get_max_nscalars_per_iter): Likewise.
13815 (vect_verify_full_masking): Likewise.
13816 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13817 retries, and free the mask rgroups before retrying. Check loop-wide
13818 reasons for disallowing fully-masked loops. Make the final decision
13819 about whether use a fully-masked loop or not.
13820 (vect_estimate_min_profitable_iters): Do not assume that peeling
13821 for the number of iterations will be needed for fully-masked loops.
13822 (vectorizable_reduction): Disable fully-masked loops.
13823 (vectorizable_live_operation): Likewise.
13824 (vect_halve_mask_nunits): New function.
13825 (vect_double_mask_nunits): Likewise.
13826 (vect_record_loop_mask): Likewise.
13827 (vect_get_loop_mask): Likewise.
13828 (vect_transform_loop): Handle the case in which the final loop
13829 iteration might handle a partial vector. Call vect_set_loop_condition
13830 instead of slpeel_make_loop_iterate_ntimes.
13831 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13832 (check_load_store_masking): New function.
13833 (prepare_load_store_mask): Likewise.
13834 (vectorizable_store): Handle fully-masked loops.
13835 (vectorizable_load): Likewise.
13836 (supportable_widening_operation): Use vect_halve_mask_nunits for
13838 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13839 (vect_gen_while): New function.
13840 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13841 (aarch64_uqdec<mode>): New insn.
13843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13844 Alan Hayward <alan.hayward@arm.com>
13845 David Sherwood <david.sherwood@arm.com>
13847 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13848 (reduc_xor_scal_optab): New optabs.
13849 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13850 (reduc_xor_scal_@var{m}): Document.
13851 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13852 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13853 internal functions.
13854 * fold-const-call.c (fold_const_call): Handle them.
13855 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13856 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13857 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13858 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13859 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13860 (UNSPEC_XORV): New unspecs.
13861 (optab): Add entries for them.
13862 (BITWISEV): New int iterator.
13863 (bit_reduc_op): New int attributes.
13865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13866 Alan Hayward <alan.hayward@arm.com>
13867 David Sherwood <david.sherwood@arm.com>
13869 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13870 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13871 * optabs.def (vec_shl_insert_optab): New optab.
13872 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13873 (duplicate_and_interleave): Likewise.
13874 * tree-vect-loop.c: Include internal-fn.h.
13875 (neutral_op_for_slp_reduction): New function, split out from
13876 get_initial_defs_for_reduction.
13877 (get_initial_def_for_reduction): Handle option 2 for variable-length
13878 vectors by loading the neutral value into a vector and then shifting
13879 the initial value into element 0.
13880 (get_initial_defs_for_reduction): Replace the code argument with
13881 the neutral value calculated by neutral_op_for_slp_reduction.
13882 Use gimple_build_vector for constant-length vectors.
13883 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13884 but the first group_size elements have a neutral value.
13885 Use duplicate_and_interleave otherwise.
13886 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13887 Update call to get_initial_defs_for_reduction. Handle SLP
13888 reductions for variable-length vectors by creating one vector
13889 result for each scalar result, with the elements associated
13890 with other scalar results stubbed out with the neutral value.
13891 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13892 Require IFN_VEC_SHL_INSERT for double reductions on
13893 variable-length vectors, or SLP reductions that have
13894 a neutral value. Require can_duplicate_and_interleave_p
13895 support for variable-length unchained SLP reductions if there
13896 is no neutral value, such as for MIN/MAX reductions. Also require
13897 the number of vector elements to be a multiple of the number of
13898 SLP statements when doing variable-length unchained SLP reductions.
13899 Update call to vect_create_epilog_for_reduction.
13900 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13901 and remove initial values.
13902 (duplicate_and_interleave): Make public.
13903 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13904 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13907 Alan Hayward <alan.hayward@arm.com>
13908 David Sherwood <david.sherwood@arm.com>
13910 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13911 (can_duplicate_and_interleave_p): New function.
13912 (vect_get_and_check_slp_defs): Take the vector of statements
13913 rather than just the current one. Remove excess parentheses.
13914 Restriction rejectinon of vect_constant_def and vect_external_def
13915 for variable-length vectors to boolean types, or types for which
13916 can_duplicate_and_interleave_p is false.
13917 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13918 (duplicate_and_interleave): New function.
13919 (vect_get_constant_vectors): Use gimple_build_vector for
13920 constant-length vectors and suitable variable-length constant
13921 vectors. Use duplicate_and_interleave for other variable-length
13922 vectors. Don't defer the update when inserting new statements.
13924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13925 Alan Hayward <alan.hayward@arm.com>
13926 David Sherwood <david.sherwood@arm.com>
13928 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13929 min_profitable_iters doesn't go negative.
13931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13932 Alan Hayward <alan.hayward@arm.com>
13933 David Sherwood <david.sherwood@arm.com>
13935 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13936 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13937 * optabs.def (vec_mask_load_lanes_optab): New optab.
13938 (vec_mask_store_lanes_optab): Likewise.
13939 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13940 (MASK_STORE_LANES): Likewise.
13941 * internal-fn.c (mask_load_lanes_direct): New macro.
13942 (mask_store_lanes_direct): Likewise.
13943 (expand_mask_load_optab_fn): Handle masked operations.
13944 (expand_mask_load_lanes_optab_fn): New macro.
13945 (expand_mask_store_optab_fn): Handle masked operations.
13946 (expand_mask_store_lanes_optab_fn): New macro.
13947 (direct_mask_load_lanes_optab_supported_p): Likewise.
13948 (direct_mask_store_lanes_optab_supported_p): Likewise.
13949 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13951 (vect_load_lanes_supported): Likewise.
13952 * tree-vect-data-refs.c (strip_conversion): New function.
13953 (can_group_stmts_p): Likewise.
13954 (vect_analyze_data_ref_accesses): Use it instead of checking
13955 for a pair of assignments.
13956 (vect_store_lanes_supported): Take a masked_p parameter.
13957 (vect_load_lanes_supported): Likewise.
13958 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13959 vect_store_lanes_supported and vect_load_lanes_supported.
13960 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13961 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13962 parameter. Don't allow gaps for masked accesses.
13963 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13964 and vect_load_lanes_supported.
13965 (get_load_store_type): Take a masked_p parameter and update
13966 call to get_group_load_store_type.
13967 (vectorizable_store): Update call to get_load_store_type.
13968 Handle IFN_MASK_STORE_LANES.
13969 (vectorizable_load): Update call to get_load_store_type.
13970 Handle IFN_MASK_LOAD_LANES.
13972 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13973 Alan Hayward <alan.hayward@arm.com>
13974 David Sherwood <david.sherwood@arm.com>
13976 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13978 * config/aarch64/aarch64-protos.h
13979 (aarch64_sve_struct_memory_operand_p): Declare.
13980 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13981 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13982 (VPRED, vpred): Handle SVE structure modes.
13983 * config/aarch64/constraints.md (Utx): New constraint.
13984 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13985 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13986 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13987 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13988 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13989 structure modes. Split into pieces after RA.
13990 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13991 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13993 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13994 SVE structure modes.
13995 (aarch64_classify_address): Likewise.
13996 (sizetochar): Move earlier in file.
13997 (aarch64_print_operand): Handle SVE register lists.
13998 (aarch64_array_mode): New function.
13999 (aarch64_sve_struct_memory_operand_p): Likewise.
14000 (TARGET_ARRAY_MODE): Redefine.
14002 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14003 Alan Hayward <alan.hayward@arm.com>
14004 David Sherwood <david.sherwood@arm.com>
14006 * target.def (array_mode): New target hook.
14007 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
14008 * doc/tm.texi: Regenerate.
14009 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
14010 * hooks.c (hook_optmode_mode_uhwi_none): New function.
14011 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
14012 targetm.array_mode.
14013 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14017 Alan Hayward <alan.hayward@arm.com>
14018 David Sherwood <david.sherwood@arm.com>
14020 * fold-const.c (fold_binary_loc): Check the argument types
14021 rather than the result type when testing for a vector operation.
14023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14025 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14026 * doc/tm.texi: Regenerate.
14028 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14029 Alan Hayward <alan.hayward@arm.com>
14030 David Sherwood <david.sherwood@arm.com>
14032 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14033 (sve): Document new AArch64 extension.
14034 * doc/md.texi (w): Extend the description of the AArch64
14035 constraint to include SVE vectors.
14036 (Upl, Upa): Document new AArch64 predicate constraints.
14037 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14039 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14040 (msve-vector-bits=): New option.
14041 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14042 SVE when these are disabled.
14043 (sve): New extension.
14044 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14045 modes. Adjust their number of units based on aarch64_sve_vg.
14046 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14047 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14048 aarch64_addr_query_type.
14049 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14050 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14051 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14052 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14053 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14054 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14055 (aarch64_simd_imm_zero_p): Delete.
14056 (aarch64_check_zero_based_sve_index_immediate): Declare.
14057 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14058 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14059 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14060 (aarch64_sve_float_mul_immediate_p): Likewise.
14061 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14062 rather than an rtx.
14063 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14064 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14065 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14066 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14067 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14068 (aarch64_regmode_natural_size): Likewise.
14069 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14070 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14072 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14073 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14074 for VG and the SVE predicate registers.
14075 (V_ALIASES): Add a "z"-prefixed alias.
14076 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14077 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14078 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14079 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14080 (REG_CLASS_NAMES): Add entries for them.
14081 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14082 and the predicate registers.
14083 (aarch64_sve_vg): Declare.
14084 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14085 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14086 (REGMODE_NATURAL_SIZE): Define.
14087 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14089 * config/aarch64/aarch64.c: Include cfgrtl.h.
14090 (simd_immediate_info): Add a constructor for series vectors,
14091 and an associated step field.
14092 (aarch64_sve_vg): New variable.
14093 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14094 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14095 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14096 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14097 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14098 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14099 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14100 (aarch64_get_mask_mode): New functions.
14101 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14102 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14103 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14104 predicate modes and predicate registers. Explicitly restrict
14105 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14106 to store a vector mode if it is recognized by
14107 aarch64_classify_vector_mode.
14108 (aarch64_regmode_natural_size): New function.
14109 (aarch64_hard_regno_caller_save_mode): Return the original mode
14111 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14112 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14113 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14114 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14116 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14117 does not overlap dest if the function is frame-related. Handle
14119 (aarch64_split_add_offset): New function.
14120 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14121 them aarch64_add_offset.
14122 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14123 and update call to aarch64_sub_sp.
14124 (aarch64_add_cfa_expression): New function.
14125 (aarch64_expand_prologue): Pass extra temporary registers to the
14126 functions above. Handle the case in which we need to emit new
14127 DW_CFA_expressions for registers that were originally saved
14128 relative to the stack pointer, but now have to be expressed
14129 relative to the frame pointer.
14130 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14132 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14133 IP0 and IP1 values for SVE frames.
14134 (aarch64_expand_vec_series): New function.
14135 (aarch64_expand_sve_widened_duplicate): Likewise.
14136 (aarch64_expand_sve_const_vector): Likewise.
14137 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14138 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14139 into the register, rather than emitting a SET directly.
14140 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14141 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14142 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14143 (offset_9bit_signed_scaled_p): New functions.
14144 (aarch64_replicate_bitmask_imm): New function.
14145 (aarch64_bitmask_imm): Use it.
14146 (aarch64_cannot_force_const_mem): Reject expressions involving
14147 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14148 (aarch64_classify_index): Handle SVE indices, by requiring
14149 a plain register index with a scale that matches the element size.
14150 (aarch64_classify_address): Handle SVE addresses. Assert that
14151 the mode of the address is VOIDmode or an integer mode.
14152 Update call to aarch64_classify_symbol.
14153 (aarch64_classify_symbolic_expression): Update call to
14154 aarch64_classify_symbol.
14155 (aarch64_const_vec_all_in_range_p): New function.
14156 (aarch64_print_vector_float_operand): Likewise.
14157 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14158 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14159 and the FP immediates 1.0 and 0.5.
14160 (aarch64_print_address_internal): Handle SVE addresses.
14161 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14162 (aarch64_regno_regclass): Handle predicate registers.
14163 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14165 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14166 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14167 (aarch64_convert_sve_vector_bits): New function.
14168 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14169 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14170 rather than an rtx.
14171 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14172 Handle SVE vector and predicate modes. Accept VL-based constants
14173 that need only one temporary register, and VL offsets that require
14174 no temporary registers.
14175 (aarch64_conditional_register_usage): Mark the predicate registers
14176 as fixed if SVE isn't available.
14177 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14178 Return true for SVE vector and predicate modes.
14179 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14180 rather than an unsigned int. Handle SVE modes.
14181 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14183 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14185 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14186 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14187 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14188 (aarch64_sve_float_mul_immediate_p): New functions.
14189 (aarch64_sve_valid_immediate): New function.
14190 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14191 Explicitly reject structure modes. Check for INDEX constants.
14192 Handle PTRUE and PFALSE constants.
14193 (aarch64_check_zero_based_sve_index_immediate): New function.
14194 (aarch64_simd_imm_zero_p): Delete.
14195 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14196 vector modes. Accept constants in the range of CNT[BHWD].
14197 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14198 ask for an Advanced SIMD mode.
14199 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14200 (aarch64_simd_vector_alignment): Handle SVE predicates.
14201 (aarch64_vectorize_preferred_vector_alignment): New function.
14202 (aarch64_simd_vector_alignment_reachable): Use it instead of
14204 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14205 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14207 (MAX_VECT_LEN): Delete.
14208 (expand_vec_perm_d): Add a vec_flags field.
14209 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14210 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14211 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14213 (aarch64_evpc_rev): Rename to...
14214 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14215 (aarch64_evpc_rev_global): New function.
14216 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14217 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14219 (aarch64_evpc_sve_tbl): New function.
14220 (aarch64_expand_vec_perm_const_1): Update after rename of
14221 aarch64_evpc_rev. Handle SVE permutes too, trying
14222 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14223 than aarch64_evpc_tbl.
14224 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14225 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14226 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14227 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14228 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14229 (aarch64_expand_sve_vcond): New functions.
14230 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14231 of aarch64_vector_mode_p.
14232 (aarch64_dwarf_poly_indeterminate_value): New function.
14233 (aarch64_compute_pressure_classes): Likewise.
14234 (aarch64_can_change_mode_class): Likewise.
14235 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14236 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14237 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14238 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14239 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14240 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14241 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14242 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14244 (Dn, Dl, Dr): Accept const as well as const_vector.
14245 (Dz): Likewise. Compare against CONST0_RTX.
14246 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14247 of "vector" where appropriate.
14248 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14249 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14250 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14251 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14252 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14253 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14254 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14255 (v_int_equiv): Extend to SVE modes.
14256 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14258 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14259 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14260 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14261 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14262 (SVE_COND_FP_CMP): New int iterators.
14263 (perm_hilo): Handle the new unpack unspecs.
14264 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14266 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14267 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14268 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14269 (aarch64_equality_operator, aarch64_constant_vector_operand)
14270 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14271 (aarch64_sve_nonimmediate_operand): Likewise.
14272 (aarch64_sve_general_operand): Likewise.
14273 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14274 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14275 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14276 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14277 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14278 (aarch64_sve_float_arith_immediate): Likewise.
14279 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14280 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14281 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14282 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14283 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14284 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14285 (aarch64_sve_float_arith_operand): Likewise.
14286 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14287 (aarch64_sve_float_mul_operand): Likewise.
14288 (aarch64_sve_vec_perm_operand): Likewise.
14289 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14290 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14291 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14292 as well as const_vector.
14293 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14294 in file. Use CONST0_RTX and CONSTM1_RTX.
14295 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14296 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14297 Use aarch64_simd_imm_zero.
14298 * config/aarch64/aarch64-sve.md: New file.
14299 * config/aarch64/aarch64.md: Include it.
14300 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14301 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14302 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14303 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14304 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14305 (sve): New attribute.
14306 (enabled): Disable instructions with the sve attribute unless
14308 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14309 aarch64_expand_mov_immediate.
14310 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14311 CNT[BHSD] immediates.
14312 (movti): Split CONST_POLY_INT moves into two halves.
14313 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14314 Split additions that need a temporary here if the destination
14315 is the stack pointer.
14316 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14317 (*add<mode>3_poly_1): New instruction.
14318 (set_clobber_cc): New expander.
14320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14322 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14323 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14324 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14325 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14326 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14327 Change innermode from fixed_mode_size to machine_mode.
14328 (simplify_subreg): Update call accordingly. Handle a constant-sized
14329 subreg of a variable-length CONST_VECTOR.
14331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14332 Alan Hayward <alan.hayward@arm.com>
14333 David Sherwood <david.sherwood@arm.com>
14335 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14336 (add_offset_to_base): New function, split out from...
14337 (create_mem_ref): ...here. When handling a scale other than 1,
14338 check first whether the address is valid without the offset.
14339 Add it into the base if so, leaving the index and scale as-is.
14341 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14344 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14345 fold_for_warn before checking if arg2 is INTEGER_CST.
14347 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14349 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14350 (store_multiple_operation): Delete.
14351 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14352 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14353 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14354 guarded by TARGET_STRING.
14355 (rs6000_output_load_multiple): Delete.
14356 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14357 OPTION_MASK_STRING / TARGET_STRING handling.
14358 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14359 (const rs6000_opt_masks) <"string">: Change mask to 0.
14360 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14361 (MASK_STRING): Delete.
14362 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14364 (load_multiple): Delete.
14371 (store_multiple): Delete.
14378 (movmemsi_8reg): Delete.
14379 (corresponding unnamed define_insn): Delete.
14380 (movmemsi_6reg): Delete.
14381 (corresponding unnamed define_insn): Delete.
14382 (movmemsi_4reg): Delete.
14383 (corresponding unnamed define_insn): Delete.
14384 (movmemsi_2reg): Delete.
14385 (corresponding unnamed define_insn): Delete.
14386 (movmemsi_1reg): Delete.
14387 (corresponding unnamed define_insn): Delete.
14388 * config/rs6000/rs6000.opt (mno-string): New.
14389 (mstring): Replace by deprecation warning stub.
14390 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14392 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14394 * regrename.c (regrename_do_replace): If replacing the same
14395 reg multiple times, try to reuse last created gen_raw_REG.
14398 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14399 main to workaround a bug in GDB.
14401 2018-01-12 Tom de Vries <tom@codesourcery.com>
14404 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14406 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14408 PR rtl-optimization/80481
14409 * ira-color.c (get_cap_member): New function.
14410 (allocnos_conflict_by_live_ranges_p): Use it.
14411 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14412 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14414 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14417 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14418 (*saddl_se_1): Ditto.
14419 (*ssubsi_1): Ditto.
14420 (*ssubl_se_1): Ditto.
14422 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14424 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14425 rather than wi::to_widest for DR_INITs.
14426 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14427 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14428 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14430 (vect_analyze_group_access_1): Note that here.
14432 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14434 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14435 polynomial type sizes.
14437 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14439 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14440 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14441 (gimple_add_tmp_var): Likewise.
14443 2018-01-12 Martin Liska <mliska@suse.cz>
14445 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14446 (gimple_alloc_sizes): Likewise.
14447 (dump_gimple_statistics): Use PRIu64 in printf format.
14448 * gimple.h: Change uint64_t to int.
14450 2018-01-12 Martin Liska <mliska@suse.cz>
14452 * tree-core.h: Use uint64_t instead of int.
14453 * tree.c (tree_node_counts): Likewise.
14454 (tree_node_sizes): Likewise.
14455 (dump_tree_statistics): Use PRIu64 in printf format.
14457 2018-01-12 Martin Liska <mliska@suse.cz>
14459 * Makefile.in: As qsort_chk is implemented in vec.c, add
14460 vec.o to linkage of gencfn-macros.
14461 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14462 passing the info to record_node_allocation_statistics.
14463 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14465 * ggc-common.c (struct ggc_usage): Add operator== and use
14466 it in operator< and compare function.
14467 * mem-stats.h (struct mem_usage): Likewise.
14468 * vec.c (struct vec_usage): Remove operator< and compare
14469 function. Can be simply inherited.
14471 2018-01-12 Martin Jambor <mjambor@suse.cz>
14474 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14475 * tree-ssa-math-opts.c: Include domwalk.h.
14476 (convert_mult_to_fma_1): New function.
14477 (fma_transformation_info): New type.
14478 (fma_deferring_state): Likewise.
14479 (cancel_fma_deferring): New function.
14480 (result_of_phi): Likewise.
14481 (last_fma_candidate_feeds_initial_phi): Likewise.
14482 (convert_mult_to_fma): Added deferring logic, split actual
14483 transformation to convert_mult_to_fma_1.
14484 (math_opts_dom_walker): New type.
14485 (math_opts_dom_walker::after_dom_children): New method, body moved
14486 here from pass_optimize_widening_mul::execute, added deferring logic
14488 (pass_optimize_widening_mul::execute): Moved most of code to
14489 math_opts_dom_walker::after_dom_children.
14490 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14491 * config/i386/i386.c (ix86_option_override_internal): Added
14492 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14494 2018-01-12 Richard Biener <rguenther@suse.de>
14497 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14498 inline instance vars.
14500 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14503 * config/rx/rx.c (rx_is_restricted_memory_address):
14504 Handle SUBREG case.
14506 2018-01-12 Richard Biener <rguenther@suse.de>
14508 PR tree-optimization/80846
14509 * target.def (split_reduction): New target hook.
14510 * targhooks.c (default_split_reduction): New function.
14511 * targhooks.h (default_split_reduction): Declare.
14512 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14513 target requests first reduce vectors by combining low and high
14515 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14516 (get_vectype_for_scalar_type_and_size): Export.
14517 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14518 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14519 * doc/tm.texi: Regenerate.
14520 * config/i386/i386.c (ix86_split_reduction): Implement
14521 TARGET_VECTORIZE_SPLIT_REDUCTION.
14523 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14526 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14527 in PIC mode except for TARGET_VXWORKS_RTP.
14528 * config/sparc/sparc.c: Include cfgrtl.h.
14529 (TARGET_INIT_PIC_REG): Define.
14530 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14531 (sparc_pic_register_p): New predicate.
14532 (sparc_legitimate_address_p): Use it.
14533 (sparc_legitimize_pic_address): Likewise.
14534 (sparc_delegitimize_address): Likewise.
14535 (sparc_mode_dependent_address_p): Likewise.
14536 (gen_load_pcrel_sym): Remove 4th parameter.
14537 (load_got_register): Adjust call to above. Remove obsolete stuff.
14538 (sparc_expand_prologue): Do not call load_got_register here.
14539 (sparc_flat_expand_prologue): Likewise.
14540 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14541 (sparc_use_pseudo_pic_reg): New function.
14542 (sparc_init_pic_reg): Likewise.
14543 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14544 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14546 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14548 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14549 Add item for branch_cost.
14551 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14553 PR rtl-optimization/83565
14554 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14555 not extend the result to a larger mode for rotate operations.
14556 (num_sign_bit_copies1): Likewise.
14558 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14561 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14563 Use values-Xc.o for -pedantic.
14564 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14566 2018-01-12 Martin Liska <mliska@suse.cz>
14569 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14571 (possible_polymorphic_call_targets): Use it.
14572 (ipa_devirt): Likewise.
14574 2018-01-12 Martin Liska <mliska@suse.cz>
14576 * profile-count.h (enum profile_quality): Use 0 as invalid
14577 enum value of profile_quality.
14579 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14581 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14582 -mext-string options.
14584 2018-01-12 Richard Biener <rguenther@suse.de>
14586 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14587 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14588 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14590 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14592 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14594 * configure.ac (--with-long-double-format): Add support for the
14595 configuration option to change the default long double format on
14597 * config.gcc (powerpc*-linux*-*): Likewise.
14598 * configure: Regenerate.
14599 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14600 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14601 used without modification.
14603 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14605 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14606 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14607 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14608 MISC_BUILTIN_SPEC_BARRIER.
14609 (rs6000_init_builtins): Likewise.
14610 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14612 (speculation_barrier): New define_insn.
14613 * doc/extend.texi: Document __builtin_speculation_barrier.
14615 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14618 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14619 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14620 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14622 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14623 integral modes instead of "ss" and "sd".
14624 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14625 vectors with 32-bit and 64-bit elements.
14626 (vecdupssescalarmodesuffix): New mode attribute.
14627 (vec_dup<mode>): Use it.
14629 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14632 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14633 frame if argument is passed on stack.
14635 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14638 * ree.c (combine_reaching_defs): Optimize also
14639 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14640 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14642 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14644 PR middle-end/83189
14645 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14647 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14649 PR middle-end/83718
14650 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14651 after they are computed.
14653 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14655 PR tree-optimization/83695
14656 * gimple-loop-linterchange.cc
14657 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14658 reset cached scev information after interchange.
14659 (pass_linterchange::execute): Remove call to scev_reset_htab.
14661 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14663 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14664 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14665 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14666 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14667 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14668 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14669 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14670 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14671 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14672 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14673 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14674 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14675 (V_lane_reg): Likewise.
14676 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14678 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14679 (vfmal_lane_low<mode>_intrinsic,
14680 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14681 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14682 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14683 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14684 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14685 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14687 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14689 * config/arm/arm-cpus.in (fp16fml): New feature.
14690 (ALL_SIMD): Add fp16fml.
14691 (armv8.2-a): Add fp16fml as an option.
14692 (armv8.3-a): Likewise.
14693 (armv8.4-a): Add fp16fml as part of fp16.
14694 * config/arm/arm.h (TARGET_FP16FML): Define.
14695 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14697 * config/arm/arm-modes.def (V2HF): Define.
14698 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14699 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14700 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14701 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14702 vfmsl_low, vfmsl_high): New set of builtins.
14703 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14704 (vfml_op): New code attribute.
14705 (VFMLHALVES): New int iterator.
14706 (VFML, VFMLSEL): New mode attributes.
14707 (V_reg): Define mapping for V2HF.
14708 (V_hi, V_lo): New mode attributes.
14709 (VF_constraint): Likewise.
14710 (vfml_half, vfml_half_selector): New int attributes.
14711 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14713 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14714 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14716 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14717 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14718 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14719 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14721 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14722 Document new effective target and option set.
14724 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14726 * config/arm/arm-cpus.in (armv8_4): New feature.
14727 (ARMv8_4a): New fgroup.
14728 (armv8.4-a): New arch.
14729 * config/arm/arm-tables.opt: Regenerate.
14730 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14731 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14732 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14733 Add matching rules for -march=armv8.4-a and extensions.
14734 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14736 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14739 * config/rx/rx.md (BW): New mode attribute.
14740 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14742 2018-01-11 Richard Biener <rguenther@suse.de>
14744 PR tree-optimization/83435
14745 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14746 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14747 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14749 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14750 Alan Hayward <alan.hayward@arm.com>
14751 David Sherwood <david.sherwood@arm.com>
14753 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14755 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14756 (aarch64_print_address_internal): Use it to check for a zero offset.
14758 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14759 Alan Hayward <alan.hayward@arm.com>
14760 David Sherwood <david.sherwood@arm.com>
14762 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14763 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14764 Return a poly_int64 rather than a HOST_WIDE_INT.
14765 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14766 rather than a HOST_WIDE_INT.
14767 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14768 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14769 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14770 final_offset from HOST_WIDE_INT to poly_int64.
14771 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14772 to_constant when getting the number of units in an Advanced SIMD
14774 (aarch64_builtin_vectorized_function): Check for a constant number
14776 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14778 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14779 attribute instead of GET_MODE_NUNITS.
14780 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14781 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14782 GET_MODE_SIZE for fixed-size registers.
14783 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14784 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14785 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14786 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14787 (aarch64_print_operand, aarch64_print_address_internal)
14788 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14789 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14790 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14791 Handle polynomial GET_MODE_SIZE.
14792 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14793 wider than SImode without modification.
14794 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14795 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14796 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14797 passing and returning SVE modes.
14798 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14799 rather than GEN_INT.
14800 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14801 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14802 (aarch64_allocate_and_probe_stack_space): Likewise.
14803 (aarch64_layout_frame): Cope with polynomial offsets.
14804 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14805 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14806 polynomial offsets.
14807 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14808 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14809 poly_int64 rather than a HOST_WIDE_INT.
14810 (aarch64_get_separate_components, aarch64_process_components)
14811 (aarch64_expand_prologue, aarch64_expand_epilogue)
14812 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14813 (aarch64_anchor_offset): New function, split out from...
14814 (aarch64_legitimize_address): ...here.
14815 (aarch64_builtin_vectorization_cost): Handle polynomial
14816 TYPE_VECTOR_SUBPARTS.
14817 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14819 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14820 number of elements from the PARALLEL rather than the mode.
14821 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14822 rather than GET_MODE_BITSIZE.
14823 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14824 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14825 (aarch64_expand_vec_perm_const_1): Handle polynomial
14826 d->perm.length () and d->perm elements.
14827 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14828 Apply to_constant to d->perm elements.
14829 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14830 polynomial CONST_VECTOR_NUNITS.
14831 (aarch64_move_pointer): Take amount as a poly_int64 rather
14833 (aarch64_progress_pointer): Avoid temporary variable.
14834 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14835 the mode attribute instead of GET_MODE.
14837 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14838 Alan Hayward <alan.hayward@arm.com>
14839 David Sherwood <david.sherwood@arm.com>
14841 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14842 x exists before using it.
14843 (aarch64_add_constant_internal): Rename to...
14844 (aarch64_add_offset_1): ...this. Replace regnum with separate
14845 src and dest rtxes. Handle the case in which they're different,
14846 including when the offset is zero. Replace scratchreg with an rtx.
14847 Use 2 additions if there is no spare register into which we can
14848 move a 16-bit constant.
14849 (aarch64_add_constant): Delete.
14850 (aarch64_add_offset): Replace reg with separate src and dest
14851 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14852 Use aarch64_add_offset_1.
14853 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14854 an rtx rather than an int. Take the delta as a poly_int64
14855 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14856 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14857 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14858 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14859 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14860 and aarch64_add_sp.
14861 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14862 aarch64_add_constant.
14864 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14866 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14867 Use scalar_float_mode.
14869 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14871 * config/aarch64/aarch64-simd.md
14872 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14873 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14874 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14875 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14876 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14877 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14878 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14879 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14880 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14881 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14883 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14886 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14887 targ_options->x_arm_arch_string is non NULL.
14889 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14891 * config/aarch64/aarch64.h
14892 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14894 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14897 * expmed.c (emit_store_flag_force): Swap if const op0
14898 and change VOIDmode to mode of op0.
14900 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14902 PR rtl-optimization/83761
14903 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14904 than bytes to mode_for_size.
14906 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14908 PR middle-end/83189
14909 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14910 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14913 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14915 PR middle-end/83575
14916 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14917 when in layout mode.
14918 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14919 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14922 2018-01-10 Michael Collison <michael.collison@arm.com>
14924 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14925 * config/aarch64/aarch64-option-extension.def: Add
14926 AARCH64_OPT_EXTENSION of 'fp16fml'.
14927 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14928 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14929 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14930 * config/aarch64/constraints.md (Ui7): New constraint.
14931 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14932 (VFMLA_SEL_W): Ditto.
14935 (VFMLA16_LOW): New int iterator.
14936 (VFMLA16_HIGH): Ditto.
14937 (UNSPEC_FMLAL): New unspec.
14938 (UNSPEC_FMLSL): Ditto.
14939 (UNSPEC_FMLAL2): Ditto.
14940 (UNSPEC_FMLSL2): Ditto.
14941 (f16mac): New code attribute.
14942 * config/aarch64/aarch64-simd-builtins.def
14943 (aarch64_fmlal_lowv2sf): Ditto.
14944 (aarch64_fmlsl_lowv2sf): Ditto.
14945 (aarch64_fmlalq_lowv4sf): Ditto.
14946 (aarch64_fmlslq_lowv4sf): Ditto.
14947 (aarch64_fmlal_highv2sf): Ditto.
14948 (aarch64_fmlsl_highv2sf): Ditto.
14949 (aarch64_fmlalq_highv4sf): Ditto.
14950 (aarch64_fmlslq_highv4sf): Ditto.
14951 (aarch64_fmlal_lane_lowv2sf): Ditto.
14952 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14953 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14954 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14955 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14956 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14957 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14958 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14959 (aarch64_fmlal_lane_highv2sf): Ditto.
14960 (aarch64_fmlsl_lane_highv2sf): Ditto.
14961 (aarch64_fmlal_laneq_highv2sf): Ditto.
14962 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14963 (aarch64_fmlalq_lane_highv4sf): Ditto.
14964 (aarch64_fmlsl_lane_highv4sf): Ditto.
14965 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14966 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14967 * config/aarch64/aarch64-simd.md:
14968 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14969 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14970 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14971 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14972 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14973 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14974 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14975 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14976 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14977 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14978 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14979 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14980 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14981 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14982 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14983 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14984 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14985 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14986 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14987 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14988 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14989 (vfmlsl_low_u32): Ditto.
14990 (vfmlalq_low_u32): Ditto.
14991 (vfmlslq_low_u32): Ditto.
14992 (vfmlal_high_u32): Ditto.
14993 (vfmlsl_high_u32): Ditto.
14994 (vfmlalq_high_u32): Ditto.
14995 (vfmlslq_high_u32): Ditto.
14996 (vfmlal_lane_low_u32): Ditto.
14997 (vfmlsl_lane_low_u32): Ditto.
14998 (vfmlal_laneq_low_u32): Ditto.
14999 (vfmlsl_laneq_low_u32): Ditto.
15000 (vfmlalq_lane_low_u32): Ditto.
15001 (vfmlslq_lane_low_u32): Ditto.
15002 (vfmlalq_laneq_low_u32): Ditto.
15003 (vfmlslq_laneq_low_u32): Ditto.
15004 (vfmlal_lane_high_u32): Ditto.
15005 (vfmlsl_lane_high_u32): Ditto.
15006 (vfmlal_laneq_high_u32): Ditto.
15007 (vfmlsl_laneq_high_u32): Ditto.
15008 (vfmlalq_lane_high_u32): Ditto.
15009 (vfmlslq_lane_high_u32): Ditto.
15010 (vfmlalq_laneq_high_u32): Ditto.
15011 (vfmlslq_laneq_high_u32): Ditto.
15012 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
15013 (AARCH64_FL_FOR_ARCH8_4): New.
15014 (AARCH64_ISA_F16FML): New ISA flag.
15015 (TARGET_F16FML): New feature flag for fp16fml.
15016 (doc/invoke.texi): Document new fp16fml option.
15018 2018-01-10 Michael Collison <michael.collison@arm.com>
15020 * config/aarch64/aarch64-builtins.c:
15021 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15022 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15023 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15024 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15025 (AARCH64_ISA_SHA3): New ISA flag.
15026 (TARGET_SHA3): New feature flag for sha3.
15027 * config/aarch64/iterators.md (sha512_op): New int attribute.
15028 (CRYPTO_SHA512): New int iterator.
15029 (UNSPEC_SHA512H): New unspec.
15030 (UNSPEC_SHA512H2): Ditto.
15031 (UNSPEC_SHA512SU0): Ditto.
15032 (UNSPEC_SHA512SU1): Ditto.
15033 * config/aarch64/aarch64-simd-builtins.def
15034 (aarch64_crypto_sha512hqv2di): New builtin.
15035 (aarch64_crypto_sha512h2qv2di): Ditto.
15036 (aarch64_crypto_sha512su0qv2di): Ditto.
15037 (aarch64_crypto_sha512su1qv2di): Ditto.
15038 (aarch64_eor3qv8hi): Ditto.
15039 (aarch64_rax1qv2di): Ditto.
15040 (aarch64_xarqv2di): Ditto.
15041 (aarch64_bcaxqv8hi): Ditto.
15042 * config/aarch64/aarch64-simd.md:
15043 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15044 (aarch64_crypto_sha512su0qv2di): Ditto.
15045 (aarch64_crypto_sha512su1qv2di): Ditto.
15046 (aarch64_eor3qv8hi): Ditto.
15047 (aarch64_rax1qv2di): Ditto.
15048 (aarch64_xarqv2di): Ditto.
15049 (aarch64_bcaxqv8hi): Ditto.
15050 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15051 (vsha512h2q_u64): Ditto.
15052 (vsha512su0q_u64): Ditto.
15053 (vsha512su1q_u64): Ditto.
15054 (veor3q_u16): Ditto.
15055 (vrax1q_u64): Ditto.
15056 (vxarq_u64): Ditto.
15057 (vbcaxq_u16): Ditto.
15058 * config/arm/types.md (crypto_sha512): New type attribute.
15059 (crypto_sha3): Ditto.
15060 (doc/invoke.texi): Document new sha3 option.
15062 2018-01-10 Michael Collison <michael.collison@arm.com>
15064 * config/aarch64/aarch64-builtins.c:
15065 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15066 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15067 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15068 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15069 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15070 (AARCH64_ISA_SM4): New ISA flag.
15071 (TARGET_SM4): New feature flag for sm4.
15072 * config/aarch64/aarch64-simd-builtins.def
15073 (aarch64_sm3ss1qv4si): Ditto.
15074 (aarch64_sm3tt1aq4si): Ditto.
15075 (aarch64_sm3tt1bq4si): Ditto.
15076 (aarch64_sm3tt2aq4si): Ditto.
15077 (aarch64_sm3tt2bq4si): Ditto.
15078 (aarch64_sm3partw1qv4si): Ditto.
15079 (aarch64_sm3partw2qv4si): Ditto.
15080 (aarch64_sm4eqv4si): Ditto.
15081 (aarch64_sm4ekeyqv4si): Ditto.
15082 * config/aarch64/aarch64-simd.md:
15083 (aarch64_sm3ss1qv4si): Ditto.
15084 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15085 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15086 (aarch64_sm4eqv4si): Ditto.
15087 (aarch64_sm4ekeyqv4si): Ditto.
15088 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15089 (sm3part_op): Ditto.
15090 (CRYPTO_SM3TT): Ditto.
15091 (CRYPTO_SM3PART): Ditto.
15092 (UNSPEC_SM3SS1): New unspec.
15093 (UNSPEC_SM3TT1A): Ditto.
15094 (UNSPEC_SM3TT1B): Ditto.
15095 (UNSPEC_SM3TT2A): Ditto.
15096 (UNSPEC_SM3TT2B): Ditto.
15097 (UNSPEC_SM3PARTW1): Ditto.
15098 (UNSPEC_SM3PARTW2): Ditto.
15099 (UNSPEC_SM4E): Ditto.
15100 (UNSPEC_SM4EKEY): Ditto.
15101 * config/aarch64/constraints.md (Ui2): New constraint.
15102 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15103 * config/arm/types.md (crypto_sm3): New type attribute.
15104 (crypto_sm4): Ditto.
15105 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15106 (vsm3tt1aq_u32): Ditto.
15107 (vsm3tt1bq_u32): Ditto.
15108 (vsm3tt2aq_u32): Ditto.
15109 (vsm3tt2bq_u32): Ditto.
15110 (vsm3partw1q_u32): Ditto.
15111 (vsm3partw2q_u32): Ditto.
15112 (vsm4eq_u32): Ditto.
15113 (vsm4ekeyq_u32): Ditto.
15114 (doc/invoke.texi): Document new sm4 option.
15116 2018-01-10 Michael Collison <michael.collison@arm.com>
15118 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15119 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15120 (AARCH64_FL_FOR_ARCH8_4): New.
15121 (AARCH64_FL_V8_4): New flag.
15122 (doc/invoke.texi): Document new armv8.4-a option.
15124 2018-01-10 Michael Collison <michael.collison@arm.com>
15126 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15127 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15128 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15129 * config/aarch64/aarch64-option-extension.def: Add
15130 AARCH64_OPT_EXTENSION of 'sha2'.
15131 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15132 (crypto): Disable sha2 and aes if crypto disabled.
15133 (crypto): Enable aes and sha2 if enabled.
15134 (simd): Disable sha2 and aes if simd disabled.
15135 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15137 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15138 (TARGET_SHA2): New feature flag for sha2.
15139 (TARGET_AES): New feature flag for aes.
15140 * config/aarch64/aarch64-simd.md:
15141 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15142 conditional on TARGET_AES.
15143 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15144 (aarch64_crypto_sha1hsi): Make pattern conditional
15146 (aarch64_crypto_sha1hv4si): Ditto.
15147 (aarch64_be_crypto_sha1hv4si): Ditto.
15148 (aarch64_crypto_sha1su1v4si): Ditto.
15149 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15150 (aarch64_crypto_sha1su0v4si): Ditto.
15151 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15152 (aarch64_crypto_sha256su0v4si): Ditto.
15153 (aarch64_crypto_sha256su1v4si): Ditto.
15154 (doc/invoke.texi): Document new aes and sha2 options.
15156 2018-01-10 Martin Sebor <msebor@redhat.com>
15158 PR tree-optimization/83781
15159 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15162 2018-01-11 Martin Sebor <msebor@gmail.com>
15163 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15165 PR tree-optimization/83501
15166 PR tree-optimization/81703
15168 * tree-ssa-strlen.c (get_string_cst): Rename...
15169 (get_string_len): ...to this. Handle global constants.
15170 (handle_char_store): Adjust.
15172 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15173 Jim Wilson <jimw@sifive.com>
15175 * config/riscv/riscv-protos.h (riscv_output_return): New.
15176 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15177 (riscv_attribute_table, riscv_output_return),
15178 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15179 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15180 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15182 (riscv_expand_prologue): Add early return for naked function.
15183 (riscv_expand_epilogue): Likewise.
15184 (riscv_function_ok_for_sibcall): Return false for naked function.
15185 (riscv_set_current_function): New.
15186 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15187 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15188 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15189 * doc/extend.texi (RISC-V Function Attributes): New.
15191 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15193 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15194 check for 128-bit long double before checking TCmode.
15195 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15196 128-bit long doubles before checking TFmode or TCmode.
15197 (FLOAT128_IBM_P): Likewise.
15199 2018-01-10 Martin Sebor <msebor@redhat.com>
15201 PR tree-optimization/83671
15202 * builtins.c (c_strlen): Unconditionally return zero for the empty
15204 Use -Warray-bounds for warnings.
15205 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15206 for non-constant array indices with COMPONENT_REF, arrays of
15207 arrays, and pointers to arrays.
15208 (gimple_fold_builtin_strlen): Determine and set length range for
15209 non-constant character arrays.
15211 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15213 PR middle-end/81897
15214 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15217 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15219 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15221 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15224 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15225 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15226 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15227 indexed_or_indirect_operand predicate.
15228 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15229 (*vsx_le_perm_load_v8hi): Likewise.
15230 (*vsx_le_perm_load_v16qi): Likewise.
15231 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15232 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15233 (*vsx_le_perm_store_v8hi): Likewise.
15234 (*vsx_le_perm_store_v16qi): Likewise.
15235 (eight unnamed splitters): Likewise.
15237 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15239 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15240 * config/rs6000/emmintrin.h: Likewise.
15241 * config/rs6000/mmintrin.h: Likewise.
15242 * config/rs6000/xmmintrin.h: Likewise.
15244 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15247 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15249 * tree.c (tree_nop_conversion): Return true for location wrapper
15251 (maybe_wrap_with_location): New function.
15252 (selftest::check_strip_nops): New function.
15253 (selftest::test_location_wrappers): New function.
15254 (selftest::tree_c_tests): Call it.
15255 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15256 (maybe_wrap_with_location): New decl.
15257 (EXPR_LOCATION_WRAPPER_P): New macro.
15258 (location_wrapper_p): New inline function.
15259 (tree_strip_any_location_wrapper): New inline function.
15261 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15264 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15265 stack_realign_offset for the largest alignment of stack slot
15267 (ix86_find_max_used_stack_alignment): New function.
15268 (ix86_finalize_stack_frame_flags): Use it. Set
15269 max_used_stack_alignment if we don't realign stack.
15270 * config/i386/i386.h (machine_function): Add
15271 max_used_stack_alignment.
15273 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15275 * config/arm/arm.opt (-mbranch-cost): New option.
15276 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15279 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15282 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15283 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15285 2018-01-10 Richard Biener <rguenther@suse.de>
15288 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15289 early out so it also covers the case where we have a non-NULL
15292 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15294 PR tree-optimization/83753
15295 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15296 for non-strided grouped accesses if the number of elements is 1.
15298 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15301 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15302 * i386.h (TARGET_USE_GATHER): Define.
15303 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15305 2018-01-10 Martin Liska <mliska@suse.cz>
15308 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15309 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15311 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15312 CLEANUP_NO_PARTITIONING is not set.
15314 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15316 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15317 for vectors, as a partial revert of r254296.
15318 * rtl.h (const_vec_p): Delete.
15319 (const_vec_duplicate_p): Don't test for vector CONSTs.
15320 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15321 * expmed.c (make_tree): Likewise.
15324 * common.md (E, F): Use CONSTANT_P instead of checking for
15326 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15327 checking for CONST_VECTOR.
15329 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15331 PR middle-end/83575
15332 * predict.c (force_edge_cold): Handle in more sane way edges
15333 with no prediction.
15335 2018-01-09 Carl Love <cel@us.ibm.com>
15337 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15339 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15340 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15341 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15342 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15343 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15344 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15345 * config/rs6000/rs6000-protos.h: Add extern defition for
15346 rs6000_generate_float2_double_code.
15347 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15349 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15350 (float2_v2df): Add define_expand.
15352 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15355 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15356 op_mode in the force_to_mode call.
15358 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15360 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15361 instead of checking each element individually.
15362 (aarch64_evpc_uzp): Likewise.
15363 (aarch64_evpc_zip): Likewise.
15364 (aarch64_evpc_ext): Likewise.
15365 (aarch64_evpc_rev): Likewise.
15366 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15367 instead of checking each element individually. Return true without
15369 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15370 whether all selected elements come from the same input, instead of
15371 checking each element individually. Remove calls to gen_rtx_REG,
15372 start_sequence and end_sequence and instead assert that no rtl is
15375 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15377 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15378 order of HIGH and CONST checks.
15380 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15382 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15383 if the destination isn't an SSA_NAME.
15385 2018-01-09 Richard Biener <rguenther@suse.de>
15387 PR tree-optimization/83668
15388 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15390 (canonicalize_loop_form): ... here, renamed from ...
15391 (canonicalize_loop_closed_ssa_form): ... this and amended to
15392 swap successor edges for loop exit blocks to make us use
15393 the RPO order we need for initial schedule generation.
15395 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15397 PR tree-optimization/64811
15398 * match.pd: When optimizing comparisons with Inf, avoid
15399 introducing or losing exceptions from comparisons with NaN.
15401 2018-01-09 Martin Liska <mliska@suse.cz>
15404 * asan.c (shadow_mem_size): Add gcc_assert.
15406 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15408 Don't save registers in main().
15411 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15412 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15413 * config/avr/avr.c (avr_set_current_function): Don't error if
15414 naked, OS_task or OS_main are specified at the same time.
15415 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15417 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15419 * common/config/avr/avr-common.c (avr_option_optimization_table):
15420 Switch on -mmain-is-OS_task for optimizing compilations.
15422 2018-01-09 Richard Biener <rguenther@suse.de>
15424 PR tree-optimization/83572
15425 * graphite.c: Include cfganal.h.
15426 (graphite_transform_loops): Connect infinite loops to exit
15427 and remove fake edges at the end.
15429 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15431 * ipa-inline.c (edge_badness): Revert accidental checkin.
15433 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15436 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15437 symbols; not inline clones.
15439 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15442 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15443 hard registers. Formatting fixes.
15445 PR preprocessor/83722
15446 * gcc.c (try_generate_repro): Pass
15447 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15448 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15451 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15452 Kito Cheng <kito.cheng@gmail.com>
15454 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15455 (riscv_leaf_function_p): Delete.
15456 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15458 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15460 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15462 (do_ifelse): New function.
15463 (do_isel): New function.
15464 (do_sub3): New function.
15465 (do_add3): New function.
15466 (do_load_mask_compare): New function.
15467 (do_overlap_load_compare): New function.
15468 (expand_compare_loop): New function.
15469 (expand_block_compare): Call expand_compare_loop() when appropriate.
15470 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15471 option description.
15472 (-mblock-compare-inline-loop-limit): New option.
15474 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15477 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15478 Reverse order of second and third operands in first alternative.
15479 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15480 of first and second elements in UNSPEC_VPERMR vector.
15481 (altivec_expand_vec_perm_le): Likewise.
15483 2018-01-08 Jeff Law <law@redhat.com>
15485 PR rtl-optimizatin/81308
15486 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15487 (process_switch): If group_case_labels makes a change, then set
15489 (pass_convert_switch::execute): If a switch is converted, then
15490 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15492 PR rtl-optimization/81308
15493 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15496 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15498 PR target/83663 - Revert r255946
15499 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15500 generation for cases where splatting a value is not useful.
15501 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15502 across a vec_duplicate and a paradoxical subreg forming a vector
15503 mode to a vec_concat.
15505 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15507 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15508 -march=armv8.3-a variants.
15509 * config/arm/t-multilib: Likewise.
15510 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15512 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15514 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15516 (cceq_ior_compare_complement): Give it a name so I can use it, and
15517 change boolean_or_operator predicate to boolean_operator so it can
15518 be used to generate a crand.
15519 (eqne): New code iterator.
15520 (bd/bd_neg): New code_attrs.
15521 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15522 a single define_insn.
15523 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15524 decrement (bdnzt/bdnzf/bdzt/bdzf).
15525 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15526 with the new names of the branch decrement patterns, and added the
15527 names of the branch decrement conditional patterns.
15529 2018-01-08 Richard Biener <rguenther@suse.de>
15531 PR tree-optimization/83563
15532 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15535 2018-01-08 Richard Biener <rguenther@suse.de>
15537 PR middle-end/83713
15538 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15540 2018-01-08 Richard Biener <rguenther@suse.de>
15542 PR tree-optimization/83685
15543 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15544 references to abnormals.
15546 2018-01-08 Richard Biener <rguenther@suse.de>
15549 * dwarf2out.c (output_indirect_strings): Handle empty
15550 skeleton_debug_str_hash.
15551 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15553 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15555 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15556 (emit_store_direct): Likewise.
15557 (arc_trampoline_adjust_address): Likewise.
15558 (arc_asm_trampoline_template): New function.
15559 (arc_initialize_trampoline): Use asm_trampoline_template.
15560 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15561 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15562 * config/arc/arc.md (flush_icache): Delete pattern.
15564 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15566 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15567 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15570 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15573 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15574 by not USED_FOR_TARGET.
15575 (make_pass_resolve_sw_modes): Likewise.
15577 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15579 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15582 2018-01-08 Richard Biener <rguenther@suse.de>
15584 PR middle-end/83580
15585 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15587 2018-01-08 Richard Biener <rguenther@suse.de>
15589 PR middle-end/83517
15590 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15592 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15594 PR middle-end/81897
15595 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15596 basic blocks with a small number of successors.
15597 (convert_control_dep_chain_into_preds): Improve handling of
15599 (dump_predicates): Split apart into...
15600 (dump_pred_chain): ...here...
15601 (dump_pred_info): ...and here.
15602 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15603 (can_chain_union_be_invalidated_p): Improve check for invalidation
15605 (uninit_uses_cannot_happen): Avoid unnecessary if
15606 convert_control_dep_chain_into_preds yielded nothing.
15608 2018-01-06 Martin Sebor <msebor@redhat.com>
15610 PR tree-optimization/83640
15611 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15612 subtracting negative offset from size.
15613 (builtin_access::overlap): Adjust offset bounds of the access to fall
15614 within the size of the object if possible.
15616 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15618 PR rtl-optimization/83699
15619 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15620 extract_bit_field_as_subreg to cases in which the extracted
15621 value is also a vector.
15623 * lra-constraints.c (process_alt_operands): Test for the equivalence
15624 substitutions when detecting a possible reload cycle.
15626 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15629 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15630 by default if flag_selective_schedling{,2}. Formatting fixes.
15632 PR rtl-optimization/83682
15633 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15634 if it has non-VECTOR_MODE element mode.
15635 (vec_duplicate_p): Likewise.
15637 PR middle-end/83694
15638 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15639 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15641 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15644 * config/i386/i386-builtin.def
15645 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15646 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15647 Require also OPTION_MASK_ISA_AVX512F in addition to
15648 OPTION_MASK_ISA_GFNI.
15649 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15650 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15651 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15652 to OPTION_MASK_ISA_GFNI.
15653 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15654 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15655 OPTION_MASK_ISA_AVX512BW.
15656 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15657 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15658 addition to OPTION_MASK_ISA_GFNI.
15659 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15660 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15661 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15662 to OPTION_MASK_ISA_GFNI.
15663 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15664 a requirement for all ISAs rather than any of them with a few
15666 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15668 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15669 bitmasks to be enabled with 3 exceptions, instead of requiring any
15670 enabled ISA with lots of exceptions.
15671 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15672 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15673 Change avx512bw in isa attribute to avx512f.
15674 * config/i386/sgxintrin.h: Add license boilerplate.
15675 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15676 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15677 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15678 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15680 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15681 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15682 temporarily sse2 rather than sse if not enabled already.
15685 * config/i386/sse.md (VI248_VLBW): Rename to ...
15686 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15687 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15688 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15689 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15690 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15691 mode iterator instead of VI248_VLBW.
15693 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15695 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15696 (record_modified): Skip clobbers; add debug output.
15697 (param_change_prob): Use sreal frequencies.
15699 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15701 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15702 punt for user-aligned variables.
15704 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15706 * tree-chrec.c (chrec_contains_symbols): Return true for
15709 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15712 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15713 of (x|y) == x for BICS pattern.
15715 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15717 PR tree-optimization/83605
15718 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15719 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15722 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15724 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15725 * config/epiphany/rtems.h: New file.
15727 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15728 Uros Bizjak <ubizjak@gmail.com>
15731 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15732 QIreg_operand instead of register_operand predicate.
15733 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15734 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15735 comments instead of -fmitigate[-_]rop.
15737 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15740 * cgraphunit.c (symbol_table::compile): Switch to text_section
15741 before calling assembly_start debug hook.
15742 * run-rtl-passes.c (run_rtl_passes): Likewise.
15745 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15747 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15748 range_int_cst_p rather than !symbolic_range_p before calling
15749 extract_range_from_multiplicative_op_1.
15751 2018-01-04 Jeff Law <law@redhat.com>
15753 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15754 redundant test in assertion.
15756 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15758 * doc/rtl.texi: Document machine_mode wrapper classes.
15760 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15762 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15763 using tree_to_uhwi.
15765 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15767 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15768 the VEC_PERM_EXPR fold to fail.
15770 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15773 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15774 to switched_sections.
15776 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15779 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15780 test for d.testing.
15782 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15785 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15786 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15788 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15791 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15792 is BLKmode and bitpos not zero or mode change is needed.
15794 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15797 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15800 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15803 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15804 instead of MULT rtx. Update all corresponding splitters.
15805 (*saddl_se): Ditto.
15806 (*ssub<modesuffix>): Ditto.
15807 (*ssubl_se): Ditto.
15808 (*cmp_sadd_di): Update split patterns.
15809 (*cmp_sadd_si): Ditto.
15810 (*cmp_sadd_sidi): Ditto.
15811 (*cmp_ssub_di): Ditto.
15812 (*cmp_ssub_si): Ditto.
15813 (*cmp_ssub_sidi): Ditto.
15814 * config/alpha/predicates.md (const23_operand): New predicate.
15815 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15816 Look for ASHIFT, not MULT inner operand.
15817 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15819 2018-01-04 Martin Liska <mliska@suse.cz>
15821 PR gcov-profile/83669
15822 * gcov.c (output_intermediate_file): Add version to intermediate
15824 * doc/gcov.texi: Document new field 'version' in intermediate
15825 file format. Fix location of '-k' option of gcov command.
15827 2018-01-04 Martin Liska <mliska@suse.cz>
15830 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15832 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15834 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15836 2018-01-03 Martin Sebor <msebor@redhat.com>
15838 PR tree-optimization/83655
15839 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15840 checking calls with invalid arguments.
15842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15844 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15845 (vectorizable_mask_load_store): Delete.
15846 (vectorizable_call): Return false for masked loads and stores.
15847 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15848 instead of gimple_assign_rhs1.
15849 (vectorizable_load): Handle IFN_MASK_LOAD.
15850 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15854 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15856 (vectorizable_mask_load_store): ...here.
15857 (vectorizable_load): ...and here.
15859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15861 * tree-vect-stmts.c (vect_build_all_ones_mask)
15862 (vect_build_zero_merge_argument): New functions, split out from...
15863 (vectorizable_load): ...here.
15865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15867 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15869 (vectorizable_mask_load_store): ...here.
15870 (vectorizable_store): ...and here.
15872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15874 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15876 (vectorizable_mask_load_store): ...here.
15878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15880 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15881 (vect_model_store_cost): Take a vec_load_store_type instead of a
15883 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15884 (vect_model_store_cost): Take a vec_load_store_type instead of a
15886 (vectorizable_mask_load_store): Update accordingly.
15887 (vectorizable_store): Likewise.
15888 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15892 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15893 IFN_MASK_LOAD calls here rather than...
15894 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15897 Alan Hayward <alan.hayward@arm.com>
15898 David Sherwood <david.sherwood@arm.com>
15900 * expmed.c (extract_bit_field_1): For vector extracts,
15901 fall back to extract_bit_field_as_subreg if vec_extract
15904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15905 Alan Hayward <alan.hayward@arm.com>
15906 David Sherwood <david.sherwood@arm.com>
15908 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15909 they are variable or constant sized.
15910 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15911 slots for constant-sized data.
15913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15914 Alan Hayward <alan.hayward@arm.com>
15915 David Sherwood <david.sherwood@arm.com>
15917 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15918 handling COND_EXPRs with boolean comparisons, try to find a better
15919 basis for the mask type than the boolean itself.
15921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15923 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15924 is calculated and how it can be overridden.
15925 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15926 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15928 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15932 Alan Hayward <alan.hayward@arm.com>
15933 David Sherwood <david.sherwood@arm.com>
15935 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15936 Remove the mode argument.
15937 (aarch64_simd_valid_immediate): Remove the mode and inverse
15939 * config/aarch64/iterators.md (bitsize): New iterator.
15940 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15941 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15942 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15943 aarch64_simd_valid_immediate.
15944 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15945 (aarch64_reg_or_bic_imm): Likewise.
15946 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15947 with an insn_type enum and msl with a modifier_type enum.
15948 Replace element_width with a scalar_mode. Change the shift
15949 to unsigned int. Add constructors for scalar_float_mode and
15950 scalar_int_mode elements.
15951 (aarch64_vect_float_const_representable_p): Delete.
15952 (aarch64_can_const_movi_rtx_p)
15953 (aarch64_simd_scalar_immediate_valid_for_move)
15954 (aarch64_simd_make_constant): Update call to
15955 aarch64_simd_valid_immediate.
15956 (aarch64_advsimd_valid_immediate_hs): New function.
15957 (aarch64_advsimd_valid_immediate): Likewise.
15958 (aarch64_simd_valid_immediate): Remove mode and inverse
15959 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15960 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15961 and aarch64_float_const_representable_p on the result.
15962 (aarch64_output_simd_mov_immediate): Remove mode argument.
15963 Update call to aarch64_simd_valid_immediate and use of
15964 simd_immediate_info.
15965 (aarch64_output_scalar_simd_mov_immediate): Update call
15968 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15969 Alan Hayward <alan.hayward@arm.com>
15970 David Sherwood <david.sherwood@arm.com>
15972 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15973 (mode_nunits): Likewise CONST_MODE_NUNITS.
15974 * machmode.def (ADJUST_NUNITS): Document.
15975 * genmodes.c (mode_data::need_nunits_adj): New field.
15976 (blank_mode): Update accordingly.
15977 (adj_nunits): New variable.
15978 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15980 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15981 listed in adj_nunits.
15982 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15983 listed in adj_nunits. Don't emit case statements for such modes.
15984 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15985 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15986 nothing if adj_nunits is nonnull.
15987 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15988 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15989 (emit_mode_fbit): Update use of print_maybe_const_decl.
15990 (emit_move_size): Likewise. Treat the array as non-const
15992 (emit_mode_adjustments): Handle adj_nunits.
15994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15996 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15997 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15998 (VECTOR_MODES): Use it.
15999 (make_vector_modes): Take the prefix as an argument.
16001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16002 Alan Hayward <alan.hayward@arm.com>
16003 David Sherwood <david.sherwood@arm.com>
16005 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
16006 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
16007 for MODE_VECTOR_BOOL.
16008 * machmode.def (VECTOR_BOOL_MODE): Document.
16009 * genmodes.c (VECTOR_BOOL_MODE): New macro.
16010 (make_vector_bool_mode): New function.
16011 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
16013 * lto-streamer-in.c (lto_input_mode_table): Likewise.
16014 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16016 * stor-layout.c (int_mode_for_mode): Likewise.
16017 * tree.c (build_vector_type_for_mode): Likewise.
16018 * varasm.c (output_constant_pool_2): Likewise.
16019 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16020 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16021 for MODE_VECTOR_BOOL.
16022 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16023 of mode class checks.
16024 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16025 instead of a list of mode class checks.
16026 (expand_vector_scalar_condition): Likewise.
16027 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16030 Alan Hayward <alan.hayward@arm.com>
16031 David Sherwood <david.sherwood@arm.com>
16033 * machmode.h (mode_size): Change from unsigned short to
16035 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16036 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16037 or if measurement_type is not polynomial.
16038 (fixed_size_mode::includes_p): Check for constant-sized modes.
16039 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16040 return a poly_uint16 rather than an unsigned short.
16041 (emit_mode_size): Change the type of mode_size from unsigned short
16042 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16043 (emit_mode_adjustments): Cope with polynomial vector sizes.
16044 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16046 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16048 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16049 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16050 * caller-save.c (setup_save_areas): Likewise.
16051 (replace_reg_with_saved_mem): Likewise.
16052 * calls.c (emit_library_call_value_1): Likewise.
16053 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16054 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16055 (gen_lowpart_for_combine): Likewise.
16056 * convert.c (convert_to_integer_1): Likewise.
16057 * cse.c (equiv_constant, cse_insn): Likewise.
16058 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16059 (cselib_subst_to_values): Likewise.
16060 * dce.c (word_dce_process_block): Likewise.
16061 * df-problems.c (df_word_lr_mark_ref): Likewise.
16062 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16063 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16064 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16065 (rtl_for_decl_location): Likewise.
16066 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16067 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16068 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16069 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16070 (expand_expr_real_1): Likewise.
16071 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16072 (pad_below): Likewise.
16073 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16074 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16075 * ira.c (get_subreg_tracking_sizes): Likewise.
16076 * ira-build.c (ira_create_allocno_objects): Likewise.
16077 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16078 (ira_sort_regnos_for_alter_reg): Likewise.
16079 * ira-costs.c (record_operand_costs): Likewise.
16080 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16081 (resolve_simple_move): Likewise.
16082 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16083 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16084 (lra_constraints): Likewise.
16085 (CONST_POOL_OK_P): Reject variable-sized modes.
16086 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16087 (add_pseudo_to_slot, lra_spill): Likewise.
16088 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16089 * optabs-query.c (get_best_extraction_insn): Likewise.
16090 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16091 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16092 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16093 * recog.c (offsettable_address_addr_space_p): Likewise.
16094 * regcprop.c (maybe_mode_change): Likewise.
16095 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16096 * regrename.c (build_def_use): Likewise.
16097 * regstat.c (dump_reg_info): Likewise.
16098 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16099 (find_reloads, find_reloads_subreg_address): Likewise.
16100 * reload1.c (eliminate_regs_1): Likewise.
16101 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16102 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16103 (simplify_binary_operation_1, simplify_subreg): Likewise.
16104 * targhooks.c (default_function_arg_padding): Likewise.
16105 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16106 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16107 (verify_gimple_assign_ternary): Likewise.
16108 * tree-inline.c (estimate_move_cost): Likewise.
16109 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16110 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16111 (get_address_cost_ainc): Likewise.
16112 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16113 (vect_supportable_dr_alignment): Likewise.
16114 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16115 (vectorizable_reduction): Likewise.
16116 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16117 (vectorizable_operation, vectorizable_load): Likewise.
16118 * tree.c (build_same_sized_truth_vector_type): Likewise.
16119 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16120 * var-tracking.c (emit_note_insn_var_location): Likewise.
16121 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16122 (ADDR_VEC_ALIGN): Likewise.
16124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16125 Alan Hayward <alan.hayward@arm.com>
16126 David Sherwood <david.sherwood@arm.com>
16128 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16130 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16131 or if measurement_type is polynomial.
16132 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16133 * combine.c (make_extraction): Likewise.
16134 * dse.c (find_shift_sequence): Likewise.
16135 * dwarf2out.c (mem_loc_descriptor): Likewise.
16136 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16137 (extract_bit_field, extract_low_bits): Likewise.
16138 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16139 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16140 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16141 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16142 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16143 * reload.c (find_reloads): Likewise.
16144 * reload1.c (alter_reg): Likewise.
16145 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16146 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16147 * tree-if-conv.c (predicate_mem_writes): Likewise.
16148 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16149 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16150 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16151 * valtrack.c (dead_debug_insert_temp): Likewise.
16152 * varasm.c (mergeable_constant_section): Likewise.
16153 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_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 * expr.c (expand_assignment): Cope with polynomial mode sizes
16160 when assigning to a CONCAT.
16162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16163 Alan Hayward <alan.hayward@arm.com>
16164 David Sherwood <david.sherwood@arm.com>
16166 * machmode.h (mode_precision): Change from unsigned short to
16168 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16170 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16171 or if measurement_type is not polynomial.
16172 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16173 in which the mode is already known to be a scalar_int_mode.
16174 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16175 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16177 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16178 for GET_MODE_PRECISION.
16179 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16180 for GET_MODE_PRECISION.
16181 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16183 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16184 (expand_field_assignment, make_extraction): Likewise.
16185 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16186 (get_last_value): Likewise.
16187 * convert.c (convert_to_integer_1): Likewise.
16188 * cse.c (cse_insn): Likewise.
16189 * expr.c (expand_expr_real_1): Likewise.
16190 * lra-constraints.c (simplify_operand_subreg): Likewise.
16191 * optabs-query.c (can_atomic_load_p): Likewise.
16192 * optabs.c (expand_atomic_load): Likewise.
16193 (expand_atomic_store): Likewise.
16194 * ree.c (combine_reaching_defs): Likewise.
16195 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16196 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16197 * tree.h (type_has_mode_precision_p): Likewise.
16198 * ubsan.c (instrument_si_overflow): Likewise.
16200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16201 Alan Hayward <alan.hayward@arm.com>
16202 David Sherwood <david.sherwood@arm.com>
16204 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16205 polynomial numbers of units.
16206 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16207 (valid_vector_subparts_p): New function.
16208 (build_vector_type): Remove temporary shim and take the number
16209 of units as a poly_uint64 rather than an int.
16210 (build_opaque_vector_type): Take the number of units as a
16211 poly_uint64 rather than an int.
16212 * tree.c (build_vector_from_ctor): Handle polynomial
16213 TYPE_VECTOR_SUBPARTS.
16214 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16215 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16216 (build_vector_from_val): If the number of units is variable,
16217 use build_vec_duplicate_cst for constant operands and
16218 VEC_DUPLICATE_EXPR otherwise.
16219 (make_vector_type): Remove temporary is_constant ().
16220 (build_vector_type, build_opaque_vector_type): Take the number of
16221 units as a poly_uint64 rather than an int.
16222 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16224 * cfgexpand.c (expand_debug_expr): Likewise.
16225 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16226 (store_constructor, expand_expr_real_1): Likewise.
16227 (const_scalar_mask_from_tree): Likewise.
16228 * fold-const-call.c (fold_const_reduction): Likewise.
16229 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16230 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16231 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16232 (fold_relational_const): Likewise.
16233 (native_interpret_vector): Likewise. Change the size from an
16234 int to an unsigned int.
16235 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16236 TYPE_VECTOR_SUBPARTS.
16237 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16238 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16239 duplicating a non-constant operand into a variable-length vector.
16240 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16241 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16242 * ipa-icf.c (sem_variable::equals): Likewise.
16243 * match.pd: Likewise.
16244 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16245 * print-tree.c (print_node): Likewise.
16246 * stor-layout.c (layout_type): Likewise.
16247 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16248 * tree-cfg.c (verify_gimple_comparison): Likewise.
16249 (verify_gimple_assign_binary): Likewise.
16250 (verify_gimple_assign_ternary): Likewise.
16251 (verify_gimple_assign_single): Likewise.
16252 * tree-pretty-print.c (dump_generic_node): Likewise.
16253 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16254 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16255 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16256 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16257 (vect_shift_permute_load_chain): Likewise.
16258 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16259 (expand_vector_condition, optimize_vector_constructor): Likewise.
16260 (lower_vec_perm, get_compute_type): Likewise.
16261 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16262 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16263 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16264 (vect_recog_mask_conversion_pattern): Likewise.
16265 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16266 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16267 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16268 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16269 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16270 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16271 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16272 (supportable_widening_operation): Likewise.
16273 (supportable_narrowing_operation): Likewise.
16274 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16276 * varasm.c (output_constant): Likewise.
16278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16279 Alan Hayward <alan.hayward@arm.com>
16280 David Sherwood <david.sherwood@arm.com>
16282 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16283 so that both the length == 3 and length != 3 cases set up their
16284 own permute vectors. Add comments explaining why we know the
16285 number of elements is constant.
16286 (vect_permute_load_chain): Likewise.
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_nunits): Change from unsigned char to
16294 (ONLY_FIXED_SIZE_MODES): New macro.
16295 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16296 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16297 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16299 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16300 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16301 or if measurement_type is not polynomial.
16302 * genmodes.c (ZERO_COEFFS): New macro.
16303 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16305 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16306 Use ZERO_COEFFS when emitting initializers.
16307 * data-streamer.h (bp_pack_poly_value): New function.
16308 (bp_unpack_poly_value): Likewise.
16309 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16310 for GET_MODE_NUNITS.
16311 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16312 for GET_MODE_NUNITS.
16313 * tree.c (make_vector_type): Remove temporary shim and make
16314 the real function take the number of units as a poly_uint64
16315 rather than an int.
16316 (build_vector_type_for_mode): Handle polynomial nunits.
16317 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16318 * emit-rtl.c (const_vec_series_p_1): Likewise.
16319 (gen_rtx_CONST_VECTOR): Likewise.
16320 * fold-const.c (test_vec_duplicate_folding): Likewise.
16321 * genrecog.c (validate_pattern): Likewise.
16322 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16323 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16324 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16325 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16326 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16327 * rtlanal.c (subreg_get_info): Likewise.
16328 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16329 (vect_grouped_load_supported): Likewise.
16330 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16331 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16332 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16333 (simplify_const_unary_operation, simplify_binary_operation_1)
16334 (simplify_const_binary_operation, simplify_ternary_operation)
16335 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16336 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16337 instead of CONST_VECTOR_NUNITS.
16338 * varasm.c (output_constant_pool_2): Likewise.
16339 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16340 explicit-encoded elements in the XVEC for variable-length vectors.
16342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16344 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16347 Alan Hayward <alan.hayward@arm.com>
16348 David Sherwood <david.sherwood@arm.com>
16350 * coretypes.h (fixed_size_mode): Declare.
16351 (fixed_size_mode_pod): New typedef.
16352 * builtins.h (target_builtins::x_apply_args_mode)
16353 (target_builtins::x_apply_result_mode): Change type to
16354 fixed_size_mode_pod.
16355 * builtins.c (apply_args_size, apply_result_size, result_vector)
16356 (expand_builtin_apply_args_1, expand_builtin_apply)
16357 (expand_builtin_return): Update accordingly.
16359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16361 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16362 * cselib.c (cselib_hash_rtx): Likewise.
16363 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16364 CONST_VECTOR encoding.
16366 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16367 Jeff Law <law@redhat.com>
16370 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16371 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16372 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16373 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16376 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16377 explicitly probe *sp in a noreturn function if there were any callee
16378 register saves or frame pointer is needed.
16380 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16383 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16384 BLKmode for ternary, binary or unary expressions.
16387 * var-tracking.c (delete_vta_debug_insn): New inline function.
16388 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16389 insns from get_insns () to NULL instead of each bb separately.
16390 Use delete_vta_debug_insn. No longer static.
16391 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16392 delete_vta_debug_insns callers.
16393 * rtl.h (delete_vta_debug_insns): Declare.
16394 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16395 instead of variable_tracking_main.
16397 2018-01-03 Martin Sebor <msebor@redhat.com>
16399 PR tree-optimization/83603
16400 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16401 arguments past the endof the argument list in functions declared
16402 without a prototype.
16403 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16404 Avoid checking when arguments are null.
16406 2018-01-03 Martin Sebor <msebor@redhat.com>
16409 * doc/extend.texi (attribute const): Fix a typo.
16410 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16411 issuing -Wsuggest-attribute for void functions.
16413 2018-01-03 Martin Sebor <msebor@redhat.com>
16415 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16416 offset_int::from instead of wide_int::to_shwi.
16417 (maybe_diag_overlap): Remove assertion.
16418 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16419 * gimple-ssa-sprintf.c (format_directive): Same.
16420 (parse_directive): Same.
16421 (sprintf_dom_walker::compute_format_length): Same.
16422 (try_substitute_return_value): Same.
16424 2018-01-03 Jeff Law <law@redhat.com>
16426 PR middle-end/83654
16427 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16428 non-constant residual for zero at runtime and avoid probing in
16429 that case. Reorganize code for trailing problem to mirror handling
16432 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16434 PR tree-optimization/83501
16435 * tree-ssa-strlen.c (get_string_cst): New.
16436 (handle_char_store): Call get_string_cst.
16438 2018-01-03 Martin Liska <mliska@suse.cz>
16440 PR tree-optimization/83593
16441 * tree-ssa-strlen.c: Include tree-cfg.h.
16442 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16443 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16444 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16446 (strlen_dom_walker::before_dom_children): Call
16447 gimple_purge_dead_eh_edges. Dump tranformation with details
16449 (strlen_dom_walker::before_dom_children): Update call by adding
16450 new argument cleanup_eh.
16451 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16453 2018-01-03 Martin Liska <mliska@suse.cz>
16456 * cif-code.def (VARIADIC_THUNK): New enum value.
16457 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16460 2018-01-03 Jan Beulich <jbeulich@suse.com>
16462 * sse.md (mov<mode>_internal): Tighten condition for when to use
16463 vmovdqu<ssescalarsize> for TI and OI modes.
16465 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16467 Update copyright years.
16469 2018-01-03 Martin Liska <mliska@suse.cz>
16472 * ipa-visibility.c (function_and_variable_visibility): Skip
16473 functions with noipa attribure.
16475 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16477 * gcc.c (process_command): Update copyright notice dates.
16478 * gcov-dump.c (print_version): Ditto.
16479 * gcov.c (print_version): Ditto.
16480 * gcov-tool.c (print_version): Ditto.
16481 * gengtype.c (create_file): Ditto.
16482 * doc/cpp.texi: Bump @copying's copyright year.
16483 * doc/cppinternals.texi: Ditto.
16484 * doc/gcc.texi: Ditto.
16485 * doc/gccint.texi: Ditto.
16486 * doc/gcov.texi: Ditto.
16487 * doc/install.texi: Ditto.
16488 * doc/invoke.texi: Ditto.
16490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16492 * vector-builder.h (vector_builder::m_full_nelts): Change from
16493 unsigned int to poly_uint64.
16494 (vector_builder::full_nelts): Update prototype accordingly.
16495 (vector_builder::new_vector): Likewise.
16496 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16497 (vector_builder::operator ==): Likewise.
16498 (vector_builder::finalize): Likewise.
16499 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16500 Take the number of elements as a poly_uint64 rather than an
16502 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16503 from unsigned int to poly_uint64.
16504 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16505 (vec_perm_indices::new_vector): Likewise.
16506 (vec_perm_indices::length): Likewise.
16507 (vec_perm_indices::nelts_per_input): Likewise.
16508 (vec_perm_indices::input_nelts): Likewise.
16509 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16510 number of elements per input as a poly_uint64 rather than an
16511 unsigned int. Use the original encoding for variable-length
16512 vectors, rather than clamping each individual element.
16513 For the second and subsequent elements in each pattern,
16514 clamp the step and base before clamping their sum.
16515 (vec_perm_indices::series_p): Handle polynomial element counts.
16516 (vec_perm_indices::all_in_range_p): Likewise.
16517 (vec_perm_indices_to_tree): Likewise.
16518 (vec_perm_indices_to_rtx): Likewise.
16519 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16520 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16521 (tree_vector_builder::new_binary_operation): Handle polynomial
16522 element counts. Return false if we need to know the number
16523 of elements at compile time.
16524 * fold-const.c (fold_vec_perm): Punt if the number of elements
16525 isn't known at compile time.
16527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16529 * vec-perm-indices.h (vec_perm_builder): Change element type
16530 from HOST_WIDE_INT to poly_int64.
16531 (vec_perm_indices::element_type): Update accordingly.
16532 (vec_perm_indices::clamp): Handle polynomial element_types.
16533 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16534 (vec_perm_indices::all_in_range_p): Likewise.
16535 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16537 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16538 polynomial vec_perm_indices element types.
16539 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16540 * fold-const.c (fold_vec_perm): Likewise.
16541 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16542 * tree-vect-generic.c (lower_vec_perm): Likewise.
16543 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16544 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16545 element type to HOST_WIDE_INT.
16547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16548 Alan Hayward <alan.hayward@arm.com>
16549 David Sherwood <david.sherwood@arm.com>
16551 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16552 rather than an int. Use plus_constant.
16553 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16554 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16557 Alan Hayward <alan.hayward@arm.com>
16558 David Sherwood <david.sherwood@arm.com>
16560 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16561 a HOST_WIDE_INT to a poly_int64.
16563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16564 Alan Hayward <alan.hayward@arm.com>
16565 David Sherwood <david.sherwood@arm.com>
16567 * calls.c (load_register_parameters): Cope with polynomial
16568 mode sizes. Require a constant size for BLKmode parameters
16569 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16570 forces a parameter to be padded at the lsb end in order to
16571 fill a complete number of words, require the parameter size
16572 to be ordered wrt UNITS_PER_WORD.
16574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16575 Alan Hayward <alan.hayward@arm.com>
16576 David Sherwood <david.sherwood@arm.com>
16578 * reload1.c (spill_stack_slot_width): Change element type
16579 from unsigned int to poly_uint64_pod.
16580 (alter_reg): Treat mode sizes as polynomial.
16582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16583 Alan Hayward <alan.hayward@arm.com>
16584 David Sherwood <david.sherwood@arm.com>
16586 * reload.c (complex_word_subreg_p): New function.
16587 (reload_inner_reg_of_subreg, push_reload): Use it.
16589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16590 Alan Hayward <alan.hayward@arm.com>
16591 David Sherwood <david.sherwood@arm.com>
16593 * lra-constraints.c (process_alt_operands): Reject matched
16594 operands whose sizes aren't ordered.
16595 (match_reload): Refer to this check here.
16597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16598 Alan Hayward <alan.hayward@arm.com>
16599 David Sherwood <david.sherwood@arm.com>
16601 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16602 that the mode size is in the set {1, 2, 4, 8, 16}.
16604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16605 Alan Hayward <alan.hayward@arm.com>
16606 David Sherwood <david.sherwood@arm.com>
16608 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16609 Use plus_constant instead of gen_rtx_PLUS.
16611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16612 Alan Hayward <alan.hayward@arm.com>
16613 David Sherwood <david.sherwood@arm.com>
16615 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16616 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16617 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16618 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16619 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16620 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16621 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16622 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16623 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16624 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16626 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16627 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16628 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16629 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16630 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16631 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16632 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16633 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16634 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16635 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16637 * expr.c (emit_move_resolve_push): Treat the input and result
16638 of PUSH_ROUNDING as a poly_int64.
16639 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16640 (emit_push_insn): Likewise.
16641 * lra-eliminations.c (mark_not_eliminable): Likewise.
16642 * recog.c (push_operand): Likewise.
16643 * reload1.c (elimination_effects): Likewise.
16644 * rtlanal.c (nonzero_bits1): Likewise.
16645 * calls.c (store_one_arg): Likewise. Require the padding to be
16646 known at compile time.
16648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16649 Alan Hayward <alan.hayward@arm.com>
16650 David Sherwood <david.sherwood@arm.com>
16652 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16653 Use plus_constant instead of gen_rtx_PLUS.
16655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16656 Alan Hayward <alan.hayward@arm.com>
16657 David Sherwood <david.sherwood@arm.com>
16659 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16660 rather than an int.
16662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16663 Alan Hayward <alan.hayward@arm.com>
16664 David Sherwood <david.sherwood@arm.com>
16666 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16667 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16668 via stack temporaries. Treat the mode size as polynomial too.
16670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16671 Alan Hayward <alan.hayward@arm.com>
16672 David Sherwood <david.sherwood@arm.com>
16674 * expr.c (expand_expr_real_2): When handling conversions involving
16675 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16676 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16677 as a poly_uint64 too.
16679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16680 Alan Hayward <alan.hayward@arm.com>
16681 David Sherwood <david.sherwood@arm.com>
16683 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16686 Alan Hayward <alan.hayward@arm.com>
16687 David Sherwood <david.sherwood@arm.com>
16689 * combine.c (can_change_dest_mode): Handle polynomial
16690 REGMODE_NATURAL_SIZE.
16691 * expmed.c (store_bit_field_1): Likewise.
16692 * expr.c (store_constructor): Likewise.
16693 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16694 and polynomial REGMODE_NATURAL_SIZE.
16695 (gen_lowpart_common): Likewise.
16696 * reginfo.c (record_subregs_of_mode): Likewise.
16697 * rtlanal.c (read_modify_subreg_p): Likewise.
16699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16700 Alan Hayward <alan.hayward@arm.com>
16701 David Sherwood <david.sherwood@arm.com>
16703 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16704 numbers of elements.
16706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16707 Alan Hayward <alan.hayward@arm.com>
16708 David Sherwood <david.sherwood@arm.com>
16710 * match.pd: Cope with polynomial numbers of vector elements.
16712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16713 Alan Hayward <alan.hayward@arm.com>
16714 David Sherwood <david.sherwood@arm.com>
16716 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16717 in a POINTER_PLUS_EXPR.
16719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16720 Alan Hayward <alan.hayward@arm.com>
16721 David Sherwood <david.sherwood@arm.com>
16723 * omp-simd-clone.c (simd_clone_subparts): New function.
16724 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16725 (ipa_simd_modify_function_body): Likewise.
16727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16728 Alan Hayward <alan.hayward@arm.com>
16729 David Sherwood <david.sherwood@arm.com>
16731 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16732 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16733 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16734 (expand_vector_condition, vector_element): Likewise.
16735 (subparts_gt): New function.
16736 (get_compute_type): Use subparts_gt.
16737 (count_type_subparts): Delete.
16738 (expand_vector_operations_1): Use subparts_gt instead of
16739 count_type_subparts.
16741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16742 Alan Hayward <alan.hayward@arm.com>
16743 David Sherwood <david.sherwood@arm.com>
16745 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16746 (vect_compile_time_alias): ...this new function. Do the calculation
16747 on poly_ints rather than trees.
16748 (vect_prune_runtime_alias_test_list): Update call accordingly.
16750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16751 Alan Hayward <alan.hayward@arm.com>
16752 David Sherwood <david.sherwood@arm.com>
16754 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16756 (vect_schedule_slp_instance): Likewise.
16758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16759 Alan Hayward <alan.hayward@arm.com>
16760 David Sherwood <david.sherwood@arm.com>
16762 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16763 constant and extern definitions for variable-length vectors.
16764 (vect_get_constant_vectors): Note that the number of units
16765 is known to be constant.
16767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16768 Alan Hayward <alan.hayward@arm.com>
16769 David Sherwood <david.sherwood@arm.com>
16771 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16772 of units as polynomial. Choose between WIDE and NARROW based
16775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16776 Alan Hayward <alan.hayward@arm.com>
16777 David Sherwood <david.sherwood@arm.com>
16779 * tree-vect-stmts.c (simd_clone_subparts): New function.
16780 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16783 Alan Hayward <alan.hayward@arm.com>
16784 David Sherwood <david.sherwood@arm.com>
16786 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16787 vectors as polynomial. Use build_index_vector for
16788 IFN_GOMP_SIMD_LANE.
16790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16791 Alan Hayward <alan.hayward@arm.com>
16792 David Sherwood <david.sherwood@arm.com>
16794 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16795 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16796 for variable-length vectors.
16797 (vectorizable_mask_load_store): Treat the number of units as
16798 polynomial, asserting that it is constant if the condition has
16799 already been enforced.
16800 (vectorizable_store, vectorizable_load): Likewise.
16802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16803 Alan Hayward <alan.hayward@arm.com>
16804 David Sherwood <david.sherwood@arm.com>
16806 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16807 of units as polynomial. Punt if we can't tell at compile time
16808 which vector contains the final result.
16810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16811 Alan Hayward <alan.hayward@arm.com>
16812 David Sherwood <david.sherwood@arm.com>
16814 * tree-vect-loop.c (vectorizable_induction): Treat the number
16815 of units as polynomial. Punt on SLP inductions. Use an integer
16816 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16817 cast of such a series for variable-length floating-point
16820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16821 Alan Hayward <alan.hayward@arm.com>
16822 David Sherwood <david.sherwood@arm.com>
16824 * tree.h (build_index_vector): Declare.
16825 * tree.c (build_index_vector): New function.
16826 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16827 of units as polynomial, forcibly converting it to a constant if
16828 vectorizable_reduction has already enforced the condition.
16829 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16830 to create a {1,2,3,...} vector.
16831 (vectorizable_reduction): Treat the number of units as polynomial.
16832 Choose vectype_in based on the largest scalar element size rather
16833 than the smallest number of units. Enforce the restrictions
16836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16837 Alan Hayward <alan.hayward@arm.com>
16838 David Sherwood <david.sherwood@arm.com>
16840 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16841 number of units as polynomial.
16843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16844 Alan Hayward <alan.hayward@arm.com>
16845 David Sherwood <david.sherwood@arm.com>
16847 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16848 * target.def (autovectorize_vector_sizes): Return the vector sizes
16849 by pointer, using vector_sizes rather than a bitmask.
16850 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16851 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16852 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16854 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16855 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16856 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16857 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16858 * omp-general.c (omp_max_vf): Likewise.
16859 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16860 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16861 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16862 * tree-vect-slp.c (vect_slp_bb): Likewise.
16863 * doc/tm.texi: Regenerate.
16864 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16866 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16867 the vector size as a poly_uint64 rather than an unsigned int.
16868 (current_vector_size): Change from an unsigned int to a poly_uint64.
16869 (get_vectype_for_scalar_type): Update accordingly.
16870 * tree.h (build_truth_vector_type): Take the size and number of
16871 units as a poly_uint64 rather than an unsigned int.
16872 (build_vector_type): Add a temporary overload that takes
16873 the number of units as a poly_uint64 rather than an unsigned int.
16874 * tree.c (make_vector_type): Likewise.
16875 (build_truth_vector_type): Take the number of units as a poly_uint64
16876 rather than an unsigned int.
16878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16879 Alan Hayward <alan.hayward@arm.com>
16880 David Sherwood <david.sherwood@arm.com>
16882 * target.def (get_mask_mode): Take the number of units and length
16883 as poly_uint64s rather than unsigned ints.
16884 * targhooks.h (default_get_mask_mode): Update accordingly.
16885 * targhooks.c (default_get_mask_mode): Likewise.
16886 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16887 * doc/tm.texi: Regenerate.
16889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16890 Alan Hayward <alan.hayward@arm.com>
16891 David Sherwood <david.sherwood@arm.com>
16893 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16894 * omp-general.c (omp_max_vf): Likewise.
16895 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16896 (expand_omp_simd): Handle polynomial safelen.
16897 * omp-low.c (omplow_simd_context): Add a default constructor.
16898 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16899 (lower_rec_simd_input_clauses): Update accordingly.
16900 (lower_rec_input_clauses): Likewise.
16902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16903 Alan Hayward <alan.hayward@arm.com>
16904 David Sherwood <david.sherwood@arm.com>
16906 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16907 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16908 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16909 (vect_analyze_slp_cost): Likewise.
16910 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16911 (vect_model_load_cost): Likewise.
16913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16914 Alan Hayward <alan.hayward@arm.com>
16915 David Sherwood <david.sherwood@arm.com>
16917 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16918 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16919 from an unsigned int * to a poly_uint64_pod *.
16920 (calculate_unrolling_factor): New function.
16921 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16924 Alan Hayward <alan.hayward@arm.com>
16925 David Sherwood <david.sherwood@arm.com>
16927 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16928 from an unsigned int to a poly_uint64.
16929 (_loop_vec_info::slp_unrolling_factor): Likewise.
16930 (_loop_vec_info::vectorization_factor): Change from an int
16932 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16933 (vect_get_num_vectors): New function.
16934 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16935 (vect_get_num_copies): Use vect_get_num_vectors.
16936 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16937 to an unsigned int *.
16938 (vect_analyze_data_refs): Change min_vf from an int * to a
16940 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16941 than an unsigned HOST_WIDE_INT.
16942 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16943 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16944 to an unsigned int *.
16945 (vect_analyze_data_ref_dependences): Likewise.
16946 (vect_compute_data_ref_alignment): Handle polynomial vf.
16947 (vect_enhance_data_refs_alignment): Likewise.
16948 (vect_prune_runtime_alias_test_list): Likewise.
16949 (vect_shift_permute_load_chain): Likewise.
16950 (vect_supportable_dr_alignment): Likewise.
16951 (dependence_distance_ge_vf): Take the vectorization factor as a
16952 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16953 (vect_analyze_data_refs): Change min_vf from an int * to a
16955 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16956 vfm1 as a poly_uint64 rather than an int. Make the same change
16957 for the returned bound_scalar.
16958 (vect_gen_vector_loop_niters): Handle polynomial vf.
16959 (vect_do_peeling): Likewise. Update call to
16960 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16961 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16963 * tree-vect-loop.c (vect_determine_vectorization_factor)
16964 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16965 (vect_get_known_peeling_cost): Likewise.
16966 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16967 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16968 (vect_transform_loop): Likewise. Use the lowest possible VF when
16969 updating the upper bounds of the loop.
16970 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16971 rather than an int.
16972 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16973 polynomial unroll factors.
16974 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16975 (vect_make_slp_decision): Likewise.
16976 (vect_supported_load_permutation_p): Likewise, and polynomial
16978 (vect_analyze_slp_cost): Handle polynomial vf.
16979 (vect_slp_analyze_node_operations): Likewise.
16980 (vect_slp_analyze_bb_1): Likewise.
16981 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16982 than an unsigned HOST_WIDE_INT.
16983 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16984 (vectorizable_load): Handle polynomial vf.
16985 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16987 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16990 Alan Hayward <alan.hayward@arm.com>
16991 David Sherwood <david.sherwood@arm.com>
16993 * match.pd: Handle bit operations involving three constants
16994 and try to fold one pair.
16996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16998 * tree-vect-loop-manip.c: Include gimple-fold.h.
16999 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
17000 niters_maybe_zero parameters. Handle other cases besides a step of 1.
17001 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
17002 Add a path that uses a step of VF instead of 1, but disable it
17004 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
17005 and niters_no_overflow parameters. Update calls to
17006 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
17007 Create a new SSA name if the latter choses to use a ste other
17008 than zero, and return it via niters_vector_mult_vf_var.
17009 * tree-vect-loop.c (vect_transform_loop): Update calls to
17010 vect_do_peeling, vect_gen_vector_loop_niters and
17011 slpeel_make_loop_iterate_ntimes.
17012 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
17013 (vect_gen_vector_loop_niters): Update declarations after above changes.
17015 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17017 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17018 128-bit round to integer instructions.
17019 (ceil<mode>2): Likewise.
17020 (btrunc<mode>2): Likewise.
17021 (round<mode>2): Likewise.
17023 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17025 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17026 unaligned VSX load/store on P8/P9.
17027 (expand_block_clear): Allow the use of unaligned VSX
17028 load/store on P8/P9.
17030 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17032 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17034 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17035 swap associated with both a load and a store.
17037 2018-01-02 Andrew Waterman <andrew@sifive.com>
17039 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17040 * config/riscv/riscv.md (clear_cache): Use it.
17042 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17044 * web.c: Remove out-of-date comment.
17046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17048 * expr.c (fixup_args_size_notes): Check that any existing
17049 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17050 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17051 (emit_single_push_insn): ...here.
17053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17055 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17056 (const_vector_encoded_nelts): New function.
17057 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17058 (const_vector_int_elt, const_vector_elt): Declare.
17059 * emit-rtl.c (const_vector_int_elt_1): New function.
17060 (const_vector_elt): Likewise.
17061 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17062 of CONST_VECTOR_ELT.
17064 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17066 * expr.c: Include rtx-vector-builder.h.
17067 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17068 directly on the tree encoding.
17069 (const_vector_from_tree): Likewise.
17070 * optabs.c: Include rtx-vector-builder.h.
17071 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17072 sequence of "u" values.
17073 * vec-perm-indices.c: Include rtx-vector-builder.h.
17074 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17075 directly on the vec_perm_indices encoding.
17077 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17079 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17080 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17081 * rtx-vector-builder.h: New file.
17082 * rtx-vector-builder.c: Likewise.
17083 * rtl.h (rtx_def::u2): Add a const_vector field.
17084 (CONST_VECTOR_NPATTERNS): New macro.
17085 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17086 (CONST_VECTOR_DUPLICATE_P): Likewise.
17087 (CONST_VECTOR_STEPPED_P): Likewise.
17088 (CONST_VECTOR_ENCODED_ELT): Likewise.
17089 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17090 (unwrap_const_vec_duplicate): Likewise.
17091 (const_vec_series_p): Check for a non-duplicated vector encoding.
17092 Say that the function only returns true for integer vectors.
17093 * emit-rtl.c: Include rtx-vector-builder.h.
17094 (gen_const_vec_duplicate_1): Delete.
17095 (gen_const_vector): Call gen_const_vec_duplicate instead of
17096 gen_const_vec_duplicate_1.
17097 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17098 (gen_const_vec_duplicate): Use rtx_vector_builder.
17099 (gen_const_vec_series): Likewise.
17100 (gen_rtx_CONST_VECTOR): Likewise.
17101 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17102 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17103 Build a new vector rather than modifying a CONST_VECTOR in-place.
17104 (handle_special_swappables): Update call accordingly.
17105 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17106 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17107 Build a new vector rather than modifying a CONST_VECTOR in-place.
17108 (handle_special_swappables): Update call accordingly.
17110 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17112 * simplify-rtx.c (simplify_const_binary_operation): Use
17113 CONST_VECTOR_ELT instead of XVECEXP.
17115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17117 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17118 the selector elements to be different from the data elements
17119 if the selector is a VECTOR_CST.
17120 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17121 ssizetype for the selector.
17123 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17125 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17126 before testing each element individually.
17127 * tree-vect-generic.c (lower_vec_perm): Likewise.
17129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17131 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17132 * selftest-run-tests.c (selftest::run_tests): Call it.
17133 * vector-builder.h (vector_builder::operator ==): New function.
17134 (vector_builder::operator !=): Likewise.
17135 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17136 (vec_perm_indices::all_from_input_p): New function.
17137 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17138 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17139 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17140 instead of reading the VECTOR_CST directly. Detect whether both
17141 vector inputs are the same before constructing the vec_perm_indices,
17142 and update the number of inputs argument accordingly. Use the
17143 utility functions added above. Only construct sel2 if we need to.
17145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17147 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17148 the broadcast of the low byte.
17149 (expand_mult_highpart): Use an explicit encoding for the permutes.
17150 * optabs-query.c (can_mult_highpart_p): Likewise.
17151 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17152 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17153 (vectorizable_bswap): Likewise.
17154 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17155 explicit encoding for the power-of-2 permutes.
17156 (vect_permute_store_chain): Likewise.
17157 (vect_grouped_load_supported): Likewise.
17158 (vect_permute_load_chain): Likewise.
17160 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17162 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17163 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17164 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17165 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17166 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17167 (vect_gen_perm_mask_any): Likewise.
17169 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17171 * int-vector-builder.h: New file.
17172 * vec-perm-indices.h: Include int-vector-builder.h.
17173 (vec_perm_indices): Redefine as an int_vector_builder.
17174 (auto_vec_perm_indices): Delete.
17175 (vec_perm_builder): Redefine as a stand-alone class.
17176 (vec_perm_indices::vec_perm_indices): New function.
17177 (vec_perm_indices::clamp): Likewise.
17178 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17179 (vec_perm_indices::new_vector): New function.
17180 (vec_perm_indices::new_expanded_vector): Update for new
17181 vec_perm_indices class.
17182 (vec_perm_indices::rotate_inputs): New function.
17183 (vec_perm_indices::all_in_range_p): Operate directly on the
17184 encoded form, without computing elided elements.
17185 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17186 encoding. Update for new vec_perm_indices class.
17187 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17188 the given vec_perm_builder.
17189 (expand_vec_perm_var): Update vec_perm_builder constructor.
17190 (expand_mult_highpart): Use vec_perm_builder instead of
17191 auto_vec_perm_indices.
17192 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17193 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17194 or double series encoding as appropriate.
17195 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17196 vec_perm_indices instead of auto_vec_perm_indices.
17197 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17198 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17199 (vect_permute_store_chain): Likewise.
17200 (vect_grouped_load_supported): Likewise.
17201 (vect_permute_load_chain): Likewise.
17202 (vect_shift_permute_load_chain): Likewise.
17203 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17204 (vect_transform_slp_perm_load): Likewise.
17205 (vect_schedule_slp_instance): Likewise.
17206 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17207 (vectorizable_mask_load_store): Likewise.
17208 (vectorizable_bswap): Likewise.
17209 (vectorizable_store): Likewise.
17210 (vectorizable_load): Likewise.
17211 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17212 vec_perm_indices instead of auto_vec_perm_indices. Use
17213 tree_to_vec_perm_builder to read the vector from a tree.
17214 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17215 vec_perm_builder instead of a vec_perm_indices.
17216 (have_whole_vector_shift): Use vec_perm_builder and
17217 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17218 truncation to calc_vec_perm_mask_for_shift.
17219 (vect_create_epilog_for_reduction): Likewise.
17220 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17221 from auto_vec_perm_indices to vec_perm_indices.
17222 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17223 instead of changing individual elements.
17224 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17225 the vector in d.perm.
17226 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17227 from auto_vec_perm_indices to vec_perm_indices.
17228 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17229 instead of changing individual elements.
17230 (arm_vectorize_vec_perm_const): Use new_vector to install
17231 the vector in d.perm.
17232 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17233 Update vec_perm_builder constructor.
17234 (rs6000_expand_interleave): Likewise.
17235 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17236 (rs6000_expand_interleave): Likewise.
17238 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17240 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17241 to qimode could truncate the indices.
17242 * optabs.c (expand_vec_perm_var): Likewise.
17244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17246 * Makefile.in (OBJS): Add vec-perm-indices.o.
17247 * vec-perm-indices.h: New file.
17248 * vec-perm-indices.c: Likewise.
17249 * target.h (vec_perm_indices): Replace with a forward class
17251 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17252 * optabs.h: Include vec-perm-indices.h.
17253 (expand_vec_perm): Delete.
17254 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17255 (expand_vec_perm_const): Declare.
17256 * target.def (vec_perm_const_ok): Replace with...
17257 (vec_perm_const): ...this new hook.
17258 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17259 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17260 * doc/tm.texi: Regenerate.
17261 * optabs.def (vec_perm_const): Delete.
17262 * doc/md.texi (vec_perm_const): Likewise.
17263 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17264 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17265 expand_vec_perm for constant permutation vectors. Assert that
17266 the mode of variable permutation vectors is the integer equivalent
17267 of the mode that is being permuted.
17268 * optabs-query.h (selector_fits_mode_p): Declare.
17269 * optabs-query.c: Include vec-perm-indices.h.
17270 (selector_fits_mode_p): New function.
17271 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17272 is defined, instead of checking whether the vec_perm_const_optab
17273 exists. Use targetm.vectorize.vec_perm_const instead of
17274 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17275 fit in the vector mode before using a variable permute.
17276 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17277 vec_perm_indices instead of an rtx.
17278 (expand_vec_perm): Replace with...
17279 (expand_vec_perm_const): ...this new function. Take the selector
17280 as a vec_perm_indices rather than an rtx. Also take the mode of
17281 the selector. Update call to shift_amt_for_vec_perm_mask.
17282 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17283 Use vec_perm_indices::new_expanded_vector to expand the original
17284 selector into bytes. Check whether the indices fit in the vector
17285 mode before using a variable permute.
17286 (expand_vec_perm_var): Make global.
17287 (expand_mult_highpart): Use expand_vec_perm_const.
17288 * fold-const.c: Includes vec-perm-indices.h.
17289 * tree-ssa-forwprop.c: Likewise.
17290 * tree-vect-data-refs.c: Likewise.
17291 * tree-vect-generic.c: Likewise.
17292 * tree-vect-loop.c: Likewise.
17293 * tree-vect-slp.c: Likewise.
17294 * tree-vect-stmts.c: Likewise.
17295 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17297 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17298 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17299 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17300 (aarch64_vectorize_vec_perm_const): ...this new function.
17301 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17302 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17303 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17304 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17305 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17306 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17307 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17309 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17310 check for NEON modes.
17311 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17312 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17313 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17314 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17316 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17317 the old VEC_PERM_CONST conditions.
17318 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17319 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17320 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17321 (ia64_vectorize_vec_perm_const_ok): Merge into...
17322 (ia64_vectorize_vec_perm_const): ...this new function.
17323 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17324 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17325 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17326 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17327 * config/mips/mips.c (mips_expand_vec_perm_const)
17328 (mips_vectorize_vec_perm_const_ok): Merge into...
17329 (mips_vectorize_vec_perm_const): ...this new function.
17330 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17331 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17332 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17333 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17334 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17335 (rs6000_expand_vec_perm_const): Delete.
17336 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17338 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17339 (altivec_expand_vec_perm_const_le): Take each operand individually.
17340 Operate on constant selectors rather than rtxes.
17341 (altivec_expand_vec_perm_const): Likewise. Update call to
17342 altivec_expand_vec_perm_const_le.
17343 (rs6000_expand_vec_perm_const): Delete.
17344 (rs6000_vectorize_vec_perm_const_ok): Delete.
17345 (rs6000_vectorize_vec_perm_const): New function.
17346 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17347 an element count and rtx array.
17348 (rs6000_expand_extract_even): Update call accordingly.
17349 (rs6000_expand_interleave): Likewise.
17350 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17351 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17352 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17353 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17354 (rs6000_expand_vec_perm_const): Delete.
17355 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17356 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17357 (altivec_expand_vec_perm_const_le): Take each operand individually.
17358 Operate on constant selectors rather than rtxes.
17359 (altivec_expand_vec_perm_const): Likewise. Update call to
17360 altivec_expand_vec_perm_const_le.
17361 (rs6000_expand_vec_perm_const): Delete.
17362 (rs6000_vectorize_vec_perm_const_ok): Delete.
17363 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17364 reference to the SPE evmerge intructions.
17365 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17366 an element count and rtx array.
17367 (rs6000_expand_extract_even): Update call accordingly.
17368 (rs6000_expand_interleave): Likewise.
17369 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17370 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17372 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17374 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17376 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17377 vector mode and that that mode matches the mode of the data
17379 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17380 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17381 directly using expand_vec_perm_1 when forcing selectors into
17383 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17385 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17387 * optabs-query.h (can_vec_perm_p): Delete.
17388 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17389 * optabs-query.c (can_vec_perm_p): Split into...
17390 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17391 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17392 particular selector is valid.
17393 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17394 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17395 (vect_grouped_load_supported): Likewise.
17396 (vect_shift_permute_load_chain): Likewise.
17397 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17398 (vect_transform_slp_perm_load): Likewise.
17399 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17400 (vectorizable_bswap): Likewise.
17401 (vect_gen_perm_mask_checked): Likewise.
17402 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17403 implementations of variable permutation vectors into account
17404 when deciding which selector to use.
17405 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17406 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17407 with a false third argument.
17408 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17409 to test whether the constant selector is valid and can_vec_perm_var_p
17410 to test whether a variable selector is valid.
17412 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17414 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17415 * optabs-query.c (can_vec_perm_p): Likewise.
17416 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17417 instead of vec_perm_indices.
17418 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17419 (vect_gen_perm_mask_checked): Likewise,
17420 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17421 (vect_gen_perm_mask_checked): Likewise,
17423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17425 * optabs-query.h (qimode_for_vec_perm): Declare.
17426 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17427 (qimode_for_vec_perm): ...this new function.
17428 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17430 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17432 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17433 does not have a conditional at the top.
17435 2018-01-02 Richard Biener <rguenther@suse.de>
17437 * ipa-inline.c (big_speedup_p): Fix expression.
17439 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17442 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17445 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17449 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17450 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17451 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17452 cond_taken_branch_cost 3->4.
17454 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17456 PR tree-optimization/83581
17457 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17458 TODO_cleanup_cfg if any changes have been made.
17460 PR middle-end/83608
17461 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17462 convert_modes if target mode has the right side, but different mode
17465 PR middle-end/83609
17466 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17467 last argument when extracting from CONCAT. If either from_real or
17468 from_imag is NULL, use expansion through memory. If result is not
17469 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17470 the parts directly to inner mode, if even that fails, use expansion
17473 PR middle-end/83623
17474 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17475 check for bswap in mode rather than HImode and use that in expand_unop
17478 Copyright (C) 2018 Free Software Foundation, Inc.
17480 Copying and distribution of this file, with or without modification,
17481 are permitted in any medium without royalty provided the copyright
17482 notice and this notice are preserved.