1 2018-06-21 Richard Biener <rguenther@suse.de>
3 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
4 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
5 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
6 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8 (vect_record_base_alignments): Do not record base alignment
9 for gathers or scatters.
10 (vect_compute_data_ref_alignment): Drop return value that is always
11 true. Bail out early for gathers or scatters.
12 (vect_enhance_data_refs_alignment): Bail out early for gathers
14 (vect_find_same_alignment_drs): Likewise.
15 (vect_analyze_data_refs_alignment): Remove dead code.
16 (vect_slp_analyze_and_verify_node_alignment): Likewise.
17 (vect_analyze_data_refs): For possible gathers or scatters do
18 not create an alternate DR, just check their possible validity
19 and mark them. Adjust DECL_NONALIASED handling to not rely
21 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
22 update inits of gathers or scatters.
23 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
24 Also copy gather/scatter flag to pattern vinfo.
26 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
28 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
29 behavior of vec_packsu (vector unsigned long long, vector unsigned
30 long long) to match behavior of vec_packs with same signature.
32 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
33 Thomas Schwinge <thomas@codesourcery.com>
34 Cesar Philippidis <cesar@codesourcery.com>
36 * gimplify.c (gimplify_scan_omp_clauses): Add support for
37 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
38 (gimplify_adjust_omp_clauses): Likewise.
39 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
40 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
41 (gimplify_omp_target_update): Update handling of acc update and
43 * omp-low.c (install_var_field): Remove unused parameter
44 base_pointers_restrict.
45 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
46 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
48 (omp_target_base_pointers_restrict_p): Delete.
49 (scan_omp_target): Update call to scan_sharing_clauses.
50 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
52 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
53 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
54 (convert_local_omp_clauses): Likewise.
55 * tree-pretty-print.c (dump_omp_clause): Likewise.
56 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
58 (omp_clause_code_name): Likewise.
60 2018-06-20 Jakub Jelinek <jakub@redhat.com>
63 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
66 PR tree-optimization/86231
67 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
68 anti-range don't overwrite *vr0min before using it to compute *vr0max.
70 2018-06-20 Tom de Vries <tdevries@suse.de>
72 PR tree-optimization/86097
73 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
74 iv type if signedness of iv type is not the same as that of *nit.
76 2018-06-20 Jakub Jelinek <jakub@redhat.com>
78 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
79 EDGE_EH edges, verify they are all EDGE_EH.
81 2018-06-20 Maya Rashish <coypu@sdf.org>
83 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
85 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
87 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
88 * config/aarch64/aarch64.c (xgene1_tunings): Add
89 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
90 (aarch64_mode_valid_for_sched_fusion_p):
92 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
93 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
95 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
97 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
98 * config/aarch64/iterators.md (VQ2): New mode iterator.
100 2018-06-20 Martin Liska <mliska@suse.cz>
102 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
103 Change default ratio from 10 to 8.
105 2018-06-20 Martin Liska <mliska@suse.cz>
107 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
109 (bit_test_cluster::find_bit_tests): Likewise.
110 (switch_decision_tree::analyze_switch_statement): Find clusters.
111 * tree-switch-conversion.h (struct jump_table_cluster): Document
114 2018-06-20 Martin Liska <mliska@suse.cz>
116 * tree-switch-conversion.c (switch_conversion::collect):
117 Record m_uniq property.
118 (switch_conversion::expand): Bail out for special conditions.
119 (group_cluster::~group_cluster): New.
120 (group_cluster::group_cluster): Likewise.
121 (group_cluster::dump): Likewise.
122 (jump_table_cluster::emit): New.
123 (switch_decision_tree::fix_phi_operands_for_edges): New.
124 (struct case_node): Remove struct.
125 (jump_table_cluster::can_be_handled): New.
126 (case_values_threshold): Moved to header.
127 (reset_out_edges_aux): Likewise.
128 (jump_table_cluster::is_beneficial): New.
129 (bit_test_cluster::can_be_handled): Likewise.
130 (add_case_node): Remove.
131 (bit_test_cluster::is_beneficial): New.
132 (case_bit_test::cmp): New.
133 (bit_test_cluster::emit): New.
134 (expand_switch_as_decision_tree_p): Remove.
135 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
136 (fix_phi_operands_for_edge): Likewise.
137 (switch_decision_tree::analyze_switch_statement): New.
138 (compute_cases_per_edge): Move ...
139 (switch_decision_tree::compute_cases_per_edge): ... here.
140 (try_switch_expansion): Likewise.
141 (switch_decision_tree::try_switch_expansion): Likewise.
142 (record_phi_operand_mapping): Likewise.
143 (switch_decision_tree::record_phi_operand_mapping): Likewise.
144 (emit_case_decision_tree): Likewise.
145 (switch_decision_tree::emit): Likewise.
146 (balance_case_nodes): Likewise.
147 (switch_decision_tree::balance_case_nodes): Likewise.
148 (dump_case_nodes): Likewise.
149 (switch_decision_tree::dump_case_nodes): Likewise.
150 (emit_jump): Likewise.
151 (switch_decision_tree::emit_jump): Likewise.
152 (emit_cmp_and_jump_insns): Likewise.
153 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
154 (emit_case_nodes): Likewise.
155 (switch_decision_tree::emit_case_nodes): Likewise.
156 (conditional_probability): Remove.
157 * tree-switch-conversion.h (enum cluster_type): New.
159 (struct cluster): Likewise.
160 (cluster::cluster): Likewise.
161 (struct simple_cluster): Likewise.
162 (simple_cluster::simple_cluster): Likewise.
163 (struct group_cluster): Likewise.
164 (struct jump_table_cluster): Likewise.
165 (struct bit_test_cluster): Likewise.
166 (struct min_cluster_item): Likewise.
167 (struct case_tree_node): Likewise.
168 (case_tree_node::case_tree_node): Likewise.
169 (jump_table_cluster::case_values_threshold): Likewise.
170 (struct case_bit_test): Likewise.
171 (struct switch_decision_tree): Likewise.
172 (struct switch_conversion): Likewise.
173 (switch_decision_tree::reset_out_edges_aux): Likewise.
175 2018-06-20 Martin Liska <mliska@suse.cz>
177 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
178 (hoist_edge_and_branch_if_true): Likewise.
179 (expand_switch_using_bit_tests_p): Likewise.
180 (struct case_bit_test): Likewise.
181 (case_bit_test_cmp): Likewise.
182 (emit_case_bit_tests): Likewise.
183 (switch_conversion::switch_conversion): New class.
184 (struct switch_conv_info): Remove old struct.
185 (collect_switch_conv_info): More to ...
186 (switch_conversion::collect): ... this.
187 (check_range): Likewise.
188 (switch_conversion::check_range): Likewise.
189 (check_all_empty_except_final): Likewise.
190 (switch_conversion::check_all_empty_except_final): Likewise.
191 (check_final_bb): Likewise.
192 (switch_conversion::check_final_bb): Likewise.
193 (create_temp_arrays): Likewise.
194 (switch_conversion::create_temp_arrays): Likewise.
195 (free_temp_arrays): Likewise.
196 (gather_default_values): Likewise.
197 (switch_conversion::gather_default_values): Likewise.
198 (build_constructors): Likewise.
199 (switch_conversion::build_constructors): Likewise.
200 (constructor_contains_same_values_p): Likewise.
201 (switch_conversion::contains_same_values_p): Likewise.
202 (array_value_type): Likewise.
203 (switch_conversion::array_value_type): Likewise.
204 (build_one_array): Likewise.
205 (switch_conversion::build_one_array): Likewise.
206 (build_arrays): Likewise.
207 (switch_conversion::build_arrays): Likewise.
208 (gen_def_assigns): Likewise.
209 (switch_conversion::gen_def_assigns): Likewise.
210 (prune_bbs): Likewise.
211 (switch_conversion::prune_bbs): Likewise.
212 (fix_phi_nodes): Likewise.
213 (switch_conversion::fix_phi_nodes): Likewise.
214 (gen_inbound_check): Likewise.
215 (switch_conversion::gen_inbound_check): Likewise.
216 (process_switch): Use the newly created class.
217 (switch_conversion::expand): New.
218 (switch_conversion::~switch_conversion): New.
219 * tree-switch-conversion.h: New file.
221 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
223 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
224 tree-vect-patterns.c.
225 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
226 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
227 (vect_recog_sad_pattern): Likewise.
228 (vect_recog_widen_sum_pattern): Likewise.
229 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
230 (vect_recog_widen_shift_pattern): Remove the type_in argument.
231 (vect_recog_rotate_pattern): Likewise.
232 (vect_recog_mult_pattern): Likewise.
233 (vect_recog_vector_vector_shift_pattern): Likewise.
234 (vect_recog_divmod_pattern): Likewise.
235 (vect_recog_mixed_size_cond_pattern): Likewise.
236 (vect_recog_bool_pattern): Likewise.
237 (vect_recog_mask_conversion_pattern): Likewise.
238 (vect_try_gather_scatter_pattern): Likewise.
239 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
240 (vect_recog_over_widening_pattern): Likewise.
241 (vect_recog_gather_scatter_pattern): Likewise.
242 (vect_recog_func_ptr): Move from tree-vectorizer.h
243 (vect_vect_recog_func_ptrs): Move further down the file.
244 (vect_recog_func): Likewise. Remove the third argument.
245 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
246 (vect_pattern_recog_1): Expect the pattern function to do any
247 necessary target tests. Also expect it to provide a vector type.
248 Remove the type_in handling.
250 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
252 * tree-vect-patterns.c (vect_pattern_detected): New function.
253 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
254 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
255 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
256 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
257 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
258 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
259 (vect_recog_mask_conversion_pattern)
260 (vect_try_gather_scatter_pattern): Likewise.
262 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
264 * tree-vect-patterns.c (vect_get_internal_def): New function.
265 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
266 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
267 (search_type_for_mask_1): Use it.
269 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
271 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
272 redundant WIDEN_SUM_EXPR handling.
273 (vect_recog_sad_pattern): Likewise.
275 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
277 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
278 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
279 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
280 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
281 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
283 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
285 * tree-vect-stmts.c (vectorizable_call): Make sure that we
286 use the stmt_vec_info of the original bb statement for the
287 new zero assignment, even if the call is part of a pattern.
289 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
291 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
292 that the sequence is attached to the original statement rather
293 than the pattern statement.
294 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
295 PATTERN_DEF_SEQ from the original statement rather than
296 the main pattern statement.
297 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
298 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
299 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
301 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
303 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
304 definition statements before the early exit for statements that aren't
306 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
308 (vect_transform_loop): ...here. Process pattern definition
309 statements without first checking whether the main pattern
310 statement is live or relevant.
312 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
314 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
315 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
317 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
319 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
320 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
321 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
322 (expand_block_compare): Change select_block_compare_mode call.
323 (expand_strncmp_align_check): Use new functions, fix comment.
324 (emit_final_str_compare_gpr): New function.
325 (expand_strn_compare): Refactor and clean up code.
326 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
328 2018-06-19 Tony Reix <tony.reix@atos.com>
329 Damien Bergamini <damien.bergamini@atos.com>
330 David Edelsohn <dje.gcc@gmail.com>
332 * collect2.c (static_obj): New variable.
333 (static_libs): New variable.
334 (is_in_list): Uncomment declaration.
335 (main): Track AIX libraries linked statically.
336 (is_in_list): Uncomment definition.
337 (scan_prog_file): Don't add AIX shared libraries initializer
338 to constructor list if linking statically.
340 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
342 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
344 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
346 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
348 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
351 2018-06-19 Martin Liska <mliska@suse.cz>
353 * config/i386/i386.c (ix86_can_inline_p): Do not use
354 ipa_fn_summaries::get_create.
355 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
357 (devirtualization_time_bonus): Likewise.
358 (ipcp_propagate_stage): Likewise.
359 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
360 (edge_set_predicate): Likewise.
361 (evaluate_conditions_for_known_args): Likewise.
362 (evaluate_properties_for_edge): Likewise.
363 (ipa_call_summary::reset): Tranform to ...
364 (ipa_call_summary::~ipa_call_summary): ... this.
365 (ipa_fn_summary::reset): Transform to ...
366 (ipa_fn_summary::~ipa_fn_summary): ... this.
367 (ipa_fn_summary_t::remove): Rename to ...
368 (ipa_fn_summary_t::remove_callees): ... this.
369 (ipa_fn_summary_t::duplicate): Use placement new
370 instead of memory copy.
371 (ipa_call_summary_t::duplicate): Likewise.
372 (ipa_call_summary_t::remove): Remove.
373 (dump_ipa_call_summary): Change get_create to get.
374 (ipa_dump_fn_summary): Dump only when summary exists.
375 (analyze_function_body): Use symbol_summary::get instead
377 (compute_fn_summary): Likewise.
378 (estimate_edge_devirt_benefit): Likewise.
379 (estimate_edge_size_and_time): Likewise.
380 (inline_update_callee_summaries): Likewise.
381 (remap_edge_change_prob): Likewise.
382 (remap_edge_summaries): Likewise.
383 (ipa_merge_fn_summary_after_inlining): Likewise.
384 (write_ipa_call_summary): Likewise.
385 (ipa_fn_summary_write): Likewise.
386 (ipa_free_fn_summary): Likewise.
387 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
388 (struct ipa_call_summary): Likewise.
389 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
391 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
392 (estimate_size_after_inlining): Likewise.
393 (estimate_growth): Likewise.
394 (growth_likely_positive): Likewise.
395 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
396 (inline_call): Likewise.
397 * ipa-inline.c (caller_growth_limits): Likewise.
398 (can_inline_edge_p): Likewise.
399 (can_inline_edge_by_limits_p): Likewise.
400 (compute_uninlined_call_time): Likewise.
401 (compute_inlined_call_time): Likewise.
402 (want_inline_small_function_p): Likewise.
403 (edge_badness): Likewise.
404 (update_caller_keys): Likewise.
405 (update_callee_keys): Likewise.
406 (inline_small_functions): Likewise.
407 (inline_to_all_callers_1): Likewise.
408 (dump_overall_stats): Likewise.
409 (early_inline_small_functions): Likewise.
410 (early_inliner): Likewise.
411 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
412 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
413 * ipa-pure-const.c (malloc_candidate_p): Likewise.
414 * ipa-split.c (execute_split_functions): Likewise.
415 * symbol-summary.h: Likewise.
416 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
418 2018-06-19 Richard Biener <rguenther@suse.de>
420 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
421 (vectorize_loops): ... here. Fix dbgcnt handling.
422 (try_vectorize_loop): Wrap try_vectorize_loop_1.
424 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
427 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
428 ieee128 argument takes up only one (vector) register, not two (floating
431 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
433 * gimplify.c (gimplify_init_constructor): Really never clear for an
434 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
436 2018-06-19 Richard Biener <rguenther@suse.de>
438 PR tree-optimization/86179
439 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
440 after failed recognition.
442 2018-06-18 Martin Sebor <msebor@redhat.com>
445 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
446 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
447 Handle integer subtraction.
448 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
449 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
451 2018-06-18 David Malcolm <dmalcolm@redhat.com>
453 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
454 param from rtx to rtx_insn *.
455 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
457 (frv_ifcvt_modify_insn): Likwise.
458 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
459 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
460 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
461 as_a <rtx_insn *> cast to local "unprotected_region" once
462 it's been established that it's not NULL or pc_rtx.
463 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
464 param "sethi" from rtx to rtx_insn *.
465 (nds32_group_float_insns): Likewise for param "insn".
466 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
468 (vax_output_int_subtract): Likewise.
469 * config/vax/vax.c (vax_output_int_add): Likewise for param
471 (vax_output_int_subtract): Likewise.
472 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
473 (emit_pattern_after): Likewise for param "after".
474 (emit_insn_after): Likewise.
475 (emit_jump_insn_after): Likewise.
476 (emit_call_insn_after): Likewise.
477 (emit_debug_insn_after): Likewise.
478 (emit_pattern_before): Likewise for param "before".
479 (emit_insn_before): Likewise.
480 (emit_jump_insn_before): Likewise.
481 * final.c (get_insn_template): Likewise for param "insn", removing
483 * output.h (get_insn_template): Likewise for 2nd param.
484 * rtl.h (emit_insn_before): Likewise.
485 (emit_jump_insn_before): Likewise.
486 (emit_debug_insn_before_noloc): Likewise.
487 (emit_insn_after): Likewise.
488 (emit_jump_insn_after): Likewise.
489 (emit_call_insn_after): Likewise.
490 (emit_debug_insn_after): Likewise.
491 (set_insn_deleted): Likewise for param.
493 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
496 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
497 floating point modes, so that IFmode is numerically greater than
498 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
499 to declare the ordering. This prevents IFmode from being
500 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
501 machine. Include rs6000-modes.h to share the fractional values
502 between genmodes* and the rest of the compiler.
506 * config/rs6000/rs6000-modes.h: New file.
507 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
508 meaning of rs6000_long_double_size so that 126..128 selects an
509 appropriate 128-bit floating point type.
510 (rs6000_option_override_internal): Likewise.
511 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
512 (TARGET_LONG_DOUBLE_128): Change the meaning of
513 rs6000_long_double_size so that 126..128 selects an appropriate
514 128-bit floating point type.
515 (LONG_DOUBLE_TYPE_SIZE): Update comment.
516 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
517 source and destination to match the standard usage.
518 (truncifkf2): Likewise.
519 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
520 ISA 2.07 to use an explicit clobber, instead of passing in a
522 (copysign<mode>3_soft): Likewise.
524 2018-06-18 David Malcolm <dmalcolm@redhat.com>
526 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
527 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
528 (vect_slp_analyze_instance_dependence): Likewise.
529 (vect_enhance_data_refs_alignment): Likewise.
530 (vect_analyze_data_refs_alignment): Likewise.
531 (vect_slp_analyze_and_verify_instance_alignment
532 (vect_analyze_data_ref_accesses): Likewise.
533 (vect_prune_runtime_alias_test_list): Likewise.
534 (vect_analyze_data_refs): Likewise.
535 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
536 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
537 (vect_analyze_scalar_cycles_1): Likewise.
538 (vect_get_loop_niters): Likewise.
539 (vect_analyze_loop_form_1): Likewise.
540 (vect_update_vf_for_slp): Likewise.
541 (vect_analyze_loop_operations): Likewise.
542 (vect_analyze_loop): Likewise.
543 (vectorizable_induction): Likewise.
544 (vect_transform_loop): Likewise.
545 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
546 * tree-vect-slp.c (vect_analyze_slp): Likewise.
547 (vect_make_slp_decision): Likewise.
548 (vect_detect_hybrid_slp): Likewise.
549 (vect_slp_analyze_operations): Likewise.
550 (vect_slp_bb): Likewise.
551 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
552 (vectorizable_bswap): Likewise.
553 (vectorizable_call): Likewise.
554 (vectorizable_simd_clone_call): Likewise.
555 (vectorizable_conversion): Likewise.
556 (vectorizable_assignment): Likewise.
557 (vectorizable_shift): Likewise.
558 (vectorizable_operation): Likewise.
559 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
561 2018-06-18 Martin Sebor <msebor@redhat.com>
563 PR tree-optimization/81384
564 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
565 * builtins.c (expand_builtin_strnlen): New function.
566 (expand_builtin): Call it.
567 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
568 * builtins.def (BUILT_IN_STRNLEN): New.
569 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
570 Warn for bounds in excess of maximum object size.
571 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
572 single-value ranges. Handle strnlen.
573 (handle_builtin_strlen): Handle strnlen.
574 (strlen_check_and_optimize_stmt): Same.
575 * doc/extend.texi (Other Builtins): Document strnlen.
577 2018-06-18 Maya Rashish <coypu@sdf.org>
579 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
580 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
581 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
583 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
585 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
587 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
589 * tree.c (escaped_string::escape): Replace cast to char * by
590 const_cast<char *> (unescaped).
592 2018-06-18 Nick Clifton <nickc@redhat.com>
595 * tree.c (escaped_string): New class. Converts an unescaped
596 string into its escaped equivalent.
597 (warn_deprecated_use): Use the new class to convert the
598 deprecation message, if present.
599 (test_escaped_strings): New self test.
600 (test_c_tests): Add test_escaped_strings.
601 * doc/extend.texi (deprecated): Add a note that the
602 deprecation message is affected by the -fmessage-length
603 option, and that control characters will be escaped.
604 (#pragma GCC error): Document this pragma.
605 (#pragma GCC warning): Likewise.
606 * doc/invoke.texi (-fmessage-length): Document this option's
607 effect on the #warning and #error preprocessor directives and
608 the deprecated attribute.
610 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
612 * tree.c (decl_value_expr_lookup): Revert latest change.
613 (decl_value_expr_insert): Likewise.
615 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
617 * gimplify.c (nonlocal_vlas): Delete.
618 (nonlocal_vla_vars): Likewise.
619 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
621 (gimplify_body): Do not create and destroy nonlocal_vlas.
622 * tree-nested.c: Include diagnostic.h.
623 (use_pointer_in_frame): Tweak.
624 (lookup_field_for_decl): Add assertion and declare the transformation.
625 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
626 internal error when the reference is in a wrong context. Do not
627 create a debug decl by default.
628 (note_nonlocal_block_vlas): Delete.
629 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
630 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
631 create a debug decl by default.
632 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
633 call is in a wrong context.
634 (fixup_vla_decls): New function.
635 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
636 debug variables were created.
637 * tree.c (decl_value_expr_lookup): Add checking assertion.
638 (decl_value_expr_insert): Likewise.
640 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
643 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
644 * tree-scalar-evolution.c (interpret_expr): Likewise.
645 (expression_expensive_p): Likewise.
646 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
647 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
648 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
649 (ssa_defined_by_minus_one_stmt_p): New.
651 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
654 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
655 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
656 * dojump.c (do_jump): Likewise.
657 * expr.c (expand_expr_real_2): Check operand type's sign.
658 * fold-const.c (const_unop): Handle ABSU_EXPR.
659 (fold_abs_const): Likewise.
660 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
661 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
662 (strip_sign_op_1): Likesise.
663 * match.pd: Add new pattern to generate ABSU_EXPR.
664 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
665 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
666 * tree-eh.c (operation_could_trap_helper_p): Likewise.
667 * tree-inline.c (estimate_operator_cost): Likewise.
668 * tree-pretty-print.c (dump_generic_node): Likewise.
669 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
670 * tree.def (ABSU_EXPR): New.
672 2018-06-16 Jakub Jelinek <jakub@redhat.com>
675 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
676 documented as preserved for backward compatibility only.
677 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
679 PR rtl-optimization/86108
680 * bb-reorder.c (create_forwarder_block): Renamed to ...
681 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
682 jump from new landing pad to the second part.
683 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
686 2018-06-15 Jakub Jelinek <jakub@redhat.com>
689 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
690 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
691 Only call store_expr for halves if the mode is the same.
694 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
695 Fix up comment formatting.
697 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
699 * typed-splay-tree.h (typed_splay_tree::remove): New function.
700 (typed_splay_tree::closure,
701 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
702 (typed_splay_tree::typed_splay_tree,
703 typed_splay_tree::operator =): Declared private.
704 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
705 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
706 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
707 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
708 typed_splay_tree::splay_tree_splay,
709 typed_splay_tree::splay_tree_foreach_helper,
710 typed_splay_tree::splay_tree_insert,
711 typed_splay_tree::splay_tree_remove,
712 typed_splay_tree::splay_tree_lookup,
713 typed_splay_tree::splay_tree_predecessor,
714 typed_splay_tree::splay_tree_successor,
715 typed_splay_tree::splay_tree_min,
716 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
717 (typed_splay_tree::root, typed_splay_tree::comp,
718 typed_splay_tree::delete_key,
719 typed_splay_tree::delete_value): New data members.
720 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
721 typed_splay_tree::remove.
723 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
725 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
726 -mginv and -mno-ginv to the assembler.
727 * config/mips/mips.opt (-mcrc): New option.
729 * doc/invoke.text (-mcrc): Document.
732 2018-06-15 Nick Clifton <nickc@redhat.com>
735 * tree.c (escaped_string): New class. Converts an unescaped
736 string into its escaped equivalent.
737 (warn_deprecated_use): Use the new class to convert the
738 deprecation message, if present.
739 (test_escaped_strings): New self test.
740 (test_c_tests): Add test_escaped_strings.
741 * doc/extend.texi (deprecated): Add a note that the
742 deprecation message is affected by the -fmessage-length
743 option, and that control characters will be escaped.
744 (#pragma GCC error): Document this pragma.
745 (#pragma GCC warning): Likewise.
746 * doc/invoke.texi (-fmessage-length): Document this option's
747 effect on the #warning and #error preprocessor directives and
748 the deprecated attribute.
750 2018-06-15 Richard Biener <rguenther@suse.de>
752 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
753 here, also noting vector size used.
754 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
755 size used in MSG_OPTIMIZED_LOCATIONS dump.
756 (pass_slp_vectorize::execute): Adjust.
758 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
761 * config/arc/arc.c (arc_return_address_register): Fix
764 2018-06-15 Richard Biener <rguenther@suse.de>
767 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
768 leave useless conversion stripping to force_gimple_operand_gsi.
769 (gimplify_build2): Likewise.
770 (gimplify_build1): Likewise.
772 2018-06-15 Richard Biener <rguenther@suse.de>
775 * tree-cfg.c (move_stmt_op): unshare invariant addresses
776 before adjusting their block.
778 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
780 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
781 multilibs for *-*-rtems*.
782 * config/riscv/t-rtems: New file.
784 2018-06-14 Jakub Jelinek <jakub@redhat.com>
787 * match.pd ((A +- CST1) +- CST2): Punt if last resort
788 unsigned_type_for returns NULL.
791 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
792 subregs of multi-word pseudos unless the float mode has word size.
794 2018-06-14 Richard Biener <rguenther@suse.de>
797 * tree-vect-generic.c (build_word_mode_vector_type): Remove
798 duplicate and harmful type_hash_canon.
799 * tree.c (type_hash_canon): Assert we didn't find ourselves.
801 2018-06-14 Richard Biener <rguenther@suse.de>
804 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
807 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
809 * config/rtems.h (STDINT_LONG32): Define.
811 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
812 Prachi Godbole <prachi.godbole@imgtec.com>
814 * config/mips/mips-cpus.def: Define P6600.
815 * config/mips/mips-tables.opt: Regenerate.
816 * config/mips/mips.c (mips_ucbranch_type): New enum.
817 (mips_rtx_cost_data): Add support for P6600.
818 (mips_issue_rate): Likewise.
819 (mips_multipass_dfa_lookahead): Likewise.
820 (mips_avoid_hazard): Likewise.
821 (mips_reorg_process_insns): Likewise.
822 (mips_classify_branch_p6600): New function.
823 * config/mips/mips.h (TUNE_P6600): New define.
824 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
825 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
826 * config/mips/mips.md: Include p6600.md.
827 (processor): Add p6600.
828 * config/mips/p6600.md: New file.
829 * doc/invoke.texi: Add p6600 to supported architectures.
831 2018-06-13 Martin Sebor <msebor@redhat.com>
833 PR tree-optimization/86114
834 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
836 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
838 2018-06-13 Richard Biener <rguenther@suse.de>
840 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
841 Properly set vector type of the intermediate stmt.
842 * tree-vect-stmts.c (vectorizable_operation): The destination
843 var always has vectype_out type.
845 2018-06-13 Jeff Law <law@redhat.com>
847 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
848 integer 0 for argument to print_rtl_with_bb.
849 (rl78_reorg): Likewise.
851 2018-06-13 David Malcolm <dmalcolm@redhat.com>
853 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
854 from rtx to rtx_insn *.
855 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
857 (add_sched_insns_for_speculation): Likewise for local "target",
858 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
859 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
860 from rtx_insn ** to rtx_code_label **.
861 (reorg_emit_nops): Likewise.
862 (c6x_reorg): Likewise for local "call_labels".
863 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
865 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
866 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
867 the loops over LABEL_REFS.
868 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
870 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
871 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
872 (split_branches): Strengthen local "olabel" from rtx to
873 rtx_insn *, adding a safe_as_a cast.
874 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
876 (add_insn_after): Likewise for first two params.
877 (add_insn_before): Likewise.
878 (remove_insn): Likewise for param.
879 (emit_pattern_before_noloc): Likewise for second and third params.
880 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
881 (emit_call_insn_before_noloc): Likewise.
882 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
884 (emit_barrier_before): Likewise.
885 (emit_label_before): Strengthen "label" param from "rtx" to
886 "rtx_code_label *". Strengthen "before" param from "rtx" to
888 (emit_insn_after_1): Strengthen "after" param from "rtx" to
890 (emit_pattern_after_noloc): Likewise.
891 (emit_insn_after_noloc): Likewise.
892 (emit_jump_insn_after_noloc): Likewise.
893 (emit_call_insn_after_noloc): Likewise.
894 (emit_debug_insn_after_noloc): Likewise.
895 (emit_barrier_after): Likewise.
896 (emit_label_after): Likewise for both params.
897 (emit_pattern_after_setloc): Likewise for "after" param. Convert
898 "loc" param from "int" to "location_t".
899 (emit_insn_after_setloc): Likewise.
900 (emit_jump_insn_after_setloc): Likewise.
901 (emit_call_insn_after_setloc): Likewise.
902 (emit_debug_insn_after_setloc): Likewise.
903 (emit_pattern_before_setloc): Likewise for "before" param. Convert
904 "loc" param from "int" to "location_t".
905 (emit_pattern_before): Convert NULL_RTX to NULL.
906 (emit_insn_before_setloc): Convert "loc" param from "int" to
908 (emit_jump_insn_before_setloc): Likewise.
909 (emit_call_insn_before_setloc): Likewise.
910 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
911 rtx_insn *. Convert "loc" param from "int" to "location_t".
912 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
913 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
914 Convert 3rd param from "int" to "location_t".
915 (emit_barrier_before, emit_barrier_after, next_real_insn):
916 Strengthen param from rtx to rtx_insn *.
917 (emit_label_before): Strengthen 1st param from "rtx" to
918 "rtx_code_label *". Strengthen 2nd param from "rtx" to
920 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
921 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
922 Strengthen 2nd param from "rtx" to "rtx_insn *".
923 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
924 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
925 Likewise. Convert 3rd param from "int" to "location_t".
926 (emit_label_after): Strengthen 1st param from "rtx" to
928 (next_real_insn, remove_insn): Strengthen param from "rtx" to
930 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
931 from "rtx" to "rtx_insn *".
933 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
935 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
936 bodies streamed in with -Q.
937 * dumpfile.c (dump_files): Add lto-stream-out dump file.
938 * dumpfile.h (tree_dump_index): Add lto_stream_out.
939 * gimple-streamer-out.c: Include gimple-pretty-print.h
940 (output_bb): Dump stmts streamed.
941 * lto-section-out.c: Include print-tree.h
942 (lto_begin_section): Dump sections created.
943 (lto_output_decl_index): Dump decl encoded.
944 * lto-streamer-out.c: Include print-tree.h
945 (create_output_block): Dump output block created.
946 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
947 (output_function): Dump function output.
948 (output_constructor): Dump constructor streamed.
949 (write_global_stream): Output indexes encoded.
950 (produce_asm_for_decls): Dump streams encoded.
951 * lto-streamer.c (streamer_dump_file): New global var.
952 * lto-streamer.h (streamer_dump_file): Declare.
953 * passes.c (ipa_write_summaries): Initialize streamer dump.
954 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
957 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
960 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
961 offsets for register save directives. Emit a second batch of save
962 directives, if need be, when the function accesses prior frames.
964 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
966 * config/arc/fpu.md (fmasf4): Force operand to register.
969 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
971 * config/arc/arc-protos.h (arc_pad_return): Remove.
972 * config/arc/arc.c (machine_function): Remove force_short_suffix
974 (arc_print_operand): Adjust printing of '&'.
975 (arc_verify_short): Remove conditional printing of short suffix.
976 (arc_final_prescan_insn): Remove reference to size_reason.
977 (pad_return): New function.
978 (arc_reorg): Call pad_return.
979 (arc_pad_return): Remove.
980 (arc_init_machine_status): Remove reference to force_short_suffix.
981 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
982 (attr length): When attribute iscompact is true force to 2
983 regardless; in the case of maybe check if we want to force the
984 instruction to have 4 bytes length.
985 (nopv): Change it to generate 4 byte long nop as well.
986 (blockage): New pattern.
987 (simple_return): Remove call to arc_pad_return.
988 (p_return_i): Likewise.
990 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
992 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
994 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
996 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
999 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1001 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
1002 for ARC700 and ARCv2.
1004 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
1007 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
1008 operands[2] instead of operands[1].
1011 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1013 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
1014 case, check whether the outer register overlaps an unallocatable
1015 register, not just whether it fits the required class.
1017 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
1019 * poly-int.h (can_div_trunc_p): Add new overload in which all values
1021 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
1022 (memrefs_conflict_p): Likewise.
1023 (init_alias_analysis): Likewise.
1024 * cfgexpand.c (expand_debug_expr): Likewise.
1025 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
1026 * cse.c (fold_rtx): Likewise.
1027 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
1028 * expr.c (emit_block_move_hints): Likewise.
1029 (clear_storage_hints, push_block, emit_push_insn): Likewise.
1030 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
1031 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
1032 (emit_group_store): Likewise.
1033 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
1034 to read the PRE/POST_MODIFY increment.
1035 * calls.c (store_one_arg): Use strip_offset.
1036 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
1038 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
1040 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
1041 (simplify_binary_operation_1): Extend CONST_INT handling to
1043 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
1044 than a HOST_WIDE_INT.
1045 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
1047 (adjust_mems, add_stores): Update accodingly.
1048 (vt_canonicalize_addr): Track polynomial offsets.
1049 (emit_note_insn_var_location): Likewise.
1050 (vt_add_function_parameter): Likewise.
1051 (vt_initialize): Likewise.
1053 2018-06-12 Jeff Law <law@redhat.com>
1055 * config.gcc (alpha*-*-freebsd*): Remove.
1056 * config/alpha/freebsd.h: Remove.
1058 2018-06-12 David Malcolm <dmalcolm@redhat.com>
1061 * spellcheck-tree.c (levenshtein_distance): Rename to...
1062 (get_edit_distance): ...this, and update for underlying renaming.
1063 * spellcheck-tree.h (levenshtein_distance): Rename to...
1064 (get_edit_distance): ...this.
1065 * spellcheck.c (levenshtein_distance): Rename to...
1066 (get_edit_distance): ...this. Convert from Levenshtein distance
1067 to Damerau-Levenshtein distance by supporting transpositions of
1068 adjacent characters. Rename "v1" to "v_next" and "v0" to
1070 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
1071 (selftest::test_edit_distance_unit_test_oneway): ...this, and
1072 update for underlying renaming.
1073 (selftest::levenshtein_distance_unit_test): Rename to...
1074 (selftest::test_get_edit_distance_unit): ...this, and update for
1075 underlying renaming.
1076 (selftest::test_find_closest_string): Add example from PR 69968
1077 where transposition helps
1078 (selftest::test_metric_conditions): Update for renaming.
1079 (selftest::test_metric_conditions): Likewise.
1080 (selftest::spellcheck_c_tests): Likewise.
1081 * spellcheck.h (levenshtein_distance): Rename both overloads to...
1082 (get_edit_distance): ...this.
1083 (best_match::consider): Update for renaming.
1085 2018-06-12 Martin Sebor <msebor@redhat.com>
1087 PR tree-optimization/85259
1088 * builtins.c (compute_objsize): Handle constant offsets.
1089 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
1090 true iff a warning has been issued.
1091 * gimple.h (gimple_nonartificial_location): New function.
1092 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
1093 gimple_nonartificial_location and handle -Wno-system-headers.
1094 (handle_builtin_stxncpy): Same.
1096 2018-06-12 Martin Sebor <msebor@redhat.com>
1099 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
1101 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
1103 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
1104 BUILTIN_VEC_XST entries for pointer to double and long long.
1106 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
1109 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1111 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
1114 2018-06-12 Martin Liska <mliska@suse.cz>
1116 * doc/options.texi: Document IntegerRange.
1118 2018-06-12 Martin Liska <mliska@suse.cz>
1120 * config/i386/i386.opt: Make MPX-related options as Deprecated.
1121 * opt-functions.awk: Handle Deprecated flag.
1122 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
1124 (read_cmdline_option): Report warning for a deprecated option.
1125 * opts.h (struct cl_option): Add new field cl_deprecated.
1126 (CL_ERR_DEPRECATED): New.
1128 2018-06-12 Martin Liska <mliska@suse.cz>
1130 * doc/options.texi: Document Deprecated option flag.
1132 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1134 * config/arc/arc-arch.h (arc_extras): New enum.
1135 (arc_cpu_t):Add field extra.
1136 (arc_cpu_types): Consider the extras.
1137 * config/arc/arc-cpus.def: Add extras info.
1138 * config/arc/arc-opts.h (processor_type): Consider extra field.
1139 * config/arc/arc.c (arc_override_options): Handle extra field.
1141 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
1143 * config/arc/arc-arch.h: Update ARC_OPTX macro.
1144 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
1146 * config/arc/arc.c (arc_init): Update pic warning.
1147 (irq_range): Update irq range parsing warnings.
1148 (arc_override_options): Update various warning messages.
1149 (arc_handle_aux_attribute): Likewise.
1151 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
1153 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
1155 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1157 * doc/sourcebuild.texi: Document usage of line number 0 in verify
1158 compiler messages directives.
1160 2018-06-12 Matthew Fortune <mfortune@gmail.com>
1162 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
1163 * config/mips/mips-tables.opt: Regenerate.
1164 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
1166 * doc/invoke.texi: Document -march=i6500.
1168 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
1170 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
1171 (i6400_gpmul): Add cpu_unit.
1172 (i6400_gpdiv): Likewise.
1173 (i6400_msa_add_d): Update reservations.
1174 (i6400_msa_int_add) Likewise.
1175 (i6400_msa_short_logic3) Likewise.
1176 (i6400_msa_short_logic2) Likewise.
1177 (i6400_msa_short_logic) Likewise.
1178 (i6400_msa_move) Likewise.
1179 (i6400_msa_cmp) Likewise.
1180 (i6400_msa_short_float2) Likewise.
1181 (i6400_msa_div_d) Likewise.
1182 (i6400_msa_long_logic1) Likewise.
1183 (i6400_msa_long_logic2) Likewise.
1184 (i6400_msa_mult) Likewise.
1185 (i6400_msa_long_float2) Likewise.
1186 (i6400_msa_long_float4) Likewise.
1187 (i6400_msa_long_float5) Likewise.
1188 (i6400_msa_long_float8) Likewise.
1189 (i6400_fpu_fadd): Include frint type.
1190 (i6400_fpu_store): New define_insn_reservation.
1191 (i6400_fpu_load): Likewise.
1192 (i6400_fpu_move): Likewise.
1193 (i6400_fpu_fcmp): Likewise.
1194 (i6400_fpu_fmadd): Likewise.
1195 (i6400_int_mult): Include imul3nc type and update reservation.
1196 (i6400_int_div): Include idiv3 type and update reservation.
1197 (i6400_int_load): Update to check type not move_type.
1198 (i6400_int_store): Likewise.
1199 (i6400_int_prefetch): Set zero latency.
1201 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
1203 * gcc.c: Document new %@{...} sequence.
1204 (LINK_COMMAND_SPEC): Use it for the -L switches.
1205 (cpp_unique_options): Use it for the -I switches.
1206 (at_file_argbuf): New global variable.
1207 (in_at_file): Likewise.
1208 (alloc_args): Create at_file_argbuf.
1209 (clear_args): Truncate at_file_argbuf.
1210 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
1211 (open_at_file): New function.
1212 (close_at_file): Likewise.
1213 (create_at_file): Delete.
1214 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
1217 (validate_switches_from_spec): Deal with %@{...} sequence.
1218 (validate_switches): Likewise.
1219 (driver::finalize): Call clear_args.
1221 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
1223 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
1225 2018-06-11 Martin Sebor <msebor@redhat.com>
1227 * doc/invoke.texi (-Wall): List -Wc++17-compat.
1228 (Wno-class-memaccess): Add @opindex.
1229 (Wno-templates, Wno-multiple-inheritance): Same.
1230 (Wno-virtual-inheritance, Wno-namespaces): Same.
1231 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
1232 (Wno-format-overflow, Wno-format-truncation): Same.
1233 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
1234 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
1235 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
1236 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
1237 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
1238 (Wno-misspelled-isr): Same.
1240 2018-06-11 Martin Sebor <msebor@redhat.com>
1242 * PR tree-optimization/86083
1243 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
1245 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
1247 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
1249 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
1252 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
1253 on the correct operand.
1254 (*movdi_internal64): Ditto.
1256 2018-06-11 Martin Liska <mliska@suse.cz>
1258 PR tree-optimization/86089
1259 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
1261 2018-06-11 Julia Koval <julia.koval@intel.com>
1263 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
1264 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
1265 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
1267 2018-06-11 Olivier Hainque <hainque@adacore.com>
1269 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
1270 for Ada with strict dwarf2.
1272 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
1275 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
1278 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
1280 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
1282 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
1284 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
1287 2018-06-08 Martin Liska <mliska@suse.cz>
1289 * tree-cfg.h (debug_function): Fix argument type to match
1292 2018-06-08 Martin Liska <mliska@suse.cz>
1294 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
1295 Remove usage of MPX-related (and removed) fields.
1296 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
1298 2018-06-08 David Malcolm <dmalcolm@redhat.com>
1300 * cfg.c (debug): Use TDF_NONE rather than 0.
1301 * cfghooks.c (debug): Likewise.
1302 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
1303 (struct dump_option_value_info): Convert to...
1304 (struct kv_pair): ...this template type.
1305 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
1307 (optinfo_verbosity_options): Likewise.
1308 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
1310 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
1311 than int for "optgroup_flags" param.
1312 (dump_generic_expr_loc): Use dump_flags_t rather than int for
1314 (dump_dec): Likewise.
1315 (dump_finish): Use TDF_NONE rather than 0.
1316 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
1317 rather than int for "optgroup_flags" param. Use TDF_NONE rather
1318 than 0. Update for change to option_ptr.
1319 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
1320 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
1321 0. Update for changes to optinfo_verbosity_options and
1323 (opt_info_switch_p): Convert optgroup_flags from int to
1325 (dump_basic_block): Use dump_flags_t rather than int
1326 for "dump_kind" param.
1327 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
1328 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
1329 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
1330 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
1331 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
1332 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
1333 TDF_NONE): Convert from macros to...
1334 (enum dump_flag): ...this new enum.
1335 (dump_flags_t): Update to use enum.
1336 (operator|, operator&, operator~, operator|=, operator&=):
1337 Implement for dump_flags_t.
1338 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
1339 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
1340 Convert from macros to...
1341 (enum optgroup_flag): ...this new enum.
1342 (optgroup_flags_t): New typedef.
1343 (operator|, operator|=): Implement for optgroup_flags_t.
1344 (struct dump_file_info): Convert field "alt_flags" to
1345 dump_flags_t. Convert field "optgroup_flags" to
1347 (dump_basic_block): Use dump_flags_t rather than int for param.
1348 (dump_generic_expr_loc): Likewise.
1349 (dump_dec): Likewise.
1350 (dump_register): Convert param "optgroup_flags" to
1352 (opt_info_enable_passes): Likewise.
1353 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
1355 * gimple-pretty-print.c (debug): Likewise.
1356 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
1357 (merged_store_group::apply_stores): Likewise.
1358 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
1359 * gimple.c (verify_gimple_pp): Likewise.
1360 * graphite-poly.c (print_pbb_body): Likewise.
1361 * passes.c (pass_manager::register_one_dump_file): Convert
1362 local "optgroup_flags" to optgroup_flags_t.
1363 * print-tree.c (print_node): Use TDF_NONE rather than 0.
1365 (debug_body): Likewise.
1366 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
1367 to optgroup_flags_t.
1368 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
1370 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
1371 (convert_mult_to_fma): Likewise.
1372 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
1373 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
1374 * tree-vect-data-refs.c (dump_lower_bound): Convert param
1375 "dump_kind" to dump_flags_t.
1377 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1379 * config/rs6000/rs6000.c (min, max): Delete.
1381 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
1383 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
1384 -mabi=spe and -mabi=no-spe.
1386 2018-06-08 Martin Liska <mliska@suse.cz>
1388 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
1389 where we expect an existing summary.
1391 2018-06-08 Martin Liska <mliska@suse.cz>
1393 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
1394 * ipa-inline.h (estimate_edge_growth): Likewise.
1396 2018-06-08 Martin Liska <mliska@suse.cz>
1398 * cgraph.c (function_version_hasher::hash): Use
1399 cgraph_node::get_uid ().
1400 (function_version_hasher::equal):
1401 * cgraph.h (cgraph_node::get_uid): New method.
1402 * ipa-inline.c (update_caller_keys): Use
1403 cgraph_node::get_uid ().
1404 (update_callee_keys): Likewise.
1405 * ipa-utils.c (searchc): Likewise.
1406 (ipa_reduced_postorder): Likewise.
1407 * lto-cgraph.c (input_node): Likewise.
1408 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
1409 * symbol-summary.h (symtab_insertion): Likewise.
1410 (symtab_removal): Likewise.
1411 (symtab_duplication): Likewise.
1412 * tree-pretty-print.c (dump_function_header): Likewise.
1413 * tree-sra.c (convert_callers_for_node): Likewise.
1415 2018-06-08 Martin Liska <mliska@suse.cz>
1417 * cgraph.c (symbol_table::create_edge): Always assign a new
1419 (symbol_table::free_edge): Do not recycle numbers.
1420 * cgraph.h (cgraph_edge::get): New method.
1421 * symbol-summary.h (symtab_removal): Use it.
1422 (symtab_duplication): Likewise.
1423 (call_summary::hashable_uid): Remove.
1425 2018-06-08 Martin Liska <mliska@suse.cz>
1427 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
1428 (initialize_growth_caches): Remove.
1429 (free_growth_caches): Likewise.
1430 (do_estimate_edge_time): Use edge_growth_cache.
1431 (do_estimate_edge_size): Likewise.
1432 (do_estimate_edge_hints): Likewise.
1433 * ipa-inline.c (reset_edge_caches): Likewise.
1434 (recursive_inlining): Likewise.
1435 (inline_small_functions): Likewise.
1436 * ipa-inline.h (initialize_growth_caches): Remove.
1437 (estimate_edge_size): Likewise.
1438 (estimate_edge_time): Likewise.
1439 (estimate_edge_hints): Likewise.
1440 (reset_edge_growth_cache): Likewise.
1441 * symbol-summary.h (call_summary::remove): New method.
1443 2018-06-08 Martin Liska <mliska@suse.cz>
1445 * ipa-cp.c (class edge_clone_summary): New summary.
1446 (grow_edge_clone_vectors): Remove.
1447 (ipcp_edge_duplication_hook): Remove.
1448 (class edge_clone_summary_t): New call_summary class.
1449 (ipcp_edge_removal_hook): Remove.
1450 (edge_clone_summary_t::duplicate): New function.
1451 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
1452 (create_specialized_node): Likewise.
1453 (ipcp_driver): Initialize edge_clone_summaries and do not
1456 2018-06-08 Martin Liska <mliska@suse.cz>
1458 * symbol-summary.h (get): New function.
1459 (call_summary::m_initialize_when_cloning): New class member.
1461 2018-06-08 Martin Liska <mliska@suse.cz>
1463 * cgraph.c (cgraph_node::remove): Do not recycle uid.
1464 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
1465 (symbol_table::allocate_cgraph_symbol): Do not set uid.
1466 * passes.c (uid_hash_t): Record removed_nodes by their uids.
1467 (remove_cgraph_node_from_order): Use the removed_nodes set.
1468 (do_per_function_toporder): Likwise.
1469 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
1470 instead of summary_uid.
1471 (symtab_removal): Likewise.
1472 (symtab_duplication): Likewise.
1474 2018-06-08 Martin Liska <mliska@suse.cz>
1476 * ipa-cp.c (ipcp_store_bits_results): Use
1477 ipcp_transformation_sum.
1478 (ipcp_store_vr_results): Likewise.
1479 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
1481 (ipcp_transformation_initialize): ... this.
1482 (ipa_set_node_agg_value_chain):
1483 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
1484 (write_ipcp_transformation_info): Likewise.
1485 (read_ipcp_transformation_info): Likewise.
1486 (ipcp_update_bits): Likewise.
1487 (ipcp_update_vr): Likewise.
1488 (ipcp_transform_function): Likewise.
1489 * ipa-prop.h: Rename ipcp_transformation_summary to
1490 ipcp_transformation.
1491 (class ipcp_transformation_t): New function summary.
1492 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
1493 (ipa_get_agg_replacements_for_node): Likewise.
1495 2018-06-08 Martin Liska <mliska@suse.cz>
1497 * ipa-pure-const.c (struct funct_state_d): Do it class instead
1499 (class funct_state_summary_t): New function_summary class.
1500 (has_function_state): Remove.
1501 (get_function_state): Likewise.
1502 (set_function_state): Likewise.
1503 (add_new_function): Likewise.
1504 (funct_state_summary_t::insert): New function.
1505 (duplicate_node_data): Remove.
1506 (remove_node_data): Remove.
1507 (funct_state_summary_t::duplicate): New function.
1508 (register_hooks): Create new funct_state_summaries.
1509 (pure_const_generate_summary): Use it.
1510 (pure_const_write_summary): Likewise.
1511 (pure_const_read_summary): Likewise.
1512 (propagate_pure_const): Likewise.
1513 (propagate_nothrow): Likewise.
1514 (dump_malloc_lattice): Likewise.
1515 (propagate_malloc): Likewise.
1516 (execute): Do not register hooks, just remove summary
1518 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
1521 2018-06-08 Martin Liska <mliska@suse.cz>
1523 * ipa-reference.c (remove_node_data): Remove.
1524 (duplicate_node_data): Likewise.
1525 (class ipa_ref_var_info_summary_t): New class.
1526 (class ipa_ref_opt_summary_t): Likewise.
1527 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1528 (get_reference_optimization_summary): Use
1529 ipa_ref_opt_sum_summaries.
1530 (set_reference_vars_info): Remove.
1531 (set_reference_optimization_summary): Likewise.
1532 (ipa_init): Create summaries.
1533 (init_function_info): Use function summary.
1534 (ipa_ref_opt_summary_t::duplicate): New function.
1535 (ipa_ref_opt_summary_t::remove): New function.
1536 (get_read_write_all_from_node): Fix GNU coding style.
1537 (propagate): Use function summary.
1538 (write_node_summary_p): Fix GNU coding style.
1539 (stream_out_bitmap): Likewise.
1540 (ipa_reference_read_optimization_summary): Use function summary.
1541 (ipa_reference_c_finalize): Do not release hooks.
1543 2018-06-08 Martin Liska <mliska@suse.cz>
1545 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1546 (analyze_function_body): Extract multiple calls of get_create.
1547 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1548 * ipa-inline.c (recursive_inlining): Use ::get method.
1549 * ipa-inline.h (estimate_edge_growth): Likewise.
1551 2018-06-08 Martin Liska <mliska@suse.cz>
1553 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1555 (hsa_function_summary::hsa_function_summary): Use the new enum
1557 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1558 * hsa-gen.c (hsa_get_host_function): Likewise.
1559 (get_brig_function_name): Likewise.
1560 * ipa-hsa.c (process_hsa_functions): Likewise.
1561 (ipa_hsa_write_summary): Likewise.
1562 * symbol-summary.h (symtab_duplication): Use ::get function/
1563 (get): New function.
1565 2018-06-08 Martin Liska <mliska@suse.cz>
1567 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1569 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1570 (hsa_register_kernel): Likewise.
1571 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1572 * hsa-gen.c (hsa_get_host_function): Likewise.
1573 (get_brig_function_name): Likewise.
1574 (generate_hsa): Likewise.
1575 (pass_gen_hsail::execute): Likewise.
1576 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1577 (devirtualization_time_bonus): Likewise.
1578 (ipcp_propagate_stage): Likewise.
1579 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1580 (edge_set_predicate): Likewise.
1581 (evaluate_conditions_for_known_args): Likewise.
1582 (evaluate_properties_for_edge): Likewise.
1583 (ipa_fn_summary::reset): Likewise.
1584 (ipa_fn_summary_t::duplicate): Likewise.
1585 (dump_ipa_call_summary): Likewise.
1586 (ipa_dump_fn_summary): Likewise.
1587 (analyze_function_body): Likewise.
1588 (compute_fn_summary): Likewise.
1589 (estimate_edge_devirt_benefit): Likewise.
1590 (estimate_edge_size_and_time): Likewise.
1591 (estimate_calls_size_and_time): Likewise.
1592 (estimate_node_size_and_time): Likewise.
1593 (inline_update_callee_summaries): Likewise.
1594 (remap_edge_change_prob): Likewise.
1595 (remap_edge_summaries): Likewise.
1596 (ipa_merge_fn_summary_after_inlining): Likewise.
1597 (ipa_update_overall_fn_summary): Likewise.
1598 (read_ipa_call_summary): Likewise.
1599 (inline_read_section): Likewise.
1600 (write_ipa_call_summary): Likewise.
1601 (ipa_fn_summary_write): Likewise.
1602 (ipa_free_fn_summary): Likewise.
1603 * ipa-hsa.c (process_hsa_functions): Likewise.
1604 (ipa_hsa_write_summary): Likewise.
1605 (ipa_hsa_read_section): Likewise.
1606 * ipa-icf.c (sem_function::merge): Likewise.
1607 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1608 (do_estimate_edge_time): Likewise.
1609 (estimate_size_after_inlining): Likewise.
1610 (estimate_growth): Likewise.
1611 (growth_likely_positive): Likewise.
1612 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1613 (inline_call): Likewise.
1614 * ipa-inline.c (caller_growth_limits): Likewise.
1615 (can_inline_edge_p): Likewise.
1616 (can_inline_edge_by_limits_p): Likewise.
1617 (compute_uninlined_call_time): Likewise.
1618 (compute_inlined_call_time): Likewise.
1619 (want_inline_small_function_p): Likewise.
1620 (edge_badness): Likewise.
1621 (update_caller_keys): Likewise.
1622 (update_callee_keys): Likewise.
1623 (recursive_inlining): Likewise.
1624 (inline_small_functions): Likewise.
1625 (inline_to_all_callers_1): Likewise.
1626 (dump_overall_stats): Likewise.
1627 (early_inline_small_functions): Likewise.
1628 (early_inliner): Likewise.
1629 * ipa-inline.h (estimate_edge_growth): Likewise.
1630 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1631 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1632 * ipa-prop.h (IPA_NODE_REF): Likewise.
1633 (IPA_EDGE_REF): Likewise.
1634 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1635 (propagate_malloc): Likewise.
1636 * ipa-split.c (execute_split_functions): Likewise.
1637 * symbol-summary.h: Rename get to get_create.
1639 (get_create): Likewise.
1640 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1642 2018-06-08 Martin Liska <mliska@suse.cz>
1644 * symbol-summary.h (release): Move definition out of class
1646 (symtab_removal): Likewise.
1647 (symtab_duplication): Likewise.
1649 2018-06-08 Martin Liska <mliska@suse.cz>
1651 * symbol-summary.h (function_summary): Move constructor
1652 implementation out of class declaration.
1653 (release): Likewise.
1654 (symtab_insertion): Likewise.
1655 (symtab_removal): Likewise.
1656 (symtab_duplication): Likewise.
1659 2018-06-08 Martin Liska <mliska@suse.cz>
1661 * Makefile.in: Remove support for MPX (macros, related functions,
1662 fields in cgraph_node, ...).
1663 * builtin-types.def (BT_BND): Likewise.
1664 (BT_FN_BND_CONST_PTR): Likewise.
1665 (BT_FN_CONST_PTR_BND): Likewise.
1666 (BT_FN_VOID_PTR_BND): Likewise.
1667 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1668 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1669 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1670 (expand_builtin_mempcpy_with_bounds): Likewise.
1671 (expand_builtin_memset_with_bounds): Likewise.
1672 (expand_builtin_memset_args): Likewise.
1673 (std_expand_builtin_va_start): Likewise.
1674 (expand_builtin): Likewise.
1675 (expand_builtin_with_bounds): Likewise.
1676 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1677 (DEF_LIB_BUILTIN_CHKP): Likewise.
1678 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1679 (DEF_CHKP_BUILTIN): Likewise.
1680 (BUILT_IN_MEMCPY): Likewise.
1681 (BUILT_IN_MEMMOVE): Likewise.
1682 (BUILT_IN_MEMPCPY): Likewise.
1683 (BUILT_IN_MEMSET): Likewise.
1684 (BUILT_IN_STPCPY): Likewise.
1685 (BUILT_IN_STRCAT): Likewise.
1686 (BUILT_IN_STRCHR): Likewise.
1687 (BUILT_IN_STRCPY): Likewise.
1688 (BUILT_IN_STRLEN): Likewise.
1689 (BUILT_IN_MEMCPY_CHK): Likewise.
1690 (BUILT_IN_MEMMOVE_CHK): Likewise.
1691 (BUILT_IN_MEMPCPY_CHK): Likewise.
1692 (BUILT_IN_MEMSET_CHK): Likewise.
1693 (BUILT_IN_STPCPY_CHK): Likewise.
1694 (BUILT_IN_STRCAT_CHK): Likewise.
1695 (BUILT_IN_STRCPY_CHK): Likewise.
1696 * calls.c (store_bounds): Likewise.
1697 (emit_call_1): Likewise.
1698 (special_function_p): Likewise.
1699 (maybe_warn_nonstring_arg): Likewise.
1700 (initialize_argument_information): Likewise.
1701 (finalize_must_preallocate): Likewise.
1702 (compute_argument_addresses): Likewise.
1703 (expand_call): Likewise.
1704 * cfgexpand.c (expand_call_stmt): Likewise.
1705 (expand_return): Likewise.
1706 (expand_gimple_stmt_1): Likewise.
1707 (pass_expand::execute): Likewise.
1708 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1709 (cgraph_node::remove): Likewise.
1710 (cgraph_node::dump): Likewise.
1711 (cgraph_node::verify_node): Likewise.
1712 * cgraph.h (chkp_function_instrumented_p): Likewise.
1713 (symtab_node::get_alias_target): Likewise.
1714 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1715 (cgraph_local_p): Likewise.
1716 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1717 (cgraph_edge::rebuild_references): Likewise.
1718 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1719 (walk_polymorphic_call_targets): Likewise.
1720 (cgraph_node::expand_thunk): Likewise.
1721 (symbol_table::output_weakrefs): Likewise.
1722 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1723 (ix86_handle_option): Likewise.
1724 * config/i386/constraints.md: Likewise.
1725 * config/i386/i386-builtin-types.def (BND): Likewise.
1729 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1730 (BDESC_FIRST): Likewise.
1732 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1733 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1734 * config/i386/i386.c (enum reg_class): Likewise.
1735 (ix86_target_string): Likewise.
1736 (ix86_option_override_internal): Likewise.
1737 (ix86_conditional_register_usage): Likewise.
1738 (ix86_valid_target_attribute_inner_p): Likewise.
1739 (ix86_set_indirect_branch_type): Likewise.
1740 (ix86_set_current_function): Likewise.
1741 (ix86_function_arg_regno_p): Likewise.
1742 (init_cumulative_args): Likewise.
1743 (ix86_function_arg_advance): Likewise.
1744 (ix86_function_arg): Likewise.
1745 (ix86_pass_by_reference): Likewise.
1746 (ix86_function_value_regno_p): Likewise.
1747 (ix86_function_value_1): Likewise.
1748 (ix86_function_value_bounds): Likewise.
1749 (ix86_return_in_memory): Likewise.
1750 (ix86_setup_incoming_vararg_bounds): Likewise.
1751 (ix86_va_start): Likewise.
1752 (indirect_thunk_need_prefix): Likewise.
1753 (print_reg): Likewise.
1754 (ix86_print_operand): Likewise.
1755 (ix86_expand_call): Likewise.
1756 (ix86_output_function_return): Likewise.
1757 (reg_encoded_number): Likewise.
1758 (BDESC_VERIFYS): Likewise.
1759 (ix86_init_mpx_builtins): Likewise.
1760 (ix86_init_builtins): Likewise.
1761 (ix86_emit_cmove): Likewise.
1762 (ix86_emit_move_max): Likewise.
1763 (ix86_expand_builtin): Likewise.
1764 (ix86_builtin_mpx_function): Likewise.
1765 (ix86_get_arg_address_for_bt): Likewise.
1766 (ix86_load_bounds): Likewise.
1767 (ix86_store_bounds): Likewise.
1768 (ix86_load_returned_bounds): Likewise.
1769 (ix86_store_returned_bounds): Likewise.
1770 (ix86_class_likely_spilled_p): Likewise.
1771 (ix86_hard_regno_mode_ok): Likewise.
1772 (x86_order_regs_for_local_alloc): Likewise.
1773 (ix86_mitigate_rop): Likewise.
1774 (ix86_bnd_prefixed_insn_p): Likewise.
1775 (ix86_mpx_bound_mode): Likewise.
1776 (ix86_make_bounds_constant): Likewise.
1777 (ix86_initialize_bounds): Likewise.
1778 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1779 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1780 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1781 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1782 (TARGET_CHKP_BOUND_MODE): Likewise.
1783 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1784 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1785 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1786 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1787 * config/i386/i386.h (TARGET_MPX): Likewise.
1788 (TARGET_MPX_P): Likewise.
1789 (VALID_BND_REG_MODE): Likewise.
1790 (FIRST_BND_REG): Likewise.
1791 (LAST_BND_REG): Likewise.
1792 (enum reg_class): Likewise.
1793 (BND_REG_P): Likewise.
1794 (BND_REGNO_P): Likewise.
1795 (BNDmode): Likewise.
1796 (ADJUST_INSN_LENGTH): Likewise.
1797 * config/i386/i386.md: Likewise.
1798 * config/i386/i386.opt: Likewise.
1799 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1800 (defined): Likewise.
1801 (LINK_MPX): Likewise.
1802 (MPX_SPEC): Likewise.
1803 (LIBMPX_SPEC): Likewise.
1804 (LIBMPXWRAPPERS_SPEC): Likewise.
1805 (CHKP_SPEC): Likewise.
1806 * config/i386/predicates.md: Likewise.
1807 * dbxout.c (dbxout_type): Likewise.
1808 * doc/extend.texi: Likewise.
1809 * doc/invoke.texi: Likewise.
1810 * doc/md.texi: Likewise.
1811 * doc/tm.texi: Likewise.
1812 * doc/tm.texi.in: Likewise.
1813 * dwarf2out.c (is_base_type): Likewise.
1814 (gen_formal_types_die): Likewise.
1815 (gen_subprogram_die): Likewise.
1816 (gen_type_die_with_usage): Likewise.
1817 (gen_decl_die): Likewise.
1818 (dwarf2out_late_global_decl): Likewise.
1819 * expr.c (expand_assignment): Likewise.
1820 (emit_storent_insn): Likewise.
1821 (store_expr_with_bounds): Likewise.
1822 (store_expr): Likewise.
1823 (expand_expr_real_1): Likewise.
1824 * expr.h (store_expr_with_bounds): Likewise.
1825 * function.c (use_register_for_decl): Likewise.
1826 (struct bounds_parm_data): Likewise.
1827 (assign_parms_augmented_arg_list): Likewise.
1828 (assign_parm_find_entry_rtl): Likewise.
1829 (assign_parm_is_stack_parm): Likewise.
1830 (assign_parm_load_bounds): Likewise.
1831 (assign_bounds): Likewise.
1832 (assign_parms): Likewise.
1833 (expand_function_start): Likewise.
1834 * gcc.c (CHKP_SPEC): Likewise.
1835 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1836 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1837 (wrestrict_dom_walker::check_call): Likewise.
1838 * gimple.c (gimple_build_call_from_tree): Likewise.
1839 * gimple.h (enum gf_mask): Likewise.
1840 (gimple_call_with_bounds_p): Likewise.
1841 (gimple_call_set_with_bounds): Likewise.
1842 * gimplify.c (gimplify_init_constructor): Likewise.
1843 * ipa-cp.c (initialize_node_lattices): Likewise.
1844 (propagate_constants_across_call): Likewise.
1845 (find_more_scalar_values_for_callers_subset): Likewise.
1846 * ipa-hsa.c (process_hsa_functions): Likewise.
1847 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1848 * ipa-icf.c (sem_function::merge): Likewise.
1849 * ipa-inline.c (early_inliner): Likewise.
1850 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1851 (warn_function_cold): Likewise.
1852 (propagate_pure_const): Likewise.
1853 * ipa-ref.h (enum GTY): Likewise.
1854 * ipa-split.c (find_retbnd): Likewise.
1855 (consider_split): Likewise.
1856 (split_function): Likewise.
1857 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1858 * ipa.c (walk_polymorphic_call_targets): Likewise.
1859 (symbol_table::remove_unreachable_nodes): Likewise.
1860 (process_references): Likewise.
1861 (cgraph_build_static_cdtor_1): Likewise.
1862 * lto-cgraph.c (lto_output_node): Likewise.
1863 (output_refs): Likewise.
1864 (compute_ltrans_boundary): Likewise.
1865 (input_overwrite_node): Likewise.
1866 (input_node): Likewise.
1867 (input_cgraph_1): Likewise.
1868 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1869 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1870 (class pass_chkp_instrumentation_passes): Likewise.
1871 (make_pass_chkp_instrumentation_passes): Likewise.
1872 * passes.def: Likewise.
1873 * rtl.h (struct GTY): Likewise.
1874 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1875 * stor-layout.c (layout_type): Likewise.
1876 * symtab.c: Likewise.
1877 * target.def: Likewise.
1878 * targhooks.c (default_chkp_bound_type): Likewise.
1879 (default_chkp_bound_mode): Likewise.
1880 (default_builtin_chkp_function): Likewise.
1881 (default_chkp_function_value_bounds): Likewise.
1882 (default_chkp_make_bounds_constant): Likewise.
1883 (default_chkp_initialize_bounds): Likewise.
1884 * targhooks.h (default_chkp_bound_type): Likewise.
1885 (default_chkp_bound_mode): Likewise.
1886 (default_builtin_chkp_function): Likewise.
1887 (default_chkp_function_value_bounds): Likewise.
1888 (default_chkp_make_bounds_constant): Likewise.
1889 (default_chkp_initialize_bounds): Likewise.
1890 * toplev.c (compile_file): Likewise.
1891 (process_options): Likewise.
1892 * tree-core.h (DEF_BUILTIN): Likewise.
1893 (DEF_BUILTIN_CHKP): Likewise.
1894 * tree-inline.c (declare_return_variable): Likewise.
1895 (remap_gimple_stmt): Likewise.
1896 (copy_bb): Likewise.
1897 (initialize_inlined_parameters): Likewise.
1898 (expand_call_inline): Likewise.
1899 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1900 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1901 (make_pass_ipa_chkp_produce_thunks): Likewise.
1902 (make_pass_chkp): Likewise.
1903 (make_pass_chkp_opt): Likewise.
1904 (make_pass_chkp_instrumentation_passes): Likewise.
1905 * tree-pretty-print.c (dump_generic_node): Likewise.
1906 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1907 * tree-ssa-dce.c (propagate_necessity): Likewise.
1908 (eliminate_unnecessary_stmts): Likewise.
1909 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1910 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1911 * tree-ssa-sccvn.h: Likewise.
1912 * tree-ssa-strlen.c (get_string_length): Likewise.
1913 (valid_builtin_call): Likewise.
1914 (adjust_last_stmt): Likewise.
1915 (handle_builtin_strchr): Likewise.
1916 (handle_builtin_strcpy): Likewise.
1917 (handle_builtin_stxncpy): Likewise.
1918 (handle_builtin_memcpy): Likewise.
1919 (handle_builtin_strcat): Likewise.
1920 (strlen_check_and_optimize_stmt): Likewise.
1921 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1922 * tree-streamer-in.c: Likewise.
1923 * tree-streamer.c (record_common_node): Likewise.
1924 * tree.c (tree_code_size): Likewise.
1925 (wide_int_to_tree_1): Likewise.
1926 (type_contains_placeholder_1): Likewise.
1927 (build_common_tree_nodes): Likewise.
1928 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1929 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1930 (POINTER_BOUNDS_P): Likewise.
1931 (BOUNDED_TYPE_P): Likewise.
1932 (BOUNDED_P): Likewise.
1933 (CALL_WITH_BOUNDS_P): Likewise.
1934 (pointer_bounds_type_node): Likewise.
1935 * value-prof.c (gimple_ic): Likewise.
1936 * var-tracking.c (vt_add_function_parameters): Likewise.
1937 * varasm.c (make_decl_rtl): Likewise.
1938 (assemble_start_function): Likewise.
1939 (output_constant): Likewise.
1940 (maybe_assemble_visibility): Likewise.
1941 * varpool.c (ctor_for_folding): Likewise.
1942 * chkp-builtins.def: Remove.
1943 * ipa-chkp.c: Remove.
1944 * ipa-chkp.h: Remove.
1945 * rtl-chkp.c: Remove.
1946 * rtl-chkp.h: Remove.
1947 * tree-chkp-opt.c: Remove.
1948 * tree-chkp.c: Remove.
1949 * tree-chkp.h: Remove.
1951 2018-06-07 Carl Love <cel@us.ibm.com>
1953 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1954 vextract_fp_from_shortl): Add BE support.
1956 2018-06-07 Paul Koning <ni1d@arrl.net>
1958 * compare-elim.c (try_merge_compare): Don't merge compare if
1959 address contains a side effect.
1960 (try_eliminate_compare): Likewise.
1962 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1964 * config.gcc: Support "tremont".
1965 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1966 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1968 * config/i386/i386.c (m_TREMONT): Define.
1969 (processor_target_table): Add "tremont".
1970 (PTA_TREMONT): Define.
1971 (ix86_lea_outperforms): Add TARGET_TREMONT.
1972 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1973 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1974 and M_INTEL_GOLDMONT_PLUS.
1975 (fold_builtin_cpu): Add "tremont".
1976 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1977 (ix86_option_override_internal): Add "tremont".
1978 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1979 (processor_type): Add PROCESSOR_TREMONT.
1980 * config/i386/x86-tune.def: Add m_TREMONT.
1981 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1983 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1985 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1986 symbol defined for msp430i* devices to be lower case.
1988 gcc/testsuite/gcc.target/msp430/
1989 * msp430i-device-symbol.c: New test.
1990 * msp430f-device-symbol.c: New test.
1991 * msp430.h: New test header file.
1993 2018-06-07 Richard Biener <rguenther@suse.de>
1995 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1996 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1997 Properly wrap signed arithmetic if overflow wraps.
1999 2018-06-07 Jakub Jelinek <jakub@redhat.com>
2001 PR tree-optimization/69615
2002 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
2003 of a cast from a same precision integral SSA_NAME in a bb dominated
2004 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
2005 cast to utype if rhs2 has already a compatible type.
2007 2018-06-07 Richard Biener <rguenther@suse.de>
2009 PR tree-optimization/85935
2010 * graphite-scop-detection.c (find_params_in_bb): Analyze
2011 condition operands with respect to the correct loop. Assert
2012 the analysis doesn't fail.
2014 2018-06-04 Carl Love <cel@us.ibm.com>
2016 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
2017 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
2018 as it is slightly cheaper.
2019 (first_match_or_eos_index_<mode>):
2020 Calculate index using natural element order.
2021 (first_match_index_<mode>):
2022 Calculate index using natural element order.
2023 (first_match_or_eos_index_<mode>):
2024 Calculate index using natural order.
2025 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
2026 for BE and LE modes.
2027 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
2028 P9V_BUILTIN_VCLZLSBB_V16QI.
2029 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
2032 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2034 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
2035 indentation and line wrap for many prototypes. Add missing
2036 @smallexample directives around block of prototypes for vec_xl and
2039 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
2041 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
2042 track if we pass or return IEEE 128-bit floating point.
2043 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
2044 C++ mangling that is compatible with GCC 8.1.
2045 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
2046 (init_cumulative_args): Note if we pass or return IEEE 128-bit
2047 floating point types.
2048 (rs6000_function_arg_advance_1): Likewise.
2049 (rs6000_mangle_type): Optionally generate mangled names that match
2050 what GCC 8.1 generated for IEEE 128-bit floating point types.
2051 (rs6000_globalize_decl_name): If we have an external function that
2052 passes or returns IEEE 128-bit types, generate a weak reference
2053 from the mangled name used in GCC 8.1 to the current mangled
2055 (rs6000_init_builtins): Make __ibm128 use the long double type if
2056 long double is IBM extended double. Make __float128 use the long
2057 double type if long double is IEEE 128-bit.
2060 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
2061 macro for __ibm128 built-in functions.
2062 (PACK_IF): Add __ibm128 pack/unpack functions.
2063 (UNPACK_IF): Likewise.
2064 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
2065 enable long double built-in functions if long double is IEEE
2066 128-bit floating point.
2067 (rs6000_invalid_builtin): Update long double built-in function
2069 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
2070 functions, adjust the built-in function to use the long double
2071 built-in function if __ibm128 and long double are the same type.
2072 * doc/extend.texi (PowerPC builtins): Update documention for
2073 __builtin_{,un}pack_longdouble. Add documentation for
2074 __builtin_{,un}pack_ibm128.
2076 2018-06-06 Jim Wilson <jimw@sifive.com>
2078 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
2079 (struct machine_function): New field interrupt_mode.
2080 (riscv_handle_type_attribute): New function. Add forward declaration.
2081 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
2082 (riscv_expand_epilogue): Check interrupt_mode field.
2083 (riscv_set_current_function): Check interrupt attribute args and
2084 set interrupt_mode field.
2085 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
2086 (riscv_sret, riscv_uret): New.
2087 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
2088 new arguments to interrupt attribute.
2090 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
2093 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
2094 Don't handle -mcpu=power8 if -mpower9-vector is also used.
2096 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
2098 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
2099 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
2100 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2101 several redundant entries.
2103 2018-06-06 David Malcolm <dmalcolm@redhat.com>
2105 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
2106 type from "rtx" to "rtx_insn *".
2107 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
2108 for local "call_insn", removing cast.
2109 (ix86_expand_call): Likewise, introducing a "call_insn" local.
2111 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
2113 PR tree-optimization/86066
2114 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
2115 for BIT_INSERT_EXPR stores.
2117 2018-06-06 Richard Biener <rguenther@suse.de>
2119 PR tree-optimization/86062
2120 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2121 component refs ontop
2122 of to be offsetted base.
2124 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2126 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
2127 to be static and remove check on interrupt attribute name.
2129 gcc/testsuite/gcc.target/msp430/
2130 * function-attributes-4.c: New test.
2131 * static-interrupts.c: New test.
2133 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
2135 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2136 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
2138 2018-06-05 Steve Ellcey <sellcey@cavium.com>
2141 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
2143 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
2144 Remove second argument, change how error is called.
2145 (aarch64_layout_arg): Remove second argument from
2146 aarch64_err_no_fpadvsimd call.
2147 (aarch64_init_cumulative_args): Ditto.
2148 (aarch64_gimplify_va_arg_expr): Ditto.
2149 * config/aarch64/aarch64.md (mov<mode>): Ditto.
2151 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
2153 * config/i386/i386.md (simple_return_indirect_internal): New expander.
2154 (*simple_return_indirect_internal<mode>): Rename from
2155 simple_return_indirect_internal. Use W mode iterator.
2156 (rstorssp): New expander.
2157 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
2158 (clrssbsy): New expander.
2159 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
2161 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
2163 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
2165 (cmse_check_pointed_object): Likewise.
2167 2018-06-05 Martin Liska <mliska@suse.cz>
2169 PR gcov-profile/47618
2170 * doc/invoke.texi: Document how -fprofile-dir format
2173 2018-06-05 Richard Biener <rguenther@suse.de>
2175 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
2176 removal pretend DOM info isn't available so we do not update
2177 it and only remove edges, not dominated blocks. Actually free
2178 DOM info in case we removed something. Remove unreachable blocks.
2179 (mfb_keep_latches): Work with either DOM info or marked backedges.
2180 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
2181 first. Mark backedges if DOM info isn't available.
2182 (Re-)compute DOM info after cleanup_control_flow_pre.
2184 2018-06-05 Richard Biener <rguenther@suse.de>
2186 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
2187 (locus_discrim_hasher::hash): Adjust.
2188 (locus_discrim_hasher::equal): Likewise.
2189 (next_discriminator_for_locus): Work on line directly.
2190 (same_line_p): Pass in expanded locus1 as well.
2191 (assign_discriminators): Avoid redundant location expansions.
2193 2018-06-05 Richard Biener <rguenther@suse.de>
2195 PR tree-optimization/86046
2196 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
2197 if required after clearing TREE_ADDRESSABLE.
2199 2018-06-05 Richard Biener <rguenther@suse.de>
2201 PR tree-optimization/86047
2202 * tree-ssa-loop.c (for_each_index): Glob handling of all
2203 decls and constants and really handle all of them.
2205 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2208 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
2209 qualifier_void_pointer and qualifier_const_void_pointer.
2210 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
2211 (arm_init_builtins): Handle the above.
2212 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
2213 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
2214 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
2217 2018-06-05 Martin Liska <mliska@suse.cz>
2219 * auto-profile.c (read_autofdo_file): Do not use
2220 gcov_ctr_summary struct.
2221 (afdo_callsite_hot_enough_for_early_inline): Likewise.
2222 * coverage.c (struct counts_entry): Likewise.
2223 (read_counts_file): Read just single summary entry.
2224 (get_coverage_counts): Use gcov_summary struct.
2225 * coverage.h (get_coverage_counts): Likewise.
2226 * gcov-dump.c (dump_working_sets): Likewise.
2227 (tag_summary): Dump just single summary.
2228 * gcov-io.c (gcov_write_summary): Write just histogram
2230 (gcov_read_summary): Read just single summary.
2231 (compute_working_sets): Use gcov_summary struct.
2232 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
2233 of GCOV_COUNTERS_SUMMABLE.
2234 (GCOV_COUNTERS_SUMMABLE): Remove.
2235 (GCOV_FIRST_VALUE_COUNTER): Replace with
2236 GCOV_COUNTER_V_INTERVAL.
2237 (struct gcov_ctr_summary): Remove.
2238 (struct gcov_summary): Directly use fields of former
2240 (compute_working_sets): Use gcov_summary struct.
2241 * gcov.c (read_count_file): Do not use ctrs fields.
2242 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
2244 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
2246 * profile.c: Likewise.
2247 * profile.h: Likewise.
2249 2018-06-05 Martin Liska <mliska@suse.cz>
2251 PR gcov-profile/84846
2252 * gcov.c (output_lines): Print working directory only
2253 in intermediate format.
2255 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
2257 * config/s390/s390-builtin-types.def: Add void function type.
2258 * config/s390/s390-builtins.def: Use the function type for the
2261 2018-06-04 Jim Wilson <jimw@sifive.com>
2263 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
2265 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
2266 and maybe_eh_return. Change regno to unsigned int. Use new args to
2267 handle EH_RETURN_DATA_REGNO registers properly.
2268 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
2269 (riscv_expand_epilogue): Update comment. Change argument name and
2270 type. Update code to use new name and type. Pass new args to
2271 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
2273 * config/riscv/riscv.md (NORMAL_RETURN): New.
2274 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
2275 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
2276 (eh_return): Call gen_eh_return_internal and emit barrier.
2277 (eh_return_internal): Call riscv_expand_epilogue.
2279 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
2281 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
2282 bit_insertion field and declare can_be_merged_into method.
2283 (merged_store_group::can_be_merged_into): New method.
2284 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
2285 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
2286 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
2288 2018-06-04 Richard Biener <rguenther@suse.de>
2290 PR tree-optimization/85955
2291 * builtins.c (fold_builtin_sincos): Convert pointers to
2292 destination to appropriate type before dereferencing.
2294 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
2296 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
2298 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
2300 * expr.c (expand_expr_real_1): Force the operand into memory if
2301 its TYPE_MODE is BLKmode and if there is no integer mode for
2302 the number of bits being extracted.
2304 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2308 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2309 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
2311 2018-06-04 Richard Biener <rguenther@suse.de>
2313 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
2314 (cleanup_tree_cfg_noloop): ... single caller. Do
2315 start_recording_case_labels later.
2317 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
2319 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
2320 to _IMMINTRIN_H_INCLUDED.
2321 * config/i386/pconfigintrin.h: Ditto.
2322 * config/i386/waitpkgintrin.h: Ditto.
2323 * config/i386/immintrin.h: Add includes for sgxintrin.h,
2324 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
2325 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
2326 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
2327 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
2328 waitpkgintrin.h and cldemoteintrin.h.
2330 2018-06-04 Richard Biener <rguenther@suse.de>
2332 PR tree-optimization/86038
2333 * tracer.c (find_best_successor): Check probability for
2334 being initialized, bail out if not.
2336 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
2339 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
2340 of bits to ignore when comparing architectures.
2342 2018-06-04 Jakub Jelinek <jakub@redhat.com>
2344 PR tree-optimization/69615
2345 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
2346 maximum or minimum of the type, try to merge it also as if
2347 range1 is + [-, x - 1] or + [x + 1, -].
2350 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
2352 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
2354 PR tree-optimization/86034
2355 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
2356 the unsigned bitfield type in a bit insertion sequence if it does not
2357 have a larger precision than the bitfield size.
2358 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
2360 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
2362 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
2364 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2366 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
2367 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
2368 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
2369 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
2371 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
2373 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
2374 Disable -fdelete-null-pointer-checks for ELF toolchain.
2376 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2377 Kito Cheng <kito.cheng@gmail.com>
2379 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
2380 (nds32le-*-*, nds32be-*-*): Integrate checking process.
2381 (nds32*-*-*): Add glibc and uclibc conditions.
2382 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
2383 (TARGET_EXCEPT_UNWIND_INFO): Define.
2384 * config/nds32/elf.h: New file.
2385 * config/nds32/linux.h: New file.
2386 * config/nds32/nds32-elf.opt: New file.
2387 * config/nds32/nds32-linux.opt: New file.
2388 * config/nds32/nds32-fp-as-gp.c
2389 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
2390 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
2392 (nds32_asm_file_end): Ditto.
2393 (nds32_print_operand): Ditto.
2394 (nds32_insert_attributes): Ditto.
2395 (nds32_init_libfuncs): New function.
2396 (TARGET_HAVE_TLS): Define.
2397 (TARGET_INIT_LIBFUNCS): Define.
2398 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
2400 (TARGET_ELF): Apply different mcmodel setting.
2401 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
2402 been migrated into elf.h and linux.h files.
2403 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
2404 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
2405 (mcmodel): The content has been migrated into nds32-elf.opt and
2406 nds32-linux.opt files.
2407 * config/nds32/t-elf: New file.
2408 * config/nds32/t-linux: New file.
2410 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
2411 Shiva Chen <shiva0217@gmail.com>
2413 * config/nds32/constants.md (unspec_volatile_element): Add
2414 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
2415 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
2417 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
2418 (make_pass_nds32_fp_as_gp): Declare.
2419 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
2421 (nds32_asm_function_end_prologue): Remove unused asm output.
2422 (nds32_asm_function_begin_epilogue): Remove unused asm output.
2423 (nds32_asm_file_start): Output necessary fp_as_gp information.
2424 (nds32_option_override): Adjust register usage.
2425 (nds32_expand_prologue): Consider fp_as_gp situation.
2426 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
2427 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
2430 (simple_return): Ditto.
2431 (omit_fp_begin): Output special directive for fp_as_gp.
2432 (omit_fp_end): Output special directive for fp_as_gp.
2433 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
2434 mforbid-fp-as-gp): New options.
2436 2018-06-01 Mark Wielaard <mark@klomp.org>
2438 * dwarf2out.c (dwarf2out_finish): Remove generation of
2439 DW_AT_loclists_base.
2441 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
2443 * gimple-ssa-store-merging.c: Include gimple-fold.h.
2444 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
2445 (struct merged_store_group): Add bit_insertion field.
2446 (dump_char_array): Use standard hexadecimal format.
2447 (merged_store_group::merged_store_group): Set bit_insertion to false.
2448 (merged_store_group::apply_stores): Use optimal buffer size. Deal
2449 with BIT_INSERT_EXPR stores. Move up code updating the mask and
2450 also print the mask in the dump file.
2451 (pass_store_merging::gate): Minor tweak.
2452 (imm_store_chain_info::coalesce_immediate): Fix wrong association
2453 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
2454 stores with INTEGER_CST stores.
2455 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
2456 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
2457 and use it throughout. Generate bit insertion sequences if need be.
2458 (pass_store_merging::process_store): Remove redundant condition.
2459 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
2461 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
2463 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
2464 the 128-bit floating point types. Fix function comment.
2466 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2468 * config/aarch64/aarch64-simd.md
2469 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
2471 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
2474 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2476 PR tree-optimization/85989
2477 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
2479 (backprop::intersect_uses): Check it when deciding whether this
2480 is a backedge reference.
2481 (backprop::process_block): Add each phi to m_visited_phis
2482 after visiting it, then clear it at the end.
2484 2018-06-01 Richard Biener <rguenther@suse.de>
2486 * tree-vectorizer.h (vect_dr_stmt): New function.
2487 (vect_get_load_cost): Adjust.
2488 (vect_get_store_cost): Likewise.
2489 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
2490 Use vect_dr_stmt instead of DR_SMTT.
2491 (vect_record_base_alignments): Likewise.
2492 (vect_calculate_target_alignment): Likewise.
2493 (vect_compute_data_ref_alignment): Likewise and make static.
2494 (vect_update_misalignment_for_peel): Likewise.
2495 (vect_verify_datarefs_alignment): Likewise.
2496 (vector_alignment_reachable_p): Likewise.
2497 (vect_get_data_access_cost): Likewise. Pass down
2498 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
2499 (vect_get_peeling_costs_all_drs): Likewise.
2500 (vect_peeling_hash_get_lowest_cost): Likewise.
2501 (vect_enhance_data_refs_alignment): Likewise.
2502 (vect_find_same_alignment_drs): Likewise.
2503 (vect_analyze_data_refs_alignment): Likewise.
2504 (vect_analyze_group_access_1): Likewise.
2505 (vect_analyze_group_access): Likewise.
2506 (vect_analyze_data_ref_access): Likewise.
2507 (vect_analyze_data_ref_accesses): Likewise.
2508 (vect_vfa_segment_size): Likewise.
2509 (vect_small_gap_p): Likewise.
2510 (vectorizable_with_step_bound_p): Likewise.
2511 (vect_prune_runtime_alias_test_list): Likewise.
2512 (vect_analyze_data_refs): Likewise.
2513 (vect_supportable_dr_alignment): Likewise.
2514 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2515 (vect_gen_prolog_loop_niters): Likewise.
2516 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2517 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
2519 (vect_recog_mask_conversion_pattern): Likewise.
2520 (vect_try_gather_scatter_pattern): Likewise.
2521 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
2522 to vect_get_store_cost.
2523 (vect_get_store_cost): Get stmt_info instead of DR.
2524 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2525 (vect_get_load_cost): Get stmt_info instead of DR.
2527 2018-06-01 Richard Biener <rguenther@suse.de>
2530 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2531 vars wrapped in MEM_REFs.
2533 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2535 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2536 Fix subreg tests so that we only return a choice between
2537 GENERAL_REGS and FP_REGS if the original classes included both.
2539 2018-06-01 Richard Biener <rguenther@suse.de>
2542 * tree-ssa-structalias.c (get_function_part_constraint):
2543 Handle NULL fi->decl.
2544 (find_func_aliases_for_call): Properly handle indirect
2545 fi from direct call.
2546 (find_func_clobbers): Likewise.
2547 (ipa_pta_execute): Likewise.
2548 (create_variable_info_for): For functions that are ifunc_resolver
2549 resolve to a varinfo that contains the result of the resolver call.
2550 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2553 2018-05-31 Michael Collison <michael.collison@arm.com>
2555 * config/aarch64/aarch64.md:
2556 (*fix_to_zero_extenddfdi2): New pattern.
2557 * gcc.target/aarch64/fix_extend1.c: New testcase.
2559 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2563 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2564 and BUILT_IN_STRNCMP_EQ.
2565 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2566 BUILT_IN_STRNCMP_EQ.
2567 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2568 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2569 (gimple_fold_builtin): Likewise.
2570 * tree-ssa-strlen.c (compute_string_length): New function.
2571 (determine_min_obsize): New function.
2572 (handle_builtin_string_cmp): New function to handle calls to
2573 string compare functions.
2574 (strlen_optimize_stmt): Add handling to builtin string compare
2576 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2577 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2578 * tree.c (build_common_builtin_nodes): Add new defines of
2579 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2581 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2584 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2585 functions with naked attribute.
2587 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2589 * config/i386/sse.md (avx_vec_concat<mode>):
2590 Substitute concat_tg_mode mode attribute with xtg_mode.
2591 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2592 (concat_tg_mode): Remove mode attribute.
2594 2018-05-31 Martin Sebor <msebor@redhat.com>
2597 * calls.c (alloc_max_size): Correct a logic error/typo.
2598 Treat excessive arguments as infinite. Warn for invalid arguments.
2599 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2601 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2604 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2605 and movx for Haswell.
2607 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2608 Cesar Philippidis <cesar@codesourcery.com>
2611 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2612 when emitting error on private/firstprivate reductions.
2613 * omp-low.c (lower_omp_target): Avoid reference-type processing
2614 on pointers for firstprivate clause.
2616 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2618 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2621 * config/aarch64/aarch64-simd.md
2622 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2623 (aarch64_ld1_x3_<mode>): Likewise
2624 (aarch64_st1x2<VALLDIF:mode>): Likewise
2625 (aarch64_st1_x2_<mode>): Likewise
2626 (aarch64_st1x3<VALLDIF:mode>): Likewise
2627 (aarch64_st1_x3_<mode>): Likewise
2628 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2629 (vld1_s8_x3): Likewise.
2630 (vld1_u16_x3): Likewise.
2631 (vld1_s16_x3): Likewise.
2632 (vld1_u32_x3): Likewise.
2633 (vld1_s32_x3): Likewise.
2634 (vld1_u64_x3): Likewise.
2635 (vld1_s64_x3): Likewise.
2636 (vld1_f16_x3): Likewise.
2637 (vld1_f32_x3): Likewise.
2638 (vld1_f64_x3): Likewise.
2639 (vld1_p8_x3): Likewise.
2640 (vld1_p16_x3): Likewise.
2641 (vld1_p64_x3): Likewise.
2642 (vld1q_u8_x3): Likewise.
2643 (vld1q_s8_x3): Likewise.
2644 (vld1q_u16_x3): Likewise.
2645 (vld1q_s16_x3): Likewise.
2646 (vld1q_u32_x3): Likewise.
2647 (vld1q_s32_x3): Likewise.
2648 (vld1q_u64_x3): Likewise.
2649 (vld1q_s64_x3): Likewise.
2650 (vld1q_f16_x3): Likewise.
2651 (vld1q_f32_x3): Likewise.
2652 (vld1q_f64_x3): Likewise.
2653 (vld1q_p8_x3): Likewise.
2654 (vld1q_p16_x3): Likewise.
2655 (vld1q_p64_x3): Likewise.
2656 (vst1_s64_x2): Likewise.
2657 (vst1_u64_x2): Likewise.
2658 (vst1_f64_x2): Likewise.
2659 (vst1_s8_x2): Likewise.
2660 (vst1_p8_x2): Likewise.
2661 (vst1_s16_x2): Likewise.
2662 (vst1_p16_x2): Likewise.
2663 (vst1_s32_x2): Likewise.
2664 (vst1_u8_x2): Likewise.
2665 (vst1_u16_x2): Likewise.
2666 (vst1_u32_x2): Likewise.
2667 (vst1_f16_x2): Likewise.
2668 (vst1_f32_x2): Likewise.
2669 (vst1_p64_x2): Likewise.
2670 (vst1q_s8_x2): Likewise.
2671 (vst1q_p8_x2): Likewise.
2672 (vst1q_s16_x2): Likewise.
2673 (vst1q_p16_x2): Likewise.
2674 (vst1q_s32_x2): Likewise.
2675 (vst1q_s64_x2): Likewise.
2676 (vst1q_u8_x2): Likewise.
2677 (vst1q_u16_x2): Likewise.
2678 (vst1q_u32_x2): Likewise.
2679 (vst1q_u64_x2): Likewise.
2680 (vst1q_f16_x2): Likewise.
2681 (vst1q_f32_x2): Likewise.
2682 (vst1q_f64_x2): Likewise.
2683 (vst1q_p64_x2): Likewise.
2684 (vst1_s64_x3): Likewise.
2685 (vst1_u64_x3): Likewise.
2686 (vst1_f64_x3): Likewise.
2687 (vst1_s8_x3): Likewise.
2688 (vst1_p8_x3): Likewise.
2689 (vst1_s16_x3): Likewise.
2690 (vst1_p16_x3): Likewise.
2691 (vst1_s32_x3): Likewise.
2692 (vst1_u8_x3): Likewise.
2693 (vst1_u16_x3): Likewise.
2694 (vst1_u32_x3): Likewise.
2695 (vst1_f16_x3): Likewise.
2696 (vst1_f32_x3): Likewise.
2697 (vst1_p64_x3): Likewise.
2698 (vst1q_s8_x3): Likewise.
2699 (vst1q_p8_x3): Likewise.
2700 (vst1q_s16_x3): Likewise.
2701 (vst1q_p16_x3): Likewise.
2702 (vst1q_s32_x3): Likewise.
2703 (vst1q_s64_x3): Likewise.
2704 (vst1q_u8_x3): Likewise.
2705 (vst1q_u16_x3): Likewise.
2706 (vst1q_u32_x3): Likewise.
2707 (vst1q_u64_x3): Likewise.
2708 (vst1q_f16_x3): Likewise.
2709 (vst1q_f32_x3): Likewise.
2710 (vst1q_f64_x3): Likewise.
2711 (vst1q_p64_x3): Likewise.
2713 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2715 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2716 user_label_prefix to name.
2718 * tree-core.h: Update comment about the format of NAME string
2719 passed to handler in attribute_spec.
2721 * config/msp430/msp430.md: Remove erroneous subreg expression from
2722 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2723 zero_extend{q,h}isi2.
2725 2018-05-30 Borislav Petkov <bp@suse.de>
2727 * doc/extend.texi: Document some architecture specific
2728 constraints and sort entries.
2730 2018-05-30 Martin Sebor <msebor@redhat.com>
2733 * builtins.c (expand_builtin_stpcpy_1): New function.
2734 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2735 only if the former succeeds.
2737 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2739 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2742 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2744 * doc/invoke.texi (-flinker-output): Document
2746 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2748 * passes.c (ipa_write_summaries): Only modify statements if body
2750 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2751 incrementally linking.
2752 (ipa_passes): Likewise.
2753 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2754 pass down resolution info.
2755 * common.opt (flag_incremental_link): Update info.
2756 * gcc.c (plugin specs): Turn flinker-output=* to
2757 -plugin-opt=-linker-output-known
2758 * toplev.c (compile_file): Also cut compilation when doing incremental
2760 * flag-types. (enum lto_partition_model): Add
2761 LTO_LINKER_OUTPUT_NOLTOREL.
2762 (invoke.texi): Add -flinker-output docs.
2763 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2764 link same way as WPA; do not stream in dead initializers.
2766 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2767 darf2out_register_external_decl): Support incremental link.
2769 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2771 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2773 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2775 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2776 it down to simple_object_copy_lto_debug_sections.
2777 (run_gcc): Determine incremental LTO link time and configure
2778 lto1 into non-wpa mode, disable renaming of debug sections.
2780 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2782 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2783 descriptions of various incorrectly documented functions.
2785 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2788 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2789 address check not strict.
2791 2018-05-30 Richard Biener <rguenther@suse.de>
2793 PR tree-optimization/85964
2794 * tracer.c (better_p): Drop initialized count check, we only
2795 call the function with initialized counts now.
2796 (find_best_successor): Do find a best edge if one
2797 has uninitialized count.
2798 (find_best_predecessor): Likewise. Do BB frequency check only
2799 if count is initialized.
2801 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2803 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2804 (aarch64_ldrstr_offset_compare): New.
2805 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2806 load/store orderings.
2807 (aarch64_gen_adjusted_ldpstp): Likewise.
2809 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2811 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2812 Check for subset of GENERAL_REGS and FP_REGS.
2813 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2816 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2818 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2819 and wi::to_poly_offset. Add the current offset and then check
2820 whether the sum fits, rather than using an unchecked addition of
2821 a checked term. Check for a shwi rather than a uhwi.
2822 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2823 (store_constructor): Use poly_int_tree_p.
2824 (expand_expr_real_1): Likewise.
2825 * function.c (assign_temp): Likewise.
2826 * fold-const.c (const_binop): Use poly_int_tree_p and
2828 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2830 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2831 to_poly_offset for MEM offsets.
2832 * ipa-icf.c (sem_variable::equals): Likewise.
2833 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2834 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2835 wi::to_poly_offset for BIT_FIELD_REF offsets.
2836 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2838 * var-tracking.c (emit_note_insn_var_location): Use
2839 tree_to_poly_uint64.
2841 2018-05-29 Jim Wilson <jimw@sifive.com>
2843 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2845 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2848 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2849 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2851 (sse4_1_round<mode>2): Use nonimmediate_operand
2852 for operand 1 predicate.
2854 2018-05-29 Martin Sebor <msebor@redhat.com>
2855 Richard Biener <rguenther@suse.de>
2858 * calls.c (get_size_range): Call determine_value_range instead
2859 of get_value_range..
2860 * tree-vrp.h (determine_value_range): Declared new function.
2861 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2863 2018-05-29 Richard Biener <rguenther@suse.de>
2865 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2866 sure to use non-pattern stmts for get_earlier_stmt arguments.
2867 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2868 called on pattern stmts.
2869 (get_later_stmt): Likewise.
2871 2018-05-29 Martin Liska <mliska@suse.cz>
2873 PR gcov-profile/85759
2874 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2877 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2879 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2881 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2885 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2886 VEC_PACK_FLOAT_EXPR): New tree codes.
2887 * tree-pretty-print.c (op_code_prio): Handle
2888 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2889 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2890 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2891 * tree-inline.c (estimate_operator_cost): Likewise.
2892 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2893 * fold-const.c (const_binop): Likewise.
2894 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2895 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2896 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2897 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2898 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2899 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2900 * expr.c (expand_expr_real_2): Likewise.
2901 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2902 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2903 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2905 * optabs.c (expand_widen_pattern_expr): For
2906 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2907 sign from result type rather than operand's type.
2908 (expand_binop_directly): For vec_packu_float_optab and
2909 vec_packs_float_optab allow result type to be different from operand's
2911 * optabs-tree.c (optab_for_tree_code): Handle
2912 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2913 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2914 * tree-vect-generic.c (expand_vector_operations_1): Handle
2915 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2916 VEC_PACK_FLOAT_EXPR.
2917 * tree-vect-stmts.c (supportable_widening_operation): Handle
2919 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2920 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2921 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2922 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2923 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2925 (vec_pack<floatprefix>_float_<mode>): New expander.
2926 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2928 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2929 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2930 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2931 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2932 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2934 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2935 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2936 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2937 VEC_PACK_FLOAT_EXPR): Document.
2939 2018-05-29 Richard Biener <rguenther@suse.de>
2941 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2943 (stmt_vec_info_vec): Make pointer.
2944 (init_stmt_vec_info_vec): Remove.
2945 (free_stmt_vec_info_vec): Likewise.
2946 (set_stmt_vec_info_vec): New function.
2947 (free_stmt_vec_infos): Likewise.
2948 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2949 (set_vinfo_for_stmt): Likewise.
2950 (get_earlier_stmt): Likewise.
2951 (get_later_stmt): Likewise.
2952 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2953 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2954 (vec_info::~vec_info): Free stmt_vec_infos.
2955 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2956 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2957 (pass_slp_vectorize::execute): Likewise.
2958 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2959 (free_stmt_vec_info_vec): Likewise.
2960 (set_stmt_vec_info_vec): New function.
2961 (free_stmt_vec_infos): Likewise.
2962 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2963 the global stmt_vec_info_vec.
2964 * tree-parloops.c (gather_scalar_reductions): Use
2965 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2968 2018-05-29 Richard Biener <rguenther@suse.de>
2970 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2972 2018-05-29 Martin Liska <mliska@suse.cz>
2973 David Malcolm <dmalcolm@redhat.com>
2975 * vec.c (test_reverse): New.
2976 (vec_c_tests): Add new test.
2977 * vec.h (vl_ptr>::reverse): New function.
2979 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2981 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2983 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2986 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2988 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2990 2018-05-28 Richard Biener <rguenther@suse.de>
2992 PR tree-optimization/85933
2993 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2994 look at stmts marked as vectorizable.
2996 2018-05-28 Richard Biener <rguenther@suse.de>
2998 PR tree-optimization/85934
2999 * tree-vect-generic.c (expand_vector_operations_1): Hoist
3000 vector boolean check before scalar optimization.
3002 2018-05-28 Jakub Jelinek <jakub@redhat.com>
3004 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
3007 2018-05-28 Mark Wielaard <mark@klomp.org>
3009 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
3010 if it is an expression containing a minus sign.
3012 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
3014 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
3016 2018-05-27 Paul Koning <ni1d@arrl.net>
3018 * config/pdp11/pdp11.md (truncsihi2): Remove.
3020 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
3021 Chung-Ju Wu <jasonwucj@gmail.com>
3023 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
3025 (unaligned_store_dw): Ditto.
3026 * config/nds32/nds32-memory-manipulation.c
3027 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
3028 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
3029 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
3030 (emit_setmem_word_loop): Rename to ...
3031 (emit_setmem_doubleword_loop): ... this.
3032 (nds32_gen_dup_4_byte_to_word_value): New function.
3033 (nds32_gen_dup_8_byte_to_double_word_value): New function.
3034 (nds32_expand_setmem_loop): Refine implementation.
3035 (nds32_expand_setmem_loop_v3m): Ditto.
3036 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
3039 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3041 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
3043 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
3045 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
3046 (nds32_init_machine_status): Initialize machine->attr_naked_p and
3047 machine->attr_no_prologue_p.
3048 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
3049 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
3050 (nds32_expand_epilogue): Consider attr_naked_p.
3051 (nds32_expand_epilogue_v3pop): Likewise.
3052 (nds32_can_use_return_insn): Likewise.
3053 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
3054 attr_no_prologue_p fields.
3055 * config/nds32/nds32.opt (mret-in-naked-func): New option.
3057 2018-05-27 Jakub Jelinek <jakub@redhat.com>
3060 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
3062 * config/i386/sse.md
3063 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
3065 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
3067 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
3069 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
3071 (*<floatsuffix>floatv2div2sf2): Rename to ...
3072 (*float<floatunssuffix>v2div2sf2): ... this.
3073 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
3074 (float<floatunssuffix>v2div2sf2_mask): ... this.
3075 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
3076 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
3077 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
3079 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
3081 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3083 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
3085 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3087 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
3089 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
3090 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
3091 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
3092 gen_ufix_truncv8dfv8si2.
3093 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
3094 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
3095 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
3096 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
3097 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
3098 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
3099 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
3100 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
3102 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3106 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
3108 2018-05-25 Jim Wilson <jimw@sifive.com>
3110 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
3111 * config/riscv/riscv.c (struct machine_function): Add
3112 interrupt_handler_p and attribute_checked_p fields.
3113 (riscv_attribute_table): Add interrupt.
3114 (riscv_interrupt_type_p): New.
3115 (riscv_save_reg_p): Save extra regs for interrupt handler.
3116 (riscv_use_save_libcall): Return false for interrupt handler.
3117 (riscv_first_stack_step): Add forward declaration.
3118 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
3119 for interrupt handler with large frame. Use it for saved reg list.
3120 (riscv_expand_prologue): Move flag_stack_usage_info support to
3121 eliminate duplication.
3122 (riscv_expand_epilogue): Generate mret for interrupt handler.
3123 (riscv_epilogue_uses): New.
3124 (riscv_can_use_return_insn): Return false for interrupt handler.
3125 (riscv_function_ok_for_sibcall): Likewise.
3126 (riscv_set_current_function): Add interrupt handler support.
3127 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
3128 * config/riscv/riscv.md (UNSPECV_MRET): New.
3130 (riscv_frflags, riscv_fsflags): Use tab after opcode.
3132 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
3134 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
3136 PR tree-optimization/85712
3137 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
3138 this candidate has already been replaced in-situ by a copy.
3140 2018-05-25 Jason Merrill <jason@redhat.com>
3142 PR c++/80485 - inline function non-zero address.
3143 * symtab.c (nonzero_address): Check DECL_COMDAT.
3145 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
3148 * config/alpha/alpha.md (ashlsi3): New insn pattern.
3149 (*ashlsi_se): Rename from *ashldi_se. Define as sign
3150 extension of SImode operation. Use const123_operand predicate.
3151 (*saddsi_1): Remove.
3152 (*saddl_se_1): Ditto.
3154 (*ssubl_se_1): Ditto.
3155 * config/alpha/predicates.md (const123_operand): New predicate.
3156 * config/alpha/constraints.md (P): Use IN_RANGE.
3158 2018-05-25 Richard Biener <rguenther@suse.de>
3160 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
3162 (ref_maybe_used_by_stmt_p): Likewise.
3163 (stmt_may_clobber_ref_p): Likewise.
3164 (stmt_may_clobber_ref_p_1): Likewise.
3165 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
3167 (ref_maybe_used_by_stmt_p): Likewise.
3168 (stmt_may_clobber_ref_p): Likewise.
3169 (stmt_may_clobber_ref_p_1): Likewise.
3170 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
3171 the alias oracle to disambiguate DRs with stmts DR analysis
3173 (vect_analyze_data_refs): Do not give up on not analyzable
3174 DRs for BB vectorization. Remove code truncating the dataref
3177 2018-05-25 Jakub Jelinek <jakub@redhat.com>
3180 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
3181 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
3182 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
3184 2018-05-25 Richard Biener <rguenther@suse.de>
3186 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
3187 function, combining stmt data ref gathering and fatal analysis
3189 (vect_analyze_data_refs): Remove now redudnant code and simplify.
3190 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
3191 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
3192 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
3193 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
3195 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3197 PR tree-optimization/85720
3198 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
3199 SCC if all partitions are builtins.
3200 (version_loop_by_alias_check): New parameter. Generate cancelable
3201 runtime alias check if all partitions are builtins.
3202 (distribute_loop): Update call to above function.
3204 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3206 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
3207 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
3208 (parm_default_def_partition_arg): Ditto.
3209 (set_parm_default_def_partition): Ditto.
3210 (get_parm_default_def_partitions): Ditto and make it static.
3211 (get_undefined_value_partitions): Ditto and make it static.
3212 (remove_ssa_form): Refactor call to init_var_map here.
3213 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
3214 computation for loop region.
3215 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
3216 (register_default_def): Delete.
3217 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
3218 (parm_default_def_partition_arg): Ditto.
3219 (set_parm_default_def_partition): Ditto.
3220 (get_parm_default_def_partitions): Ditto and make it static.
3221 (get_undefined_value_partitions): Ditto and make it static.
3222 (coalesce_with_default, coalesce_with_default): Update comment.
3223 (create_coalesce_list_for_region): New func factored out from
3224 create_outofssa_var_map.
3225 (populate_coalesce_list_for_outofssa): New func factored out from
3226 create_outofssa_var_map and coalesce_ssa_name.
3227 (create_outofssa_var_map): Delete.
3228 (coalesce_ssa_name): Refactor to support live range computation.
3229 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
3230 (get_parm_default_def_partitions): Delete.
3231 (get_undefined_value_partitions): Ditto.
3232 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
3233 computation for loop region.
3234 (new_tree_live_info, loe_visit_block): Ditto.
3235 (live_worklist, set_var_live_on_entry): Ditto.
3236 (calculate_live_on_exit, verify_live_on_entry): Ditto.
3237 * tree-ssa-live.h (struct _var_map): New fields.
3238 (init_var_map): Change decl.
3239 (region_contains_p): New.
3241 2018-05-25 Bin Cheng <bin.cheng@arm.com>
3243 * tree-ssa-live.h (live_merge_and_clear): Delete.
3245 2018-05-25 Richard Biener <rguenther@suse.de>
3248 * tree-dump.c (dequeue_and_dump): Remove access to removed
3249 operand 2 of a SWITCH_EXPR.
3251 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3253 * doc/sourcebuild.texi (vect_double_cond_arith): Include
3254 multiplication and division.
3255 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
3256 (cond_udiv@var{m}, cond_umod@var{m}): Document.
3257 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
3258 (cond_udiv_optab, cond_umod_optab): New optabs.
3259 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
3260 (IFN_COND_RDIV): New internal functions.
3261 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
3262 TRUNC_MOD_EXPR and RDIV_EXPR.
3263 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
3264 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
3266 (SVE_INT_BINARY): Include mult.
3267 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
3268 (optab, sve_int_op): Handle mult.
3269 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
3271 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
3272 for SVE_INT_BINARY_SD.
3274 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3276 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
3277 (optab, sve_int_op): Handle div and udiv.
3278 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
3279 for SVE_INT_BINARY_SD.
3280 (*<optab><mode>3): New insn for the same.
3282 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3284 * tree-vect-patterns.c: Include predict.h.
3285 (vect_recog_divmod_pattern): Restrict check for division support
3286 to when optimizing for size.
3288 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3290 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
3291 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
3292 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
3293 (gimple_match_op::set_op): Likewise.
3294 (gimple_resimplify4): Declare.
3295 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
3296 (expr::gen_transform): Likewise.
3297 (decision_tree::gen): Generate a simplification routine for 4 operands.
3298 * gimple-match-head.c (gimple_simplify): Add an overload for
3299 4 operands. In the top-level function, handle up to 4 call
3300 arguments and call gimple_resimplify4.
3301 (gimple_resimplify4): New function.
3302 (build_call_internal): Pass a fourth operand.
3303 (maybe_push_to_seq): Likewise.
3304 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
3305 Fold VEC_COND_EXPRs of an operation and a default value into
3306 an IFN_COND_* function if possible.
3307 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
3309 (SVE_COND_FP_BINARY): Include them.
3310 (optab, sve_fp_op): Handle them.
3311 (SVE_INT_BINARY_REV): New code iterator.
3312 (SVE_COND_FP_BINARY_REV): New int iterator.
3313 (commutative): New int attribute.
3314 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
3316 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
3318 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
3319 (*cond_<optab><mode>): New patterns for reversed operands.
3321 2018-05-25 Richard Biener <rguenther@suse.de>
3323 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
3324 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
3325 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
3326 (STMT_VINFO_GROUPED_ACCESS): Adjust.
3327 * tree-vect-data-refs.c (everywhere): Adjust users.
3328 * tree-vect-loop.c (everywhere): Likewise.
3329 * tree-vect-slp.c (everywhere): Likewise.
3330 * tree-vect-stmts.c (everywhere): Likewise.
3331 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3333 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3335 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
3337 (gcc_cv_as_section_exclude): ... this.
3338 Try Solaris as #exclude syntax.
3339 * configure: Regenerate.
3340 * config.in: Regenerate.
3341 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
3343 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
3344 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
3346 * varasm.c (default_elf_asm_named_section): Don't check if
3347 HAVE_GAS_SECTION_EXCLUDE is defined.
3349 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3351 * doc/md.texi: Update the documentation of the cond_* optabs
3352 to mention the new final operand. Fix GET_MODE_NUNITS call.
3353 Describe the scalar case too.
3354 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
3355 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
3357 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
3358 (get_conditional_internal_fn): Update comment.
3359 * tree-vect-loop.c (vectorizable_reduction): Pass the original
3360 accumulator value as a final argument to conditional functions.
3361 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
3362 a define_expand and add an "else" operand. Assert for now that
3363 the else operand is equal to operand 2. Use SVE_INT_BINARY and
3364 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
3365 (*cond_<optab><mode>): New patterns.
3366 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
3367 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3368 (UNSPEC_COND_EOR): Delete.
3369 (optab): Remove associated mappings.
3370 (SVE_INT_BINARY): New code iterator.
3371 (sve_int_op): Remove int attribute and add "minus" to the code
3373 (SVE_COND_INT_OP): Delete.
3374 (SVE_COND_FP_OP): Rename to...
3375 (SVE_COND_FP_BINARY): ...this.
3377 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
3379 * optabs.c (can_reuse_operands_p): New function.
3380 (maybe_legitimize_operands): Try to reuse the results for
3383 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3385 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
3386 Add {q} suffix to insn mnemonic.
3388 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3390 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
3391 (msp430_warn_func_return): New.
3393 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
3395 * fold-const.c (tree_nonzero_bits): New function.
3396 * fold-const.h (tree_nonzero_bits): Likewise.
3397 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
3398 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
3400 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
3404 * varasm.c (assemble_alias): Check ifunc_resolver only on
3407 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
3410 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
3411 when memory input operand is handled.
3413 2018-05-24 Luis Machado <luis.machado@linaro.org>
3415 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
3417 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
3419 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3421 * match.pd: Delay FMA folds until after vectorization.
3423 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
3426 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3427 address check not strict.
3429 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
3431 * gimple-match.h (gimple_match_op): New class.
3432 (mprts_hook): Replace parameters with a gimple_match_op *.
3433 (maybe_build_generic_op): Likewise.
3434 (gimple_simplified_result_is_gimple_val): Replace parameters with
3435 a const gimple_match_op *.
3436 (gimple_simplify): Replace code_helper * and tree * parameters with
3437 a gimple_match_op * parameter.
3438 (gimple_resimplify1): Replace code_helper *, tree and tree *
3439 parameters with a gimple_match_op * parameter.
3440 (gimple_resimplify2): Likewise.
3441 (gimple_resimplify3): Likewise.
3442 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
3443 parameters with a gimple_match_op * parameter.
3444 * gimple-match-head.c (gimple_simplify): Change prototypes of
3445 auto-generated functions to take a gimple_match_op * instead of
3446 separate code_helper * and tree * parameters. Make the same
3447 change in the top-level overload and update calls to the
3448 gimple_resimplify routines. Update calls to the auto-generated
3449 functions and to maybe_push_res_to_seq in the publicly-facing
3450 operation-specific gimple_simplify overloads.
3451 (gimple_match_op::MAX_NUM_OPS): Define.
3452 (gimple_resimplify1): Replace rcode and ops with a single res_op
3453 parameter. Update call to gimple_simplify.
3454 (gimple_resimplify2): Likewise.
3455 (gimple_resimplify3): Likewise.
3456 (mprts_hook): Replace parameters with a gimple_match_op *.
3457 (maybe_build_generic_op): Likewise.
3458 (build_call_internal): Replace type, nargs and ops with
3459 a gimple_match_op *.
3460 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
3461 with a single gimple_match_op *. Update calls to mprts_hook,
3462 build_call_internal and gimple_simplified_result_is_gimple_val.
3463 Factor out code that is common to the tree_code and combined_fn cases.
3464 * genmatch.c (expr::gen_transform): Replace tem_code and
3465 tem_ops with a gimple_match_op called tem_op. Update calls
3466 to the gimple_resimplify functions and maybe_push_res_to_seq.
3467 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
3468 res_ops. Update call to the gimple_resimplify functions.
3469 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
3470 (decision_tree::gen): Make the functions take a gimple_match_op *
3471 called res_op instead of separate res_code and res_ops parameters.
3472 Update call accordingly.
3473 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
3474 and ops with a single res_op parameter. Update calls to
3475 maybe_build_generic_op and maybe_push_res_to_seq.
3476 (fold_stmt_1): Update calls to gimple_simplify and
3477 replace_stmt_with_simplification.
3478 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
3479 and gimple_simplified_result_is_gimple_val.
3480 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
3482 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
3483 with a gimple_match_op *.
3484 (vn_nary_build_or_lookup): Likewise. Update call to
3485 vn_nary_build_or_lookup_1.
3486 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
3487 gimple_match_op *. Update calls to the gimple_resimplify routines
3488 and to gimple_simplified_result_is_gimple_val.
3489 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
3490 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
3491 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
3492 (visit_nary_op): Likewise.
3493 (visit_reference_op_load): Likewise.
3495 2018-05-23 Luis Machado <luis.machado@linaro.org>
3497 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
3498 modifier for printing the step amount.
3500 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
3503 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
3506 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
3508 * doc/sourcebuild.texi (Endianness): New subsubsection.
3510 2018-05-23 Luis Machado <luis.machado@linaro.org>
3512 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3513 <prefetch_dynamic_strides>: New const bool field.
3514 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3515 prefetch_dynamic_strides.
3516 (exynosm1_prefetch_tune): Likewise.
3517 (thunderxt88_prefetch_tune): Likewise.
3518 (thunderx_prefetch_tune): Likewise.
3519 (thunderx2t99_prefetch_tune): Likewise.
3520 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
3522 (aarch64_override_options_internal): Update to set
3523 PARAM_PREFETCH_DYNAMIC_STRIDES.
3524 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3525 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3526 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3527 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3528 prefetch-dynamic-strides setting.
3530 2018-05-23 Luis Machado <luis.machado@linaro.org>
3532 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3533 <minimum_stride>: New const int field.
3534 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3535 minimum_stride field defaulting to -1.
3536 (exynosm1_prefetch_tune): Likewise.
3537 (thunderxt88_prefetch_tune): Likewise.
3538 (thunderx_prefetch_tune): Likewise.
3539 (thunderx2t99_prefetch_tune): Likewise.
3540 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3541 <default_opt_level>: Set to 3.
3542 (aarch64_override_options_internal): Update to set
3543 PARAM_PREFETCH_MINIMUM_STRIDE.
3544 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3545 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3546 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3547 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3548 stride is constant and is below the minimum stride threshold.
3550 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3552 * config/arm/arm-cpus.in (mode26): Delete.
3553 (armv4): Delete mode26 reference.
3554 * config/arm/arm.c (arm_configure_build_target): Delete use of
3557 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3559 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3561 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3562 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3564 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3565 (floatunsdidf2): Ditto.
3567 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3569 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3570 (fixuns_trunc<mode>si2_avx512f): Ditto.
3571 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3572 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3573 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3575 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3577 PR rtl-optimization/79985
3578 * df-scan.c (df_insn_refs_collect): Remove special case for
3579 global registers and asm statements.
3581 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3583 * extend.texi (Global Register Variables): Rewrite the bullet list.
3584 Note that the register is available for allocation. Note that access
3585 via inline asm must use constraints. Add note about async-signal
3586 handlers. Remove paragraph about automagic register selection.
3588 2018-05-23 Richard Biener <rguenther@suse.de>
3590 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3591 of fixed offset from memset VN.
3593 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3595 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3597 (alloc_cand_and_find_basis): Initialize first_interp field.
3598 (slsr_process_mul): Modify first_interp field.
3599 (slsr_process_add): Likewise.
3600 (slsr_process_cast): Modify first_interp field for each new
3602 (slsr_process_copy): Likewise.
3603 (dump_candidate): Dump first_interp field.
3604 (replace_mult_candidate): Process all interpretations, not just
3606 (replace_rhs_if_not_dup): Likewise.
3607 (replace_one_candidate): Likewise.
3609 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3611 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3613 (aarch64_needs_frame_chain): New function.
3614 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3616 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3619 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3620 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3621 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3622 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3623 as true for strict-align.
3624 (aarch64_can_inline_p): Perform checks even when callee has no
3625 attributes to check for strict alignment.
3626 * doc/extend.texi (AArch64 Function Attributes): Document
3628 * doc/invoke.texi: (AArch64 Options): Likewise.
3630 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3632 PR tree-optimization/85853
3633 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3634 the handling of the root of the node to...
3635 (vect_slp_analyze_node_operations_1): ...this new function,
3636 and run the whole thing with the child nodes' def types
3637 set according to their SLP node's def type.
3639 2018-05-23 Richard Biener <rguenther@suse.de>
3642 * tree-data-ref.c (create_runtime_alias_checks): Defer
3643 and ignore overflow warnings.
3645 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3647 PR tree-optimization/85822
3648 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3651 2018-05-23 Richard Biener <rguenther@suse.de>
3653 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3654 memset constants via native_interpret_expr.
3656 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3659 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3661 (cgraph_node::create_alias): Likewise.
3662 (cgraph_node::get_availability): Check ifunc_resolver instead
3663 of looking up ifunc attribute.
3664 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3665 * varasm.c (do_assemble_alias): Likewise.
3666 (assemble_alias): Likewise.
3667 (default_binds_local_p_3): Likewise.
3668 * cgraph.h (cgraph_node): Add ifunc_resolver.
3669 (cgraph_node::only_called_directly_or_aliased_p): Return false
3671 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3673 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3674 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3675 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3676 instead of looking up ifunc attribute.
3678 2018-05-22 Luis Machado <luis.machado@linaro.org>
3680 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3682 2018-05-22 Martin Sebor <msebor@redhat.com>
3685 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3686 only when expasion succeeds.
3687 (expand_builtin_strcmp): Same.
3688 (expand_builtin_strncmp): Same.
3690 2018-05-22 Martin Sebor <msebor@redhat.com>
3692 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3694 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3695 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3697 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3698 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3700 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3701 Add check for legitimate_address.
3702 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3703 (aarch64_swap_ldrstr_operands): New.
3704 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3707 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3708 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3710 * config/aarch64/aarch64.md: New patterns to generate stp
3712 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3713 single words and double words.
3714 (load_pair_sw, load_pair_dw): Likewise.
3715 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3717 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3719 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3720 for different mode ldpstp and add peephole for merged zero stores.
3722 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3724 (aarch64_gen_store_pair): Rename calls to match new patterns.
3725 (aarch64_gen_load_pair): Rename calls to match new patterns.
3726 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3727 (load_pair<DREG:mode><DREG2:mode>): ... This.
3728 (store_pair<mode>): Rename to...
3729 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3730 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3732 (V_INT_EQUIV): Handle SImode.
3733 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3736 2018-05-22 Martin Sebor <msebor@redhat.com>
3739 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3740 or ajust the presumed bound on an operation to avoid unnecessary
3743 2018-05-22 Martin Sebor <msebor@redhat.com>
3745 PR tree-optimization/85826
3746 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3747 assuming that a DECL necesarily has a constant size.
3749 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3752 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3754 2018-05-22 Richard Biener <rguenther@suse.de>
3756 PR tree-optimization/85834
3757 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3758 non-constant and non-zero memset arguments.
3760 2018-05-22 Martin Liska <mliska@suse.cz>
3763 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3765 2018-05-22 Richard Biener <rguenther@suse.de>
3767 PR tree-optimization/85863
3768 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3769 comparisons when vectype is specified.
3770 (vectorizable_condition): Do not specify vectype for
3771 vect_is_simple_cond when SLP vectorizing.
3773 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3776 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3777 define __ibm128 as long double.
3778 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3779 as a distinct type when IEEE 128-bit support is enabled.
3780 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3781 128-bit types to use the correct functions.
3782 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3783 convert between 128-bit floating point types that have different
3784 modes but the same representation, instead of using gen_lowpart to
3786 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3788 (IFKF_reg): New attributes to give the register constraints for
3790 (extend<mode>tf2_internal): New insns to mark an explicit
3791 conversion between 128-bit floating point types that have a
3792 different mode but share the same representation.
3794 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3796 PR tree-optimization/85814
3797 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3798 a null return from get_strinfo when unsharing the next
3799 strinfo in the chain.
3801 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3804 * varasm.c (weak_finish): Clean up weak_decls.
3806 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3808 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3809 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3810 UNSPEC_UADALP values.
3811 * config/aarch64/iterators.md (ABAL): New int iterator.
3814 (sur): Add mappings for the above.
3815 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3817 (aarch64_<sur>abal<mode>_4): Likewise.
3818 (aarch64_<sur>adalp<mode>_3): Likewise.
3819 (<sur>sadv16qi): New define_expand.
3821 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3823 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3824 (*movdf_internal): Ditto.
3825 (*rcpsf2_sse): Ditto.
3826 (*rsqrtsf2_sse): Ditto.
3827 (*sqrt<mode>2_sse): Ditto.
3829 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3831 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3833 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3834 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3835 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3836 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3838 * config/aarch64/arm_neon.h: Likewise.
3839 * config/aarch64/iterators.md (VQ_I): New.
3841 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3843 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3845 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3848 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3850 * config/nds32/constraints.md (S): New constraint.
3851 * config/nds32/nds32.md (call_internal): Use constraint S.
3852 (call_value_internal): Likewise.
3853 (sibcall_internal): Likewise.
3854 (sibcall_value_internal): Likewise.
3856 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3857 Chung-Ju Wu <jasonwucj@gmail.com>
3859 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3862 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3863 Chung-Ju Wu <jasonwucj@gmail.com>
3865 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3866 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3867 (nds32_rtx_costs_impl): Simplify.
3868 (nds32_address_cost_impl): Simplify.
3869 (nds32_init_rtx_costs): New function.
3870 (nds32_rtx_costs_speed_prefer): Likewise.
3871 (nds32_rtx_costs_size_prefer): Likewise.
3872 (nds32_address_cost_speed_prefer): Likewise.
3873 (nds32_address_cost_speed_fwprop): Likewise.
3874 (nds32_address_cost_size_prefer): Likewise.
3875 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3876 * config/nds32/nds32.c (nds32_option_override): Use
3877 nds32_init_rtx_costs function.
3879 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3882 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3883 (TARGET_PIPELINE_N8): Likewise.
3884 (TARGET_PIPELINE_N10): Likewise.
3885 (TARGET_PIPELINE_N13): Likewise.
3886 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3888 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3890 * config/nds32/nds32-fpu.md: Update copyright year.
3892 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3894 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3896 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3898 * config/nds32/nds32.c
3899 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3900 * config/nds32/nds32.opt (minline-asm-r15): New option.
3902 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3904 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3906 * config/nds32/nds32.md (abssi2): New pattern.
3908 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3910 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3911 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3912 Merge insn pattern from sse_cvtsi2ss<round_name> and
3913 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3914 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3915 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3916 using SWI48 mode iterator.
3917 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3918 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3919 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3920 pattern from sse_cvttss2si<round_saeonly_name>
3921 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3922 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3923 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3924 using SWI48 mode iterator.
3925 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3926 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3927 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3928 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3929 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3930 using SWI48 mode iterator.
3931 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3932 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3933 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3934 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3935 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3936 SWI48 mode iterator.
3937 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3938 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3939 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3940 pattern from sse_cvttsd2si<round_saeonly_name>
3941 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3943 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3945 * config/nds32/nds32-md-auxiliary.c
3946 (nds32_valid_smw_lwm_base_p): Refine.
3947 (nds32_output_smw_single_word): Refine.
3948 (nds32_output_smw_double_word): New.
3949 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3951 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3953 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3954 (nds32_output_stack_pop): Refine.
3955 (nds32_expand_unaligned_load): Refine.
3956 (nds32_expand_unaligned_store): Refine.
3958 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3959 Chung-Ju Wu <jasonwucj@gmail.com>
3961 * config/nds32/constants.md: Add TP_REGNUM constant.
3962 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3963 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3965 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3966 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3967 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3968 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3969 and PIC code generation.
3970 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3972 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3974 * config/nds32/nds32.md: Support TLS and PIC.
3975 * config/nds32/nds32.c: Support TLS and PIC.
3976 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3977 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3980 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3982 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3983 mode with E_ prefix.
3985 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3986 Chung-Ju Wu <jasonwucj@gmail.com>
3988 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3989 * config/nds32/nds32-md-auxiliary.c
3990 (symbolic_reference_mentioned_p): New.
3991 (nds32_legitimize_ict_address): New.
3992 (nds32_expand_ict_move): New.
3993 (nds32_indirect_call_referenced_p): New.
3994 (nds32_symbol_binds_local_p): Delete.
3995 (nds32_long_call_p): Modify.
3996 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3997 * config/nds32/nds32-protos.h
3998 (symbolic_reference_mentioned_p): Declare.
3999 (nds32_legitimize_ict_address): Declare.
4000 (nds32_expand_ict_move): Declare.
4001 (nds32_indirect_call_referenced_p): Declare.
4002 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
4003 (nds32_relax_group): Use nds32_ict_const_p as condition.
4004 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
4005 (nds32_asm_file_start): Output ict_model directive in asm code.
4006 (nds32_legitimate_address_p): Consider indirect call.
4007 (nds32_print_operand): Consider indirect call.
4008 (nds32_print_operand_address): Consider indirect call.
4009 (nds32_insert_attributes): Handle "indirect_call" attribute.
4010 (TARGET_LEGITIMATE_ADDRESS_P): Define.
4011 (TARGET_LEGITIMATE_CONSTANT_P): Define.
4012 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
4013 (TARGET_DELEGITIMIZE_ADDRESS): Define.
4014 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
4015 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
4016 (TARGET_ICT_MODEL_SMALL): Define.
4017 (TARGET_ICT_MODEL_LARGE): Define.
4018 * config/nds32/nds32.md (movsi): Consider ict model.
4019 (call, call_value): Consider ict model.
4020 (sibcall, sibcall_value): Consider ict model.
4021 * config/nds32/nds32.opt (mict-model): New option.
4022 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
4025 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
4026 Monk Chiang <sh.chiang04@gmail.com>
4027 Jim Wilson <jimw@sifive.com>
4029 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
4030 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
4031 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
4032 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
4033 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
4034 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
4035 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
4036 compute save_libcall_adjustment properly.
4037 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
4038 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
4039 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
4040 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
4041 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
4042 (ABI_SPEC): Handle mabi=ilp32e.
4043 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
4044 (RVE): Add RVE mask.
4045 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
4046 <-march>: Add rv32e as an example.
4048 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
4051 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
4052 (intra_create_variable_infos): Handle C++ constructors.
4054 2018-05-18 Martin Liska <mliska@suse.cz>
4056 * passes.def: Remove a redundant pass.
4058 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4061 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
4063 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4065 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
4067 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
4069 (armv2, armv2a, armv3, armv3m): Delete architectures.
4070 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
4071 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
4072 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
4074 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
4075 (*mulsidi3adddi): Likewise.
4076 (mulsidi3): Likewise.
4077 (*mulsidi3_nov6): Likewise.
4078 (umulsidi3): Likewise.
4079 (umulsidi3_nov6): Likewise.
4080 (umaddsidi4): Likewise.
4081 (*umulsidi3adddi): Likewise.
4082 (smulsi3_highpart): Likewise.
4083 (*smulsi3_highpart_nov6): Likewise.
4084 (umulsi3_highpart): Likewise.
4085 (*umulsi3_highpart_nov6): Likewise.
4086 * config/arm/arm.h (arm_arch3m): Delete.
4087 * config/arm/arm.c (arm_arch3m): Delete.
4088 (arm_option_override_internal): Update armv3-related comment.
4089 (arm_configure_build_target): Delete use of isa_bit_mode32.
4090 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
4091 (arm_rtx_costs_internal): Delete check of arm_arch3m.
4092 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
4094 (mulusa3): Likewise.
4095 * config/arm/arm-protos.h (arm_arch3m): Delete.
4096 * config/arm/arm-tables.opt: Regenerate.
4097 * config/arm/arm-tune.md: Likewise.
4098 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
4099 deleted architectures.
4101 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4103 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
4104 (armv5t, armv5te): New features.
4105 (ARMv5, ARMv5e): Delete fgroups.
4106 (ARMv5t, ARMv5te): Adjust for above changes.
4108 (armv5, armv5e): Delete arches.
4109 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
4111 (*call_reg_arm): Likewise.
4112 (*call_value_reg_armv5): Likewise.
4113 (*call_value_reg_arm): Likewise.
4114 (*call_symbol): Likewise.
4115 (*call_value_symbol): Likewise.
4116 (*sibcall_insn): Likewise.
4117 (*sibcall_value_insn): Likewise.
4119 (prefetch): Likewise.
4120 (define_split and define_peephole2 dependent on arm_arch5):
4122 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
4124 (TARGET_ARM_QBIT): Likewise.
4125 (TARGET_DSP_MULTIPLY): Likewise.
4126 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
4127 (arm_arch5, arm_arch5e): Delete.
4128 (arm_arch5t, arm_arch5te): Declare.
4129 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
4130 (arm_arch5t): Declare.
4131 (arm_option_reconfigure_globals): Update for the above.
4132 (arm_options_perform_arch_sanity_checks): Update comment, replace
4133 use of arm_arch5 with arm_arch5t.
4134 (use_return_insn): Likewise.
4135 (arm_emit_call_insn): Likewise.
4136 (output_return_instruction): Likewise.
4137 (arm_final_prescan_insn): Likewise.
4138 (arm_coproc_builtin_available): Likewise.
4139 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
4140 arm_arch5e with arm_arch5t and arm_arch5te.
4141 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
4142 (arm_arch5t, arm_arch5te): Declare.
4143 * config/arm/arm-tables.opt: Regenerate.
4144 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
4145 * config/arm/t-multilib: Likewise.
4146 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
4147 instead of arm_arch5.
4148 (*call_reg_thumb1): Likewise.
4149 (*call_value_reg_thumb1_v5): Likewise.
4150 (*call_value_reg_thumb1): Likewise.
4151 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
4153 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
4155 2018-05-18 Martin Liska <mliska@suse.cz>
4157 PR gcov-profile/84846
4158 * doc/gcov.texi: Document -t option of gcov tool.
4160 2018-05-18 Martin Liska <mliska@suse.cz>
4162 PR gcov-profile/84846
4163 * gcov.c (print_usage): Add new -t option.
4164 (process_args): Handle the option.
4165 (generate_results): Use stdout as output when requested by
4168 2018-05-18 Martin Liska <mliska@suse.cz>
4170 PR gcov-profile/84846
4171 * coverage.c (coverage_init): Write PWD to .gcno file.
4172 * doc/gcov.texi: Document how working directory is printed.
4173 * gcov-dump.c (dump_gcov_file): Print PWD.
4174 * gcov.c (output_intermediate_file): Likewise.
4175 (read_graph_file): Read PWD string.
4176 (output_lines): Print PWD.
4178 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4181 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
4182 for retval and return false if all args to phi are zero.
4184 2018-05-18 Richard Biener <rguenther@suse.de>
4186 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
4188 (evrp_dom_walker::before_dom_children): Call it.
4190 2018-05-18 Richard Biener <rguenther@suse.de>
4192 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4193 results when processing array refs with variable index.
4195 2018-05-18 Toon Moene <toon@moene.org>
4197 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
4198 directly after that of -floop-interchange. Indicate that both
4199 options are enabled by default when specifying -O3.
4201 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4203 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
4204 iterator. Delete separate integer-mode vec_set<mode> expander.
4205 (aarch64_simd_vec_setv2di): Delete.
4206 (vec_setv2di): Delete.
4207 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
4208 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
4209 the "w, r" alternative.
4211 2018-05-18 Martin Liska <mliska@suse.cz>
4213 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
4214 * tree-pass.h (make_pass_lower_switch_O0): New function.
4215 * tree-switch-conversion.c (node_has_low_bound): Remove.
4216 (node_has_high_bound): Likewise.
4217 (node_is_bounded): Likewise.
4218 (class pass_lower_switch): Make it a template type and create
4220 (pass_lower_switch::execute): Add template argument.
4221 (make_pass_lower_switch): New function.
4222 (make_pass_lower_switch_O0): New function.
4223 (do_jump_if_equal): Remove.
4224 (emit_case_nodes): Simplify to just handle all 3 cases and leave
4225 all the hard work to tree optimization passes.
4227 2018-05-18 Martin Liska <mliska@suse.cz>
4229 * dbgcnt.c (limit_low): Renamed from limit.
4230 (limit_high): New variable.
4231 (dbg_cnt_is_enabled): Check for upper limit.
4232 (dbg_cnt): Adjust dumping.
4233 (dbg_cnt_set_limit_by_index): Add new argument for high
4235 (dbg_cnt_set_limit_by_name): Likewise.
4236 (dbg_cnt_process_single_pair): Parse new format.
4237 (dbg_cnt_process_opt): Use strtok.
4238 (dbg_cnt_list_all_counters): Remove 'value' and add
4240 * doc/invoke.texi: Document changes.
4242 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
4244 * doc/sourcebuild.texi (scalar_all_fma): Document.
4245 * tree.def (FMA_EXPR): Delete.
4246 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
4247 * internal-fn.c (ternary_direct): New macro.
4248 (expand_ternary_optab_fn): Likewise.
4249 (direct_ternary_optab_supported_p): Likewise.
4250 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
4251 * builtins.c (fold_builtin_fma): Delete.
4252 (fold_builtin_3): Don't call it.
4253 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
4254 * expr.c (expand_expr_real_2): Likewise.
4255 * fold-const.c (operand_equal_p): Likewise.
4256 (fold_ternary_loc): Likewise.
4257 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
4258 * gimple.c (DEFTREECODE): Likewise.
4259 * gimplify.c (gimplify_expr): Likewise.
4260 * optabs-tree.c (optab_for_tree_code): Likewise.
4261 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
4262 * tree-eh.c (operation_could_trap_p): Likewise.
4263 (stmt_could_throw_1_p): Likewise.
4264 * tree-inline.c (estimate_operator_cost): Likewise.
4265 * tree-pretty-print.c (dump_generic_node): Likewise.
4266 (op_code_prio): Likewise.
4267 * tree-ssa-loop-im.c (stmt_cost): Likewise.
4268 * tree-ssa-operands.c (get_expr_operands): Likewise.
4269 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
4270 * fold-const-call.h (fold_fma): Delete.
4271 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
4272 CFN_FNMA and CFN_FNMS.
4274 * genmatch.c (combined_fn): New enum.
4275 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
4276 (commutative_op): New function.
4277 (commutate): Use it. Handle more than 2 operands.
4278 (dt_operand::gen_gimple_expr): Use commutative_op.
4279 (parser::parse_expr): Allow :c to be used with non-binary
4280 operators if the commutative operand is known.
4281 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
4282 CFN_FMS, CFN_FNMA and CFN_FNMS.
4283 (backprop::process_assign_use): Remove FMA_EXPR handling.
4284 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
4285 (gen_hsa_fma): New function.
4286 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
4287 IFN_FNMA and IFN_FNMS.
4288 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
4289 * gimple-fold.h (follow_all_ssa_edges): Declare.
4290 * gimple-fold.c (follow_all_ssa_edges): New function.
4291 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
4292 gimple_build interface and use follow_all_ssa_edges to fold the result.
4293 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
4294 instead of checking for optabs directly.
4295 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
4296 rather than FMA_EXPRs.
4297 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
4298 call to IFN_FMA instead of an FMA_EXPR.
4300 2018-05-17 Jim Wilson <jimw@sifive.com>
4302 * expr.c (do_tablejump): When converting index to Pmode, if we have a
4303 sign extended promoted subreg, and the range does not have the sign bit
4304 set, then do a sign extend.
4306 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
4307 test, check for sign extended subreg and/or constant operands, and
4308 do a sign extend in that case.
4310 2018-05-17 Steve Ellcey <sellcey@cavium.com>
4312 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
4313 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
4315 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
4316 Change logics_shift_reg to logics_shift_imm.
4317 (thunderx2t99_fp_loadpair_basic): Delete.
4318 (thunderx2t99_fp_storepair_basic): Delete.
4319 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
4320 (thunderx2t99_asimd_polynomial): Delete.
4321 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
4322 and neon_fp_mul_d_scalar_q.
4323 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
4324 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
4325 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
4326 (thunderx2t99_asimd_lut): Add missing tbl types.
4327 (thunderx2t99_asimd_ext): Delete.
4328 (thunderx2t99_asimd_load1_1_mult): Delete.
4329 (thunderx2t99_asimd_load1_2_mult): Delete.
4330 (thunderx2t99_asimd_load1_ldp): New.
4331 (thunderx2t99_asimd_load1): New.
4332 (thunderx2t99_asimd_load2): Add missing *load2* types.
4333 (thunderx2t99_asimd_load3): New.
4334 (thunderx2t99_asimd_load4): New.
4335 (thunderx2t99_asimd_store1_1_mult): Delete.
4336 (thunderx2t99_asimd_store1_2_mult): Delete.
4337 (thunderx2t99_asimd_store2_mult): Delete.
4338 (thunderx2t99_asimd_store2_onelane): Delete.
4339 (thunderx2t99_asimd_store_stp): New.
4340 (thunderx2t99_asimd_store1): New.
4341 (thunderx2t99_asimd_store2): New.
4342 (thunderx2t99_asimd_store3): New.
4343 (thunderx2t99_asimd_store4): New.
4345 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
4347 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
4348 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
4350 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
4351 Segher Boessenkool <segher@kernel.crashing.org>
4354 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
4357 2018-05-17 Richard Biener <rguenther@suse.de>
4359 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4360 for pruning loop and prune defs feeding only already visited PHIs.
4362 2018-05-17 Richard Biener <rguenther@suse.de>
4364 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
4366 2018-05-17 Bin Cheng <bin.cheng@arm.com>
4367 Richard Biener <rguenther@suse.de>
4369 PR tree-optimization/85793
4370 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
4371 for VMAT_ELEMENTWISE.
4373 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4375 * internal-fn.h (lookup_internal_fn): Declare
4376 * internal-fn.c (lookup_internal_fn): New function.
4377 * gimple.c (gimple_build_call_from_tree): Handle calls to
4379 * gimple-pretty-print.c (dump_gimple_call): Print "." before
4380 internal function names.
4381 * tree-pretty-print.c (dump_generic_node): Likewise.
4382 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
4384 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
4386 * gimple-fold.h (gimple_build): Make the function forms take
4387 combined_fn rather than built_in_function.
4388 (gimple_simplify): Likewise.
4389 * gimple-match-head.c (gimple_simplify): Likewise.
4390 * gimple-fold.c (gimple_build): Likewise.
4391 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
4392 rather than gimple_build_call_internal.
4393 (get_initial_defs_for_reduction): Likewise.
4394 (vect_create_epilog_for_reduction): Likewise.
4395 (vectorizable_live_operation): Likewise.
4397 2018-05-17 Martin Liska <mliska@suse.cz>
4399 * gimple-ssa-sprintf.c (format_directive): Do not use
4400 space in between 'G_' and '('.
4402 2018-05-17 Jakub Jelinek <jakub@redhat.com>
4405 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
4406 even if the mask is not all ones.
4409 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
4411 (ix86_gimple_fold_builtin): Likewise.
4414 * config/i386/i386.c: Include tree-vector-builder.h.
4415 (ix86_vector_shift_count): New function.
4416 (ix86_fold_builtin): Fold shift builtins by scalar count.
4417 (ix86_gimple_fold_builtin): Likewise.
4419 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
4420 _mm512_setzero): New intrinsics.
4422 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
4423 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4425 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
4426 code generation for cases where splatting a value is not useful.
4427 * simplify-rtx.c (simplify_ternary_operation): Simplify
4428 vec_merge across a vec_duplicate and a paradoxical subreg forming
4429 a vector mode to a vec_concat.
4431 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
4433 * config.gcc: Support "goldmont-plus".
4434 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4436 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4437 PROCESSOR_GOLDMONT_PLUS.
4438 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
4439 (processor_target_table): Add "goldmont-plus".
4440 (PTA_GOLDMONT_PLUS): Define.
4441 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
4442 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
4443 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
4444 (fold_builtin_cpu): Add "goldmont-plus".
4445 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
4446 (ix86_option_override_internal): Add "goldmont-plus".
4447 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
4448 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
4449 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
4450 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
4452 2018-05-17 Richard Biener <rguenther@suse.de>
4454 PR tree-optimization/85757
4455 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
4456 remove defs that only feed that PHI from further processing.
4458 2018-05-16 Jim Wilson <jimw@sifive.com>
4460 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
4462 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
4464 2018-05-16 Mark Wielaard <mark@klomp.org>
4466 * dwarf2out.c (count_index_strings): New function.
4467 (output_indirect_strings): Call count_index_strings and generate
4468 header for dwarf_version >= 5.
4470 2018-05-16 Mark Wielaard <mark@klomp.org>
4472 * dwarf2out.c (dwarf_FORM): New function.
4473 (set_indirect_string): Use dwarf_FORM.
4474 (reset_indirect_string): Likewise.
4475 (size_of_die): Likewise.
4476 (value_format): Likewise.
4477 (output_die): Likewise.
4478 (add_skeleton_AT_string): Likewise.
4479 (output_macinfo_op): Likewise.
4480 (index_string): Likewise.
4481 (output_index_string_offset): Likewise.
4482 (output_index_string): Likewise.
4483 (count_index_strings): Likewise.
4485 2018-05-16 Carl Love <cel@us.ibm.com>
4487 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4488 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
4490 2018-05-16 Martin Jambor <mjambor@suse.cz>
4492 * ipa-prop.c (ipa_free_all_edge_args): Remove.
4493 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
4495 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
4497 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
4498 (fnma<mode>4): Likewise.
4499 (fms<mode>4): Likewise.
4500 (fnms<mode>4): Likewise.
4501 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
4502 (aarch64_fnma<mode>4): Likewise.
4503 (aarch64_fms<mode>4): Likewise.
4504 (aarch64_fnms<mode>4): Likewise.
4505 (aarch64_fnmadd<mode>4): Likewise.
4507 2018-05-16 Jason Merrill <jason@redhat.com>
4509 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
4511 2018-05-16 Richard Biener <rguenther@suse.de>
4513 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
4514 (dump_stmt_cost): Declare.
4515 (add_stmt_cost): Dump cost we add.
4516 (add_stmt_costs): New function.
4517 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
4519 (vect_analyze_stmt): Adjust prototype.
4520 (vectorizable_condition): Likewise.
4521 (vectorizable_live_operation): Likewise.
4522 (vectorizable_reduction): Likewise.
4523 (vectorizable_induction): Likewise.
4524 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4525 cost vector to pass to vectorizable_ and record afterwards.
4526 (vect_model_reduction_cost): Take cost vector argument and adjust.
4527 (vect_model_induction_cost): Likewise.
4528 (vectorizable_reduction): Likewise.
4529 (vectorizable_induction): Likewise.
4530 (vectorizable_live_operation): Likewise.
4531 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4532 SLP_TREE_NUMBER_OF_VEC_STMTS.
4533 (vect_analyze_slp_cost_1): Remove.
4534 (vect_analyze_slp_cost): Likewise.
4535 (vect_slp_analyze_node_operations): Take visited args and
4536 a target cost vector. Avoid processing already visited stmt sets.
4537 (vect_slp_analyze_operations): Use a local cost vector to gather
4538 costs and register those of non-discarded instances.
4539 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4540 (vect_schedule_slp_instance): Remove copying of
4541 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4543 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4544 adding cost. Record cost entry location.
4545 (vect_prologue_cost_for_slp_op): Function to compute cost of
4546 a constant or invariant generated for SLP vect in the prologue,
4547 split out from vect_analyze_slp_cost_1.
4548 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4549 (vect_model_promotion_demotion_cost): Likewise.
4550 (vect_model_store_cost): Likewise, make static.
4551 (vect_model_load_cost): Likewise.
4552 (vectorizable_bswap): Add cost vector arg and adjust.
4553 (vectorizable_call): Likewise.
4554 (vectorizable_simd_clone_call): Likewise.
4555 (vectorizable_conversion): Likewise.
4556 (vectorizable_assignment): Likewise.
4557 (vectorizable_shift): Likewise.
4558 (vectorizable_operation): Likewise.
4559 (vectorizable_store): Likewise.
4560 (vectorizable_load): Likewise.
4561 (vectorizable_condition): Likewise.
4562 (vectorizable_comparison): Likewise.
4563 (can_vectorize_live_stmts): Likewise.
4564 (vect_analyze_stmt): Likewise.
4565 (vect_transform_stmt): Adjust calls to vectorizable_*.
4566 * tree-vectorizer.c: Include gimple-pretty-print.h.
4567 (dump_stmt_cost): New function.
4569 2018-05-16 Richard Biener <rguenther@suse.de>
4571 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4572 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4573 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4574 (check_name): New callback.
4575 (dse_classify_store): Track cycles via a visited bitmap of PHI
4576 defs and simplify handling of in-loop and across loop dead stores
4577 and properly fail for loop-variant refs. Handle byte-tracking with
4578 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4581 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4583 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4584 (vect_get_mask_type_for_stmt): Likewise.
4585 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4587 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4588 to determine the statement's vector type and the vector type that
4589 should be used for calculating nunits. Deal with cases in which
4590 the type has to be deferred.
4591 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4592 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4593 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4594 (vect_determine_vf_for_stmt): New functions, split out from...
4595 (vect_determine_vectorization_factor): ...here.
4596 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4597 (vect_get_mask_type_for_stmt): New functions, split out from
4598 vect_determine_vectorization_factor.
4600 2018-05-16 Richard Biener <rguenther@suse.de>
4602 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4603 verify the [VEC_]COND_EXPR embedded comparison.
4605 2018-05-15 Martin Sebor <msebor@redhat.com>
4607 PR tree-optimization/85753
4608 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4609 RECORD_TYPE in addition to ARRAY_TYPE.
4611 2018-05-15 Martin Sebor <msebor@redhat.com>
4614 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4616 2018-05-15 Richard Biener <rguenther@suse.de>
4618 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4619 add by_clobber_p one. Change algorithm to collect all defs
4620 representing uses we need to walk and try reducing them to
4621 a single one before failing.
4622 (dse_dom_walker::dse_optimize_stmt): Adjust.
4624 2018-05-13 Mark Wielaard <mark@klomp.org>
4626 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4627 (size_of_loc_descr): Likewise.
4628 (output_loc_operands): Likewise.
4629 (output_loc_operands_raw): Likewise.
4630 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4631 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4632 (hash_loc_operands): Likewise.
4633 (compare_loc_operands): Likewise.
4635 2018-05-14 Mark Wielaard <mark@klomp.org>
4637 * dwarf2out.c (count_index_addrs): New function.
4638 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4640 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4642 PR tree-optimization/83648
4643 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4644 return value as malloc candidate.
4646 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4649 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4650 param as true in call to suggest_attribute.
4652 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4654 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4655 -mreadonly-in-sdata.
4657 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4659 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4661 (aarch64_crypto_aesd_fused): Likewise.
4663 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4665 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4666 (movsi_aarch64): Likewise.
4667 (load_pairsi): Likewise.
4668 (load_pairdi): Likewise.
4669 (store_pairsi): Likewise.
4670 (store_pairdi): Likewise.
4671 (load_pairsf): Likewise.
4672 (load_pairdf): Likewise.
4673 (store_pairsf): Likewise.
4674 (store_pairdf): Likewise.
4675 (zero_extend): Likewise.
4676 (trunc): Swap alternatives.
4677 (fcvt_target): Add '?' to prefer w over r.
4679 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4682 * config/i386/i386.md: Disallow non-commutative arithmetics in
4683 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4684 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4685 in the peephole2 before it.
4687 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4689 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4690 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4691 (ix86_handle_option): Handle -mcldemote.
4692 * config.gcc: New header.
4693 * config/i386/cldemoteintrin.h: New file.
4694 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4695 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4697 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4698 OPTION_MASK_ISA_CLDEMOTE.
4699 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4700 (ix86_valid_target_attribute_inner_p): Ditto.
4701 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4702 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4703 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4704 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4705 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4707 * config/i386/i386.opt: Add -mcldemote.
4708 * config/i386/x86intrin.h: New header.
4709 * doc/invoke.texi: Add -mcldemote.
4711 2018-05-14 Richard Biener <rguenther@suse.de>
4713 * doc/match-and-simplify.texi: Adjust :s documentation.
4715 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4717 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4718 intended memcpy size.
4719 (REORDER_45): Likewise.
4721 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4723 * sort.cc: New file.
4724 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4725 * vec.c (qsort_chk): Use gcc_qsort.
4726 * Makefile.in (OBJS-libcommon): Add sort.o.
4727 (build/sort.o): New target. Use it...
4728 (BUILD_RTL): ... here, and...
4729 (build/gencfn-macros): ... here, and...
4730 (build/genmatch): ... here.
4732 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4733 Chung-Ju Wu <jasonwucj@gmail.com>
4735 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4736 * config/nds32/nds32-graywolf.md: New file.
4737 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4738 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4740 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4741 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4742 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4743 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4744 * config/nds32/pipelines.md: Include n15 settings.
4746 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4747 Chung-Ju Wu <jasonwucj@gmail.com>
4749 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4750 * config/nds32/nds32-n13.md: New file.
4751 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4752 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4754 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4755 * config/nds32/nds32.md (pipeline_model): Add n13.
4756 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4757 * config/nds32/pipelines.md: Include n13 settings.
4759 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4760 Chung-Ju Wu <jasonwucj@gmail.com>
4762 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4763 * config/nds32/nds32-n10.md: New file.
4764 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4765 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4767 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4768 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4769 * config/nds32/nds32.md (pipeline_model): Add n10.
4770 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4771 * config/nds32/pipelines.md: Include n10 settings.
4773 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4774 Kito Cheng <kito.cheng@gmail.com>
4775 Chung-Ju Wu <jasonwucj@gmail.com>
4777 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4778 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4779 Add enum values for DSP extension instructions.
4780 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4782 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4783 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4785 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4786 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4787 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4788 * config/nds32/nds32-intrinsic.md: Likewise.
4789 * config/nds32/nds32_intrinsic.h: Likewise.
4790 * config/nds32/nds32-md-auxiliary.c: Likewise.
4791 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4792 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4793 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4794 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4795 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4796 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4797 TYPE_DMAC in switch statement.
4798 * config/nds32/nds32.c: New checking and implementation for DSP
4799 extension instructions.
4800 * config/nds32/nds32.h: Likewise.
4801 * config/nds32/nds32.md: Likewise.
4802 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4803 * config/nds32/predicates.md: Implement new predicates for DSP
4806 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4808 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4809 Reformat alternatives and attributes so it is easier to identify
4810 which constraints/attributes go with which instruction.
4811 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4812 (mov<mode>_softfloat32, FMOVE64): Likewise.
4813 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4814 (mov<mode>_softfloat64, FMOVE64): Likewise.
4816 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4818 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4820 (Basic PowerPC Built-in Functions): The new name of the
4821 subsection previously known as "PowerPC Built-in Functions".
4822 (Basic PowerPC Built-in Functions Available on all Configurations):
4824 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4825 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4826 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4827 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4829 2018-05-11 Martin Jambor <mjambor@suse.cz>
4832 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4835 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4838 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4840 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4842 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4843 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4844 (ix86_handle_option): Handle -mwaitpkg.
4845 * config.gcc: New header.
4846 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4847 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4848 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4850 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4851 OPTION_MASK_ISA_WAITPKG.
4852 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4853 (ix86_option_override_internal): Add PTA_WAITPKG.
4854 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4855 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4856 IX86_BUILTIN_TPAUSE.
4857 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4858 __builtin_ia32_umwait and __builtin_ia32_tpause.
4859 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4860 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4861 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4862 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4863 UNSPECV_TPAUSE): New.
4864 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4865 * config/i386/i386.opt: Add -mwaitpkg.
4866 * config/i386/waitpkgintrin.h: New file.
4867 * config/i386/x86intrin.h: New header.
4868 * doc/invoke.texi: Add -mwaitpkg.
4870 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4873 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4875 (cortex-m0): Use armv6s-m isa.
4876 (cortex-m0plus): Likewise.
4877 (cortex-m1): Likewise.
4878 (cortex-m0.small-multiply): Likewise.
4879 (cortex-m0plus.small-multiply): Likewise.
4880 (cortex-m1.small-multiply): Likewise.
4882 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4883 Jakub Jelinek <jakub@redhat.com>
4885 PR tree-optimization/85692
4886 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4887 source permute as well.
4889 2018-05-11 Martin Liska <mliska@suse.cz>
4892 * doc/extend.texi: Document LLVM style format for no_sanitize
4895 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4897 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4898 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4899 (mode_supports_vsx_dform_quad): Likewise.
4900 (mode_supports_vmx_dform): Move these functions to be next to the
4901 other mode_supports functions.
4902 (mode_supports_dq_form): Likewise.
4903 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4904 mode_supports_dq_form.
4905 (reg_offset_addressing_ok_p): Likewise.
4906 (offsettable_ok_by_alignment): Likewise.
4907 (rs6000_legitimate_offset_address_p): Likewise.
4908 (legitimate_lo_sum_address_p): Likewise.
4909 (rs6000_legitimize_address): Likewise.
4910 (rs6000_legitimize_reload_address): Likewise.
4911 (rs6000_secondary_reload_inner): Likewise.
4912 (rs6000_preferred_reload_class): Likewise.
4913 (rs6000_output_move_128bit): Likewise.
4915 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4917 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4918 Generate SImode target register for null target.
4919 <case IX86_BUILTIN_XGETBV>: Ditto.
4920 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4921 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4923 2018-05-10 Carl Love <cel@us.ibm.com>
4925 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4926 dcbtt and dcbtstt if operands[2] is 0.
4928 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4931 * config/i386/sse.md (usadv64qi): New expander.
4933 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4935 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4936 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4937 -maltivec=be support.
4938 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4939 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4940 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4941 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4942 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4943 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4944 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4945 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4946 altivec_vsumsws): Adjust.
4947 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4948 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4949 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4951 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4952 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4953 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4954 (altivec_lve<VI_char>x): Delete expand.
4955 (*altivec_lve<VI_char>x_internal): Rename to...
4956 (altivec_lve<VI_char>x): ... this.
4957 (altivec_lvxl_<mode>): Delete expand.
4958 (*altivec_lvxl_<mode>_internal): Rename to ...
4959 (altivec_lvxl_<mode>): ... this.
4960 (altivec_stvxl_<mode>): Delete expand.
4961 (*altivec_stvxl_<mode>_internal): Rename to ...
4962 (altivec_stvxl_<mode>): ... this.
4963 (altivec_stve<VI_char>x): Delete expand.
4964 (*altivec_stve<VI_char>x_internal): Rename to ...
4965 (altivec_stve<VI_char>x): ... this.
4966 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4967 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4968 reduc_plus_scal_<mode>): Adjust.
4969 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4971 (rs6000_cpu_cpp_builtins): Adjust.
4972 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4973 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4974 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4975 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4976 -maltivec=be support.
4977 (rs6000_split_vec_extract_var): Adjust.
4978 (rs6000_split_v4si_init): Adjust.
4979 (swap_selector_for_mode): Delete.
4980 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4981 altivec_expand_stvex_be): Delete.
4982 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4983 -maltivec=be support.
4984 (rs6000_gimple_fold_builtin): Ditto.
4985 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4987 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4988 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4989 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4990 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4991 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4992 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4993 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4994 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4995 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4996 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4997 anonymous split): Adjust.
4998 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4999 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
5001 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
5003 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
5004 when --with-gxx-include-dir is also specified.
5005 * configure: Regenerate.
5007 2018-05-09 Jim Wilson <jimw@sifive.com>
5010 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
5011 * config/riscv/t-withmultilib: New.
5012 * config/riscv/withmultilib.h: New.
5013 * doc/install.texi: Document RISC-V --with-multilib-list support.
5015 2018-05-09 Richard Biener <rguenther@suse.de>
5017 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
5019 (vect_bb_vectorization_profitable_p): Adjust. Compute
5020 actual scalar cost using the cost vector and the add_stmt_cost
5023 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5025 PR rtl-optimization/85645
5026 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
5027 in the REG_CFA_REGISTER note for LR, don't leave it empty.
5029 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5031 PR rtl-optimization/85645
5032 * shrink-wrap.c (spread_components): Return a boolean saying if
5033 anything was changed.
5034 (try_shrink_wrapping_separate): Iterate spread_components until
5035 nothing changes anymore.
5037 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5039 PR rtl-optimization/85645
5040 * regrename.c (build_def_use): Also kill the chains that include the
5041 destination of a REG_CFA_REGISTER note.
5043 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
5045 PR rtl-optimization/85645
5046 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
5047 insn that has a REG_CFA_REGISTER note.
5049 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
5051 * cfgexpand.c (expand_clobber): New function.
5052 (expand_gimple_stmt_1): Use it.
5053 * tree-vect-stmts.c (vect_clobber_variable): New function,
5055 (vectorizable_simd_clone_call): ...here.
5056 (vectorizable_store): Emit a clobber either side of an
5057 IFN_STORE_LANES sequence.
5058 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
5060 2018-05-09 Tom de Vries <tom@codesourcery.com>
5063 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
5064 (define_insn "trap_if_false"): Add exit after trap.
5066 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
5068 PR rtl-optimization/85638
5069 * bb-reorder.c: Include common/common-target.h.
5070 (create_forwarder_block): New function extracted from...
5071 (fix_up_crossing_landing_pad): ...here. Rename into...
5072 (dw2_fix_up_crossing_landing_pad): ...this.
5073 (sjlj_fix_up_crossing_landing_pad): New function.
5074 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
5075 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
5076 from both partitions and exit the loop after one iteration.
5078 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5081 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5083 (Basic PowerPC Built-in Functions): The new name of the
5084 subsection previously known as "PowerPC Built-in Functions".
5085 (Basic PowerPC Built-in Functions Available on all Configurations):
5087 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5089 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5091 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5093 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5096 2018-05-08 Jim Wilson <jimw@sifive.com>
5098 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
5099 (LD_EMUL_SUFFIX): New.
5100 (LINK_SPEC): Use it.
5102 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
5104 * doc/extend.texi (PowerPC Built-in Functions): Rename this
5106 (Basic PowerPC Built-in Functions): The new name of the
5107 subsection previously known as "PowerPC Built-in Functions".
5108 (Basic PowerPC Built-in Functions Available on all Configurations):
5110 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
5112 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
5114 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
5116 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
5119 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5122 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
5123 after cmpelim optimization.
5125 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
5127 * config.gcc: Support "goldmont".
5128 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
5129 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5131 * config/i386/i386.c (m_GOLDMONT): Define.
5132 (processor_target_table): Add "goldmont".
5133 (PTA_GOLDMONT): Define.
5134 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
5135 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
5136 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
5137 (fold_builtin_cpu): Add "goldmont".
5138 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
5139 (ix86_option_override_internal): Add "goldmont".
5140 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
5141 (processor_type): Add PROCESSOR_GOLDMONT.
5142 * config/i386/i386.md: Add CPU "glm".
5143 * config/i386/glm.md: New file.
5144 * config/i386/x86-tune.def: Add m_GOLDMONT.
5145 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
5147 2018-05-08 Jakub Jelinek <jakub@redhat.com>
5150 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
5152 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
5153 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
5154 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
5157 * config/i386/i386.c (ix86_fold_builtin): Handle
5158 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
5161 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
5162 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
5164 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
5167 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
5168 (check_arch): Likewise.
5169 (check_fpu): Return the result rather than printing it.
5170 (end arch): Fix operator precedence.
5171 (end cpu): Likewise.
5172 (END): Print the result from check_fpu.
5174 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5175 Alan Hayward <alan.hayward@arm.com>
5176 David Sherwood <david.sherwood@arm.com>
5178 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
5179 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
5180 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
5181 (*fcmuo<mode>_and): New patterns.
5183 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
5185 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
5186 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
5187 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
5188 (cmp_op, sve_imm_con): New code attributes.
5189 (SVE_COND_INT_CMP, imm_con): Delete.
5190 (cmp_op): Remove above unspecs from int attribute.
5191 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
5193 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
5194 comparison-specific unspecs.
5195 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
5196 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
5197 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
5198 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
5199 (*vec_fcm<cmp_op><mode>): Rename to...
5200 (*fcm<cmp_op><mode>): ...this and adjust likewise.
5201 (*vec_fcmuo<mode>): Rename to...
5202 (*fcmuo<mode>): ...this and adjust likewise.
5203 (*pred_fcm<cmp_op><mode>): New pattern.
5204 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
5205 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
5207 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
5209 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
5210 (aarch64_emit_sve_predicated_cond): New function.
5211 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
5212 (aarch64_emit_unspec_cond_or): Replace with...
5213 (aarch64_emit_sve_or_conds): ...this new function. Use
5214 aarch64_emit_sve_ptrue_op for the individual comparisons and
5215 aarch64_emit_binop to OR them together.
5216 (aarch64_emit_inverted_unspec_cond): Replace with...
5217 (aarch64_emit_sve_inverted_cond): ...this new function. Use
5218 aarch64_emit_sve_ptrue_op for the comparison and
5219 aarch64_emit_unop to invert the result.
5220 (aarch64_expand_sve_vec_cmp_float): Update after the above
5221 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
5223 2018-05-07 Nathan Sidwell <nathan@acm.org>
5225 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
5226 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
5227 (Backwards Compatibility): Likewise.
5229 2018-05-07 Luis Machado <luis.machado@linaro.org>
5233 2018-05-07 Luis Machado <luis.machado@linaro.org>
5235 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5236 <prefetch_dynamic_strides>: New const bool field.
5237 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5238 prefetch_dynamic_strides.
5239 (exynosm1_prefetch_tune): Likewise.
5240 (thunderxt88_prefetch_tune): Likewise.
5241 (thunderx_prefetch_tune): Likewise.
5242 (thunderx2t99_prefetch_tune): Likewise.
5243 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5245 (aarch64_override_options_internal): Update to set
5246 PARAM_PREFETCH_DYNAMIC_STRIDES.
5247 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5248 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5249 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5250 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5251 prefetch-dynamic-strides setting.
5253 2018-05-07 Luis Machado <luis.machado@linaro.org>
5255 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5256 <minimum_stride>: New const int field.
5257 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5258 minimum_stride field.
5259 (exynosm1_prefetch_tune): Likewise.
5260 (thunderxt88_prefetch_tune): Likewise.
5261 (thunderx_prefetch_tune): Likewise.
5262 (thunderx2t99_prefetch_tune): Likewise.
5263 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5264 (aarch64_override_options_internal): Update to set
5265 PARAM_PREFETCH_MINIMUM_STRIDE.
5266 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5267 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5268 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5269 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5270 stride is constant and is below the minimum stride threshold.
5272 2018-05-07 Luis Machado <luis.machado@linaro.org>
5274 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
5277 2018-05-07 Luis Machado <luis.machado@linaro.org>
5279 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5280 <prefetch_dynamic_strides>: New const bool field.
5281 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5282 prefetch_dynamic_strides.
5283 (exynosm1_prefetch_tune): Likewise.
5284 (thunderxt88_prefetch_tune): Likewise.
5285 (thunderx_prefetch_tune): Likewise.
5286 (thunderx2t99_prefetch_tune): Likewise.
5287 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
5289 (aarch64_override_options_internal): Update to set
5290 PARAM_PREFETCH_DYNAMIC_STRIDES.
5291 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
5292 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
5293 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
5294 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
5295 prefetch-dynamic-strides setting.
5297 2018-05-07 Luis Machado <luis.machado@linaro.org>
5299 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
5300 <minimum_stride>: New const int field.
5301 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
5302 minimum_stride field.
5303 (exynosm1_prefetch_tune): Likewise.
5304 (thunderxt88_prefetch_tune): Likewise.
5305 (thunderx_prefetch_tune): Likewise.
5306 (thunderx2t99_prefetch_tune): Likewise.
5307 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
5308 (aarch64_override_options_internal): Update to set
5309 PARAM_PREFETCH_MINIMUM_STRIDE.
5310 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
5311 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
5312 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
5313 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
5314 stride is constant and is below the minimum stride threshold.
5316 2018-05-06 Jakub Jelinek <jakub@redhat.com>
5319 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
5320 the type is addressable. Don't force op into register if it has
5323 2018-05-05 Roland McGrath <mcgrathr@google.com>
5326 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
5327 any section for which we don't know a specific type it should have,
5328 regardless of name. Previously this was done only for the exact
5329 names ".init_array", ".fini_array", and ".preinit_array".
5330 (default_elf_asm_named_section): Add comment about
5331 relationship with default_section_type_flags and SECTION_NOTYPE.
5332 (get_section): Don't consider it a type conflict if one side has
5333 SECTION_NOTYPE and the other doesn't, as long as neither has the
5334 SECTION_BSS et al used in the default_section_type_flags logic.
5336 2018-05-05 Tom de Vries <tom@codesourcery.com>
5339 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
5340 (workaround_barsyncs): New function.
5341 (nvptx_reorg): Use workaround_barsyncs.
5342 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
5343 (define_expand "nvptx_membar_cta"): New define_expand.
5344 (define_insn "*nvptx_membar_cta"): New insn.
5346 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
5348 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
5349 To improve optimization opportunities.
5350 * builtin-types.def: The new needed builtin types for the above.
5352 2018-05-04 Richard Biener <rguenther@suse.de>
5354 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
5355 * gimple-ssa-store-merging.c
5356 (imm_store_chain_info::output_merged_store): Remove redundant create,
5357 release split_store vector contents on failure.
5358 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
5359 scalar stmt vector on cache hit.
5361 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
5363 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
5365 * config.gcc (powerpc-xilinx-eabi*): Remove.
5366 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
5368 (fusion_addis_mem_combo_load): Ditto.
5369 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
5371 (rs6000_cpu_cpp_builtins): Ditto.
5372 * config/rs6000/rs6000-linux.c
5373 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
5374 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
5375 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
5377 (rs6000_setup_reg_addr_masks): Ditto.
5378 (rs6000_init_hard_regno_mode_ok): Ditto.
5379 (rs6000_option_override_internal): Ditto.
5380 (legitimate_lo_sum_address_p): Ditto.
5381 (rs6000_legitimize_address): Ditto.
5382 (rs6000_legitimize_reload_address): Ditto.
5383 (rs6000_legitimate_address_p): Ditto.
5384 (abi_v4_pass_in_fpr): Ditto.
5385 (setup_incoming_varargs): Ditto.
5386 (rs6000_gimplify_va_arg): Ditto.
5387 (rs6000_split_multireg_move): Ditto.
5388 (rs6000_savres_strategy): Ditto.
5389 (rs6000_emit_prologue_components): Ditto.
5390 (rs6000_emit_epilogue_components): Ditto.
5391 (rs6000_emit_prologue): Ditto.
5392 (rs6000_emit_epilogue): Ditto.
5393 (rs6000_elf_file_end): Ditto.
5394 (rs6000_function_value): Ditto.
5395 (rs6000_libcall_value): Ditto.
5396 * config/rs6000/rs6000.h: Ditto.
5397 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
5398 (TARGET_MINMAX): ... this. New.
5399 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
5400 * config/rs6000/rs6000.md: Remove Xilinx FP support.
5401 (*movsi_internal1_single): Delete.
5402 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
5403 mfpu=, mxilinx-fpu): Delete.
5404 * config/rs6000/singlefp.h: Delete.
5405 * config/rs6000/sysv4.h: Remove Xilinx FP support.
5406 * config/rs6000/t-rs6000: Ditto.
5407 * config/rs6000/t-xilinx: Delete.
5408 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
5409 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
5410 (VStype_simple): Delete.
5411 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
5412 * config/rs6000/xfpu.h: Delete.
5413 * config/rs6000/xfpu.md: Delete.
5414 * config/rs6000/xilinx.h: Delete.
5415 * config/rs6000/xilinx.opt: Delete.
5416 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
5417 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
5419 2018-05-04 Tom de Vries <tom@codesourcery.com>
5422 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
5425 2018-05-04 Richard Biener <rguenther@suse.de>
5428 * tree-complex.c (update_complex_assignment): We are always in SSA form.
5429 (expand_complex_div_wide): Likewise.
5430 (expand_complex_operations_1): Likewise.
5431 (expand_complex_libcall): Preserve EH info of the original stmt.
5432 (tree_lower_complex): Handle removed blocks.
5433 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
5434 on complex multiplication and division libcall builtins.
5436 2018-05-04 Richard Biener <rguenther@suse.de>
5439 * fold-const.c (negate_expr_p): Restrict negation of operand
5440 zero of a division to when we know that can happen without
5442 (fold_negate_expr_1): Likewise.
5444 2018-05-04 Jakub Jelinek <jakub@redhat.com>
5447 * real.h (real_nextafter): Declare.
5448 * real.c (real_nextafter): New function.
5449 * fold-const-call.c (fold_const_nextafter): New function.
5450 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
5451 CASE_CFN_NEXTTOWARD.
5452 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
5453 even when arg1_mode is different from arg0_mode.
5455 2018-05-03 Nathan Sidwell <nathan@acm.org>
5457 * doc/extend.texi (Deprecated Features): Remove
5459 (Backwards Compatibility): Likewise.
5460 * doc/invoke.texi (C++ Language Options): Likewise.
5461 (C++ Dialect Options): Likewise.
5463 2018-05-03 Jakub Jelinek <jakub@redhat.com>
5466 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5467 _mm512_mask_mullox_epi64): New intrinsics.
5469 2018-05-03 Tom de Vries <tom@codesourcery.com>
5472 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5473 dump files): Add offload-tree.
5475 2018-05-03 Richard Biener <rguenther@suse.de>
5477 PR tree-optimization/85615
5478 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
5479 to loops not nested in BBs loop father to avoid creating multi-entry
5482 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5484 PR tree-optimization/70291
5485 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
5486 arguments. Change return type to tree. Emit libcall as a new
5487 statement rather than replacing existing one when inplace_p is true.
5488 (expand_complex_multiplication_components): New function.
5489 (expand_complex_multiplication): Expand floating-point complex
5490 multiplication using the above.
5491 (expand_complex_division): Rename inner_type parameter to type.
5492 Update expand_complex_libcall call-site.
5493 (expand_complex_operations_1): Update expand_complex_multiplication
5494 and expand_complex_division call-sites.
5496 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5499 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5500 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5501 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
5502 the highest significant bit of the shift count mask is clear. In
5503 check whether and[sq]i3 is needed verify that all significant bits
5504 of the shift count other than the highest are set.
5506 2018-05-02 Tom de Vries <tom@codesourcery.com>
5509 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
5510 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
5511 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
5512 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
5513 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
5514 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
5515 __builtin_goacc_parlevel_size.
5517 2018-05-02 Richard Biener <rguenther@suse.de>
5519 PR tree-optimization/85597
5520 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
5521 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5524 2018-05-02 Tom de Vries <tom@codesourcery.com>
5527 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5528 dump files): Add ltrans-tree.
5530 2018-05-02 Tom de Vries <tom@codesourcery.com>
5533 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5534 dump files): Add wpa-ipa.
5536 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5538 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5539 powerpc*-*-linux*paired* target.
5540 * config/rs6000/750cl.h: Delete.
5541 * config/rs6000/paired.h: Delete.
5542 * config/rs6000/paired.md: Delete.
5543 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5545 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5546 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5547 comment. Remove paired float support.
5548 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5549 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5551 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5552 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5554 * config/rs6000/rs6000.c: Remove paired float support.
5555 (paired_expand_vector_init, paired_expand_vector_move,
5556 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5557 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5558 paired_expand_builtin, paired_expand_predicate_builtin,
5559 paired_init_builtins): Delete.
5560 * config/rs6000/rs6000.h: Remove paired float support.
5561 * config/rs6000/rs6000.md: Remove paired float support.
5562 (move_from_CR_ov_bit): Delete.
5563 * config/rs6000/rs6000.opt (mpaired): Delete.
5564 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5565 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5567 2018-05-02 Richard Biener <rguenther@suse.de>
5570 * gimplify.c (gimplify_save_expr): When in SSA form allow
5571 SAVE_EXPRs to compute to SSA vars.
5573 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5576 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5577 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5578 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5579 clobber operands[2], instead use a new pseudo. Formatting fixes.
5581 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5583 PR tree-optimization/85586
5584 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5585 exit early for statements in the same group if the accesses are
5588 2018-05-02 Tom de Vries <tom@codesourcery.com>
5591 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5594 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5596 PR tree-optimization/85143
5597 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5599 2018-05-01 Tom de Vries <tom@codesourcery.com>
5602 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5603 not found" error message.
5605 2018-05-01 Tom de Vries <tom@codesourcery.com>
5608 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5609 * vec.c (test_ordered_remove_if): New function.
5610 (vec_c_tests): Call test_ordered_remove_if.
5611 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5612 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5613 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5614 VEC_ORDERED_REMOVE_IF.
5616 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5618 PR tree-optimization/82665
5619 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5620 pointer subtraction where arguments come from a memchr call.
5622 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5624 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5625 --push-state --as-needed and --pop-state instead of --as-needed and
5626 --no-as-needed if ld supports it.
5627 * configure: Regenerated.
5630 * doc/install.texi2html: Replace _002d with - and _002a with * in
5631 generated html files using sed.
5633 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5636 * gcc-rich-location.c (blank_line_before_p): New function.
5637 (use_new_line): New function.
5638 (gcc_rich_location::add_fixit_insert_formatted): New function.
5639 * gcc-rich-location.h
5640 (gcc_rich_location::add_fixit_insert_formatted): New function.
5642 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5644 * selftest.c (assert_streq): Rename "expected" and "actual" to
5645 "val1" and "val2". Extend NULL-handling to cover both inputs
5646 symmetrically, while still requiring both to be non-NULL for a pass.
5647 * selftest.h (assert_streq): Rename "expected" and "actual" to
5649 (ASSERT_EQ): Likewise.
5650 (ASSERT_EQ_AT): Likewise.
5651 (ASSERT_KNOWN_EQ): Likewise.
5652 (ASSERT_KNOWN_EQ_AT): Likewise.
5653 (ASSERT_NE): Likewise.
5654 (ASSERT_MAYBE_NE): Likewise.
5655 (ASSERT_MAYBE_NE_AT): Likewise.
5656 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5657 the assertion to pass.
5658 (ASSERT_STREQ_AT): Likewise.
5660 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5662 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5663 interaction with -pie.
5665 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5667 * selftest.h: Fix alphabetization of per-source-file selftest
5670 2018-04-30 Jason Merrill <jason@redhat.com>
5672 PR c++/61982 - dead stores to destroyed objects.
5673 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5676 2018-04-30 Jason Merrill <jason@redhat.com>
5678 * tree.c (build_clobber): New.
5679 * tree.h: Declare it.
5680 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5682 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5684 * diagnostic-show-locus.c (layout::layout): Update for
5685 location_get_source_line returning a char_span.
5686 (struct char_span): Move to input.h.
5687 (struct correction): Update for fields in char_span becoming
5689 (struct source_line): Update for location_get_source_line
5690 returning a char_span.
5691 (layout::print_line): Likewise.
5692 * edit-context.c (edited_file::print_content): Likewise.
5693 (edited_file::print_diff_hunk): Likewise.
5694 (edited_file::print_run_of_changed_lines): Likewise.
5695 (edited_file::get_num_lines): Likewise.
5696 (edited_line::edited_line): Likewise.
5697 * final.c (asm_show_source): Likewise.
5698 * input.c (location_get_source_line): Convert return type
5699 from const char * to char_span, losing the final "line_len"
5701 (dump_location_info): Update for the above.
5702 (get_substring_ranges_for_loc): Likewise. Use a char_span
5703 when handling the literal within the line.
5704 (test_reading_source_line): Update for location_get_source_line
5705 returning a char_span.
5706 * input.h (class char_span): Move here from
5707 diagnostic-show-locus.c, converting from a struct to a class.
5708 Make data members private.
5709 (char_span::operator bool): New.
5710 (char_span::length): New.
5711 (char_span::get_buffer): New.
5712 (char_span::operator[]): New.
5713 (char_span::subspan): Make const.
5714 (char_span::xstrdup): New.
5715 (location_get_source_line): Convert return type from const char *
5716 to char_span, losing the final "line_size" param.
5718 2018-04-30 Jan Hubicka <jh@suse.cz>
5720 * lto-wrapper.c (ltrans_priorities): New static var.
5721 (cmp_priority): New.
5722 (run_gcc): Read priorities and if doing parallel build order
5723 the Makefile by them.
5725 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5727 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5729 2018-04-30 Richard Biener <rguenther@suse.de>
5731 * tree-cfg.c (verify_address): Remove base argument, add
5732 flag whether to check TREE_ADDRESSABLE and do that.
5733 (verify_expr): Remove.
5734 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5735 (verify_gimple_assign_single): Likewise.
5736 (verify_gimple_switch): Likewise.
5737 (verify_expr_location_1): Dereference tp once. Add (disabled)
5738 piece from verify_expr.
5739 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5741 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5743 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5745 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5747 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5748 (small_data_pattern): Likewise.
5749 (arc_rewrite_small_data): Likewise.
5750 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5751 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5752 (get_symbol_alignment): New function.
5753 (legitimate_small_data_address_p): Likewise.
5754 (legitimate_scaled_address): Update, call
5755 legitimate_small_data_address_p.
5756 (output_sdata): New static variable.
5757 (arc_print_operand): Update how we handle small data operands.
5758 (arc_print_operand_address): Likewise.
5759 (arc_legitimate_address_p): Update, use
5760 legitimate_small_data_address_p.
5761 (arc_rewrite_small_data_p): Remove.
5762 (arc_rewrite_small_data_1): Likewise.
5763 (arc_rewrite_small_data): Likewise.
5764 (small_data_pattern): Likewise.
5765 (compact_sda_memory_operand): Update to use
5766 legitimate_small_data_address_p and get_symbol_alignment.
5767 (prepare_move_operands): Don't rewite sdata pattern.
5768 (prepare_extend_operands): Remove.
5769 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5771 (zero_extendqisi2): Likewise.
5772 (zero_extendhisi2): Likewise.
5773 (extendqihi2): Likewise.
5774 (extendqisi2): Likewise.
5775 (extendhisi2): Likewise.
5779 * config/arc/constraints.md (Usd): Change it to memory constraint.
5781 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5783 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5784 as source of std instructions.
5785 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5786 allow 6-bit constants as source for store instructions.
5787 (movdi_insn): Update instruction pattern to allow 6-bit constants
5788 as source for store instructions.
5790 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5792 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5794 2018-04-30 Nathan Sidwell <nathan@acm.org>
5795 Sandra Loosemore <sandra@codesourcery.com>
5797 * dumpfile.c (dump_open): Allow '-' for stdout.
5798 * doc/invoke.texi (Developer Options): Document dump filename
5799 determination early. Document stdin/stdout selection.
5801 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5803 Microblaze Target: PIC data text relative
5805 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5806 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5808 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5809 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5810 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5811 New addressing mode for data-text relative position indepenedent code.
5812 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5813 'ADDRESS_SYMBOLIC_TXT_REL'.
5814 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5815 (microblaze_legitimate_pic_operand): Exclude function calls from
5816 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5817 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5819 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5820 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5821 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5822 for 'address + offset'.
5823 (microblaze_expand_prologue): Add new function prologue call for
5825 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5826 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5827 table in case of TARGET_PIC_DATA_TEXT_REL.
5828 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5829 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5830 Add new macros 'UNSPEC_TEXT',
5831 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5832 + exclude function calls from 'UNSPEC_PLT' in case of data text
5834 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5835 new target hook for generating address diff vector tables in case of
5837 * doc/tm.texi : Regenerate.
5838 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5839 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5840 of addr diff vector generation.
5841 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5842 target hook definition.
5843 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5844 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5845 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5846 option for data text relative.
5848 2018-04-30 Richard Biener <rguenther@suse.de>
5850 * tree-chrec.h (evolution_function_is_constant_p): Remove
5852 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5854 2018-04-30 Richard Biener <rguenther@suse.de>
5857 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5859 2018-04-30 Richard Biener <rguenther@suse.de>
5861 PR tree-optimization/28364
5862 PR tree-optimization/85275
5863 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5864 copying first exit test.
5866 2018-04-28 Mark Wielaard <mark@klomp.org>
5868 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5870 (dwarf_AT): Handle DW_AT_addr_base.
5871 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5873 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5876 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5877 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5878 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5879 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5881 2018-04-28 Richard Biener <rguenther@suse.de>
5883 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5884 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5885 to reflect use. Only add interesting stmts.
5887 2018-04-27 Martin Jambor <mjambor@suse.cz>
5890 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5891 the jump function allows for passing through aggregate values.
5893 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5895 * input.h (in_system_header_at): Convert from macro to inline
5897 (from_macro_expansion_at): Likewise.
5898 (from_macro_definition_at): Likewise.
5900 2018-04-27 Jeff Law <law@redhat.com>
5902 * config.gcc: Mark tile* targets as deprecated/obsolete.
5904 2018-04-27 Richard Biener <rguenther@suse.de>
5906 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5909 2018-04-27 Richard Biener <rguenther@suse.de>
5911 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5913 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5915 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5916 with Yd constraint. Set "preferred_for_speed" attribute from
5917 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5919 (*movdi_internal): Ditto.
5920 (movti_interunit splitters): Remove
5921 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5922 (movdi_interunit splitters): Ditto.
5923 * config/i386/constraints.md (Ye): Remove.
5924 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5926 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5929 * config/aarch64/constraints.md (Usg): Limit to 31.
5932 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5934 PR tree-optimization/85529
5935 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5936 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5937 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5938 zero extension or masking of the MSB bit.
5939 (optimize_range_tests): Add FIRST_BB argument, pass it through
5940 to optimize_range_tests_var_bound.
5941 (maybe_optimize_range_tests, reassociate_bb): Adjust
5942 optimize_range_tests callers.
5944 2018-04-26 Richard Biener <rguenther@suse.de>
5945 Jakub Jelinek <jakub@redhat.com>
5947 * cgraph.h (symbol_table): Just declare debug method here.
5948 * symtab.c (symbol_table::debug): Define.
5950 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5952 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5954 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5956 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5957 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5958 (*movdi_internal): Substitute Yi and Yj constraint with x
5959 and Ym and Yn constraint with y constraint. Update "isa"
5960 attribute and set "preferred_for_speed" attribute from
5961 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5962 (*movsi_internal): Ditto.
5963 (*movdf_internal): Ditto.
5964 (*movsf_internal): Ditto.
5965 (*zero_extendsidi2): Ditto.
5966 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5967 (sse2_loadld): Ditto.
5968 (*vec_extract<ssevecmodelower>_0): Ditto.
5969 (*vec_extractv4si_0_zext_sse4): Ditto.
5970 (vec_concatv2di): Ditto.
5971 (*vec_dup<mode>): Ditto.
5972 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5973 * config/i386/constraints.md (Yi): Remove.
5978 2018-04-26 Nathan Sidwell <nathan@acm.org>
5980 * dumpfile.c (dump_open): New.
5981 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5982 (dump_finish): Detect stdio/stderr by value not name.
5984 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5986 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5988 2018-04-26 Tom de Vries <tom@codesourcery.com>
5991 * config/nvptx/nvptx.c (verify_neutering_jumps)
5992 (verify_neutering_labels): New function
5993 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5995 2018-04-26 Tom de Vries <tom@codesourcery.com>
5998 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5999 (nvptx_single): Use needs_neutering_p to skip over insns that do not
6002 2018-04-26 Richard Biener <rguenther@suse.de>
6003 Tom de Vries <tom@codesourcery.com>
6006 * lto-streamer-out.c (output_function): Fixup loops if required to match
6007 discovery done in the reader.
6009 2018-04-26 Richard Biener <rguenther@suse.de>
6011 PR tree-optimization/85116
6012 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
6013 have a loop exit from the single latch predecessor. Remove
6014 case of header with just condition.
6015 (ch_base::copy_headers): Exclude infinite loops from any
6017 (pass_ch::execute): Record exits.
6019 2018-04-26 Richard Biener <rguenther@suse.de>
6021 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6022 prologue cost vector and pass it to vect_get_load_cost.
6023 (vect_get_peeling_costs_all_drs): Likewise.
6024 (vect_peeling_hash_get_lowest_cost): Likewise.
6025 (vect_enhance_data_refs_alignment): Likewise.
6027 2018-04-26 Richard Biener <rguenther@suse.de>
6030 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
6031 checking of integer<->pointer conversions.
6032 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
6033 sign-/zero-extending pointer types.
6034 (expand_omp_for_static_chunk): Likewise.
6036 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
6037 Jean Lee <xiaoyur347@gmail.com>
6039 * config/mips/mips.c (mips_asan_shadow_offset): New function.
6040 (TARGET_ASAN_SHADOW_OFFSET): Define.
6041 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
6042 true for -fsanitize=address.
6044 2018-04-25 Mark Wielaard <mark@klomp.org>
6046 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
6049 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6051 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
6052 than "alu", remove explicit "memory" and "imm_disp" attributes.
6053 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
6056 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
6057 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
6058 gen_lowpart_no_emit.
6060 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
6063 * config/i386/i386.c (ix86_expand_builtin): Change memory
6064 operand to XI, extend p0 to Pmode.
6065 * config/i386/i386.md: Change unspec volatile and operand
6066 1 mode to XI, change operand 0 mode to P.
6068 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6070 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
6071 GET_MODE_MASK before any checking.
6072 (nds32_can_use_bset_p): Likewise.
6073 (nds32_can_use_btgl_p): Likewise.
6075 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6077 * config/nds32/nds32-doubleword.md: New define_split pattern for
6078 illegal register number.
6080 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6082 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
6084 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
6086 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
6088 2018-04-25 Richard Biener <rguenther@suse.de>
6090 * lto-streamer.h (LTO_major_version): Bump to 8.
6092 2018-04-25 Jakub Jelinek <jakub@redhat.com>
6094 * BASE-VER: Set to 9.0.0.
6096 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
6098 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
6099 in __abskf2 and __powikf2.
6101 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6104 * config/aarch64/constraints.md (Usg, Usj): New constraints.
6105 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
6106 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
6107 Use the above on operand 2. Reindent.
6108 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
6110 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6113 * common/config/i386/i386-common.c (ix86_handle_option): Don't
6115 * config/i386/i386.opt (mcet): Removed.
6116 * doc/install.texi: Remove -mcet documentation.
6117 * doc/invoke.texi: Likewise.
6119 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
6122 * doc/install.texi: Remove -mcet from bootstrap-cet.
6124 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6127 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
6128 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
6132 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
6133 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
6134 containing a CONST_VECTOR.
6136 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
6138 * doc/install.texi: Update newlib dependency for nvptx.
6140 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6143 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
6144 instead of INTVAL when shifting x left.
6146 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
6148 PR tree-optimization/85478
6149 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
6150 vect_grouped_store_supported for single element vectors.
6152 2018-04-24 Richard Biener <rguenther@suse.de>
6155 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
6156 load cost increase to the case of non-constant step.
6158 2018-04-24 Jakub Jelinek <jakub@redhat.com>
6161 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
6162 destination if any_malformed_asm.
6164 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
6167 * expr.c (store_field): In the bitfield case, if the value comes from
6168 a function call and is returned in registers by means of a PARALLEL,
6169 do not change the mode of the temporary unless BLKmode and VOIDmode.
6171 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
6173 PR rtl-optimization/85423
6174 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
6175 dependencies to debug insns when the previous insn is non-debug.
6177 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6179 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
6180 enums into a single definition.
6181 (fls): Fix predicates and printing.
6184 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
6186 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
6187 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
6188 and short u6 immediate.
6189 (check_if_valid_sleep_operand): Remove.
6190 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
6192 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6194 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
6195 flag_always_save_lp condition.
6196 * config/nds32/nds32.opt (malways-save-lp): New option.
6198 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6200 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
6201 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
6202 * config/nds32/nds32.h
6203 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
6204 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
6206 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6208 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
6209 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
6211 2018-04-22 Shiva Chen <shiva0217@gmail.com>
6212 Chung-Ju Wu <jasonwucj@gmail.com>
6214 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
6216 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
6217 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
6219 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6221 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
6223 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6225 * config/nds32/nds32-protos.h (nds32_data_alignment,
6226 nds32_local_alignment): Declare.
6227 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
6228 nds32_local_alignment): New functions.
6229 (TARGET_CONSTANT_ALIGNMENT): Define.
6230 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
6232 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6234 * config/nds32/nds32.c
6235 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
6236 (TARGET_MODES_TIEABLE_P): Likewise.
6238 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
6240 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
6243 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
6244 Chung-Ju Wu <jasonwucj@gmail.com>
6246 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6247 for unaligned access.
6248 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6250 * config/nds32/nds32-intrinsic.md: Likewise.
6251 * config/nds32/nds32_intrinsic.h: Likewise.
6252 * config/nds32/nds32.h (nds32_builtins): Likewise.
6253 * config/nds32/nds32.opt (munaligned-access): New option.
6254 * config/nds32/nds32.c (nds32_asm_file_start): Display
6255 flag_unaligned_access status.
6257 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6259 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
6260 -mno-relax is present.
6261 * config/riscv/linux.h (LINK_SPEC): Ditto.
6263 2018-04-20 Martin Sebor <msebor@redhat.com>
6266 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
6268 (gimple_fold_builtin_stxcpy_chk): Same.
6269 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
6271 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
6274 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
6275 __powikf2 when long double is IEEE 128-bit.
6277 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
6279 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
6280 step to make sure stack always aligned.
6282 2018-04-20 Carl Love <cel@us.ibm.com>
6285 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
6286 size check for arg0.
6288 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
6289 Tom de Vries <tom@codesourcery.com>
6292 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
6293 Emit insns for calls too.
6294 (nvptx_find_par): Always look for worker-level predecessor insn.
6295 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
6297 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
6298 (nvptx_process_pars): Propagate frames for calls.
6300 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
6303 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
6305 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
6306 (ix86_handle_option): Don't handle OPT_mibt.
6307 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
6309 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
6311 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
6312 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
6313 (ix86_target_macros): Define __CET__ with flag_cf_protection
6314 for -fcf-protection.
6315 * config/i386/i386.c (isa2_opts): Remove -mibt.
6316 * config/i386/i386.h (TARGET_IBT): Removed.
6317 (TARGET_IBT_P): Likewise.
6318 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
6319 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
6320 * config/i386/i386.opt (mcet): Update help message.
6323 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
6324 -mcet as an alias for -mshstk.
6326 2018-04-20 Richard Biener <rguenther@suse.de>
6329 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
6330 complexity by forcing a single use of the multiply operand.
6332 2018-04-20 Martin Jambor <mjambor@suse.cz>
6335 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
6336 recursion dependency to only apply to non-clones.
6338 2018-04-20 Martin Jambor <mjambor@suse.cz>
6341 * ipa-cp.c (create_specialized_node): Check that clones of
6342 self-recursive edges exist during IPA-CP.
6344 2018-04-19 Toon Moene <toon@moene.org>
6346 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
6349 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6351 PR tree-optimization/85467
6352 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
6353 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
6354 VECTOR_CST element to type.
6356 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6359 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
6360 * config/i386/i386.md (builtin_setjmp_setup): Removed.
6361 (builtin_longjmp): Likewise.
6362 (save_stack_nonlocal): New pattern.
6363 (restore_stack_nonlocal): Likewise.
6365 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6368 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6369 Replace ASM_OUTPUT_LABEL with fprintf.
6371 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
6374 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
6375 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
6376 * config/i386/i386-c.c (ix86_target_macros_internal): Also
6377 define __IBT__ and __SHSTK__ for -fcf-protection.
6378 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
6380 (ix86_trampoline_init): Likewise.
6381 (x86_output_mi_thunk): Likewise.
6382 (ix86_notrack_prefixed_insn_p): Likewise.
6383 (ix86_option_override_internal): Don't disallow -fcf-protection.
6384 * config/i386/i386.md (rdssp<mode>): Also enable for
6386 (incssp<mode>): Likewise.
6387 (nop_endbr): Likewise.
6388 * config/i386/i386.opt (mcet): Change help message to built-in
6392 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
6393 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
6394 enable CET built-in functions.
6396 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
6398 * common/config/i386/i386-common.c
6399 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
6400 OPTION_MASK_ISA_MOVDIRI_UNSET,
6401 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
6402 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
6403 * config.gcc (movdirintrin.h): New header.
6404 * config/i386/cpuid.h (bit_MOVDIRI,
6405 bit_MOVDIR64B): New bits.
6406 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
6408 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
6409 (VOID, PVOID, PCVOID)): New function types.
6410 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
6411 __builtin_ia32_directstoreu_u64,
6412 __builtin_ia32_movdir64b): New builtins.
6413 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
6414 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
6416 (ix86_valid_target_attribute_inner_p): Ditto.
6417 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
6418 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
6419 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
6420 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
6421 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
6422 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
6423 (movdiri<mode>, movdir64b_<mode>): New.
6424 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
6425 * config/i386/immintrin.h: Include movdirintrin.h.
6426 * config/i386/movdirintrin.h: New file.
6427 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
6429 2018-04-19 Richard Biener <rguenther@suse.de>
6432 * cfg.c (clear_bb_flags): When loop state says we have
6433 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
6435 2018-04-19 Richard Biener <rguenther@suse.de>
6437 PR tree-optimization/84737
6438 * tree-vect-data-refs.c (vect_copy_ref_info): New function
6439 copying restrict info.
6440 (vect_setup_realignment): Use it.
6441 * tree-vectorizer.h (vect_copy_ref_info): Declare.
6442 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
6443 the first DR to all generated stores.
6444 (vectorizable_load): Likewise for loads.
6446 2018-04-19 Jakub Jelinek <jakub@redhat.com>
6448 PR tree-optimization/85446
6449 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
6450 the integral and pointer types to have the same precision.
6452 * doc/install.texi: Document --disable-cet being the default and
6455 2018-04-18 Martin Liska <mliska@suse.cz>
6457 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
6460 2018-04-18 Martin Liska <mliska@suse.cz>
6463 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6466 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6467 arguments if they are comparable.
6469 2018-04-18 Martin Liska <mliska@suse.cz>
6472 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6475 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6478 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
6481 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
6482 ENDBR after calling __morestack.
6484 2018-04-18 David Malcolm <dmalcolm@redhat.com>
6487 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
6488 by using gcc_base_ver to generate a gcc_driver_version, and use
6489 it when generating GCC_DRIVER_NAME.
6490 * configure: Regenerate.
6492 2018-04-18 Jakub Jelinek <jakub@redhat.com>
6495 * config.gcc: Obsolete powerpc*-*-*spe*.
6497 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6500 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
6501 (stabstr_D): Change type of unum from unsigned int to
6502 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
6505 2018-04-17 Jim Wilson <jimw@sifive.com>
6508 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
6509 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
6510 Set arg_pointer_offset after using pretend_args_size.
6512 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6514 PR rtl-optimization/85431
6515 * dse.c (record_store): Ignore zero width stores.
6518 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
6519 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
6520 __builtin_stack_restore rather than after it.
6521 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
6522 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6523 argument instead of virtual_dynamic_stack_rtx.
6525 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6527 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6529 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6530 Add note to error message to explain internal mapping of overloaded
6531 built-in function name to non-overloaded built-in function name.
6532 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6535 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6538 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6539 where the inputs overlap with the output.
6541 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6543 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6544 (=v, v) alternative and explicit "memory" attribute.
6545 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6546 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6548 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6549 "sselog1" type instead of "sselog".
6550 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6551 "sselog". Remove explicit "memory" attribute.
6552 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6553 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6555 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6556 "sselog1" type instead of "sselog". Remove explicit "memory"
6558 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6559 use "sselog1" type instead of "sselog". Remove explicit "memory"
6561 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6562 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6564 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6565 "sselog1" type instead of "sselog". Remove explicit "memory"
6567 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6568 use "sselog1" type instead of "sselog". Remove explicit "memory"
6572 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6575 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6578 2018-04-17 Martin Jambor <mjambor@suse.cz>
6581 * ipa-cp.c (create_specialized_node): Call
6582 expand_all_artificial_thunks if necessary.
6584 2018-04-17 Martin Liska <mliska@suse.cz>
6587 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6588 in message, remote space in between '_G' and '('.
6590 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6593 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6594 avx512f_vmcmp<mode>3<round_saeonly_name>,
6595 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6596 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6597 avx512f_rndscale<mode><round_saeonly_name>,
6598 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6599 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6600 Use %<iptr>2 instead of %2 for -masm=intel.
6601 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6602 avx512f_vcvttss2usi<round_saeonly_name>,
6603 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6605 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6606 avx512f_vcvttsd2usi<round_saeonly_name>,
6607 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6608 Use %q1 instead of %1 for -masm=intel.
6609 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6610 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6611 of %3 for -masm=intel.
6612 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6614 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6616 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6618 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6619 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6621 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6623 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6624 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6625 %g1 and one with %0 and %1.
6626 (avx512er_vmrcp28<mode><round_saeonly_name>,
6627 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6629 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6630 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6631 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6632 of %0 and %{%4%} for -masm=intel.
6633 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6634 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6635 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6636 order of %0 and %{%5%}%{z%} for -masm=intel.
6638 2018-04-17 Jan Hubicka <jh@suse.cz>
6641 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6643 2018-04-17 Martin Liska <mliska@suse.cz>
6646 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6647 for target_clone error message. Make default implementation
6648 clone to be a local declaration.
6649 (separate_attrs): Add new argument and check for an empty
6651 (expand_target_clones): Handle it.
6652 (ipa_target_clone): Make redirection just for target_clones
6655 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6656 Tom de Vries <tom@codesourcery.com>
6659 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6660 tiled basic blocks without omp continue statements.
6662 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6665 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6666 vec_extract expression as having side effects to make sure it gets
6669 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6672 * config/i386/i386.c (get_builtin_code_for_version): Check
6675 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6678 * gcc/config.gcc: Support "skylake".
6679 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6681 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6682 (processor_target_table): Add "skylake".
6683 (ix86_option_override_internal): Add "skylake".
6684 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6685 PROCESSOR_CANNONLAKE.
6686 (get_builtin_code_for_version): Fix priority for
6687 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6688 PROCESSOR_SKYLAKE-AVX512.
6689 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6690 (processor_type): Add PROCESSOR_SKYLAKE.
6692 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6693 Jason Merrill <jason@redhat.com>
6696 * convert.c (convert_to_integer_1): Use direct recursion for
6697 enumeral types and types with a precision less than the number
6698 of bits in their mode.
6700 2018-04-16 Julia Koval <julia.koval@intel.com>
6703 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6704 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6706 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6709 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6710 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6711 and -mno-direct-move.
6713 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6716 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6717 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6718 Ensure negative shifts result in {0}.
6720 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6722 PR rtl-optimization/79916
6723 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6724 regs (if any) to define how to gnerate SD moves when LRA is in
6727 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6729 PR rtl-optimization/85393
6730 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6731 * except.c (expand_dw2_landing_pad_for_region): Make static.
6732 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6733 a label and unconditional jump to old_bb, rather than
6734 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6737 PR rtl-optimization/85376
6738 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6739 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6740 instead of a specific value.
6742 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6743 Bin Cheng <bin.cheng@arm.com>
6745 PR tree-optimization/82965
6746 PR tree-optimization/83991
6747 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6748 by_profile_only parameter.
6749 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6750 information if the loop was predicted to iterate too many times.
6751 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6753 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6756 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6759 2018-04-13 Martin Liska <mliska@suse.cz>
6760 Jakub Jelinek <jakub@redhat.com>
6763 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6764 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6765 * builtins.c (expand_builtin_memory_copy_args): Use
6766 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6767 handle dest_addr == pc_rtx.
6769 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6772 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6773 asked to not generate direct moves.
6774 (fix_trunc<mode>si2_stfiwx): Similar.
6775 (fix_trunc<mode>si2_internal): Similar.
6777 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6780 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6781 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6782 lookup if dest in some wider mode is known to be const0_rtx and
6783 if so, record permanent equivalence for it to be ZERO_EXTEND of
6784 the narrower mode destination.
6786 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6788 * lto-streamer-out.c (output_function): Revert 259346.
6789 * omp-expand.c (expand_oacc_for): Likewise.
6791 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6793 PR rtl-optimization/85354
6794 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6795 * sel-sched.c (sel_global_init): ... here.
6797 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6800 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6801 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6802 mode for PE-COFF targets.
6803 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6804 (i386_pe_asm_lto_end): Likewise.
6805 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6806 (TARGET_ASM_LTO_END): Likewise.
6807 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6808 (i386_pe_asm_lto_start): New function.
6809 (i386_pe_asm_lto_end): Likewise.
6811 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6812 Richard Biener <rguenther@suse.de>
6815 * lto-streamer-out.c (output_function): Fix CFG loop state before
6817 * omp-expand.c (expand_oacc_for): Handle calls to internal
6818 functions like regular functions.
6820 2018-04-12 Richard Biener <rguenther@suse.de>
6823 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6824 for the early LTO debug to properly generate references to it
6825 during DIE emission. Do not re-use that for the skeleton for
6827 (dwarf2out_early_finish): Likewise.
6829 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6832 * config/i386/sse.md
6833 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6834 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6835 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6836 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6837 and output is a reg, avoid creating invalid lowpart subreg, but
6838 instead split into a 512-bit move. Don't split if not AVX512VL,
6839 input is xmm16+ reg and output is a mem.
6840 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6841 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6842 xmm16+ reg and output is a mem.
6844 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6846 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6847 also for flag_dwarf2_cfi_asm.
6849 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6851 PR rtl-optimization/85342
6852 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6853 a bool scalar var inside of the loop instead. Don't try to update
6854 recog_data.operand after failed apply_change_group.
6856 2018-04-12 Tom de Vries <tom@codesourcery.com>
6859 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6860 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6861 array with flexible array member as array without given dimension.
6862 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6863 argument for undefined param to true.
6865 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6868 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6869 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6870 from PowerPC section.
6871 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6872 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6873 help text that is too long.
6874 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6875 help text that is too long.
6876 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6877 help text that is too long.
6879 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6881 * config/alpha/alpha.md (stack_probe_internal): Rename
6882 from "probe_stack". Update all callers.
6884 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6886 PR rtl-optimization/84566
6887 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6888 sched_macro_fuse_insns.
6890 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6893 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6894 (compute_block_dependences): ... from here.
6896 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6898 PR tree-optimization/85331
6899 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6900 from int to HOST_WIDE_INT.
6902 2018-04-11 Martin Jambor <mjambor@suse.cz>
6905 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6906 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6907 not the same as the source val.
6908 (cgraph_edge_brings_value_p): New parameter.
6909 (gather_edges_for_value): Pass destination value to
6910 cgraph_edge_brings_value_p.
6911 (perhaps_add_new_callers): Likewise.
6912 (get_info_about_necessary_edges): Likewise and exclude values brought
6913 only by self-recursive edges.
6914 (create_specialized_node): Redirect only clones of self-calling edges.
6915 (+self_recursive_pass_through_p): New function.
6916 (find_more_scalar_values_for_callers_subset): Use it.
6917 (find_aggregate_values_for_callers_subset): Likewise.
6918 (known_aggs_to_agg_replacement_list): Removed.
6919 (decide_whether_version_node): Re-calculate known constants for all
6920 remaining context clones.
6922 2018-04-11 Richard Biener <rguenther@suse.de>
6925 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6926 from early DWARF output.
6927 (dwarf2out_early_finish): Output line info unconditionally into
6928 early DWARF and add reference to it.
6930 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6933 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6934 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6935 other than V2DFmode using iptr mode attribute.
6936 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6938 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6940 PR rtl-optimization/84659
6941 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6943 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6946 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6948 (output_loc_list): Pass address of a dummy size variable even in the
6949 locview handling loop.
6950 (index_location_lists): Add comment on why skip_loc_list_entry can't
6953 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6956 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6959 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6962 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6963 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6964 and -mstring-compare-inline-limit.
6966 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6969 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6970 for stack clash protection in a register whenever we need it to be in
6973 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6975 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6976 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6978 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6981 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6983 (mlong-double-): Ditto.
6984 * config/rs6000/sysv4.opt (msdata=): Ditto.
6985 (mtls-size=): Ditto.
6987 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6989 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6990 erroneous entries for
6991 "vector int vec_ldl (int, long int *)", and
6992 "vector unsigned int vec_ldl (int, unsigned long int *)".
6993 Add comments and entries for
6994 "vector bool char vec_ldl (int, bool char *)",
6995 "vector bool short vec_ldl (int, bool short *)",
6996 "vector bool int vec_ldl (int, bool int *)",
6997 "vector bool long long vec_ldl (int, bool long long *)",
6998 "vector pixel vec_ldl (int, pixel *)",
6999 "vector long long vec_ldl (int, long long *)",
7000 "vector unsigned long long vec_ldl (int, unsigned long long *)".
7001 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
7002 type tree bool_long_long_type_node and correct definition of
7003 bool_V2DI_type_node to make reference to this new type tree.
7004 (rs6000_mangle_type): Replace erroneous reference to
7005 bool_long_type_node with bool_long_long_type_node.
7006 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
7007 comments to emphasize sign distinctions for char and int types and
7008 replace RS6000_BTI_bool_long constant with
7009 RS6000_BTI_bool_long_long constant. Also add comment to restrict
7010 use of RS6000_BTI_pixel.
7011 (bool_long_type_node): Remove this macro definition.
7012 (bool_long_long_type_node): New macro definition
7014 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7016 PR rtl-optimization/85300
7017 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
7018 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
7019 simplify_unary_operation fails.
7021 2018-04-10 Martin Liska <mliska@suse.cz>
7023 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
7024 cgraph_edge and ipa_ref.
7026 2018-04-10 Jakub Jelinek <jakub@redhat.com>
7030 * config/i386/sse.md
7031 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
7032 computation of the VEC_MERGE selector from mask.
7033 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
7034 Fix decoding of the VEC_MERGE selector into mask.
7036 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7038 PR tree-optimization/85286
7039 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
7041 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
7043 * final.c (final_1): Set insn_last_address as well as
7044 insn_current_address.
7046 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7049 * explow.c (emit_stack_probe): Call validize_mem on memory location
7050 before passing it to gen_probe_stack. Create address operand and
7051 legitimize it for the probe_stack_address case.
7053 2018-04-09 Jan Hubicka <jh@suse.cz>
7056 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
7057 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
7058 * tree.c (free_lang_data_in_type): Fix handling of binfos;
7060 (free_lang_data): Rebuild type inheritance graph.
7062 2018-04-09 Martin Sebor <msebor@redhat.com>
7064 * invoke.texi (-finline-small-functions): Mention other optimization
7066 (-findirect-inlining, -fpartial-inlining): Same.
7067 (-finline-functions-called-once): Same.
7068 (-freorder-blocks-and-partition): Same.
7070 2018-04-09 Jan Hubicka <jh@suse.cz>
7073 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
7074 jumps; choose last target that matches the criteria (i.e.
7075 no partition changes for non-crossing jumps).
7076 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
7077 support for redirecting crossing jumps to non-crossing.
7079 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
7081 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
7082 also for naked functions.
7084 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
7086 * config/arc/arc.md (add_shift): New pattern.
7087 (add_shift2): Likewise.
7088 (sub_shift): Likewise.
7089 (sub_shift_cmp0_noout): Likewise.
7090 (compare_si_ashiftsi): Likewise.
7091 (xbfu_cmp0_noout): New combine pattern.
7092 (xbfu_cmp0"): Likewise.
7093 (movsi_set_cc_insn): Place the predicable variant first.
7094 (commutative_binary_cmp0_noout): Remove clobber.
7095 (commutative_binary_cmp0): New pattern.
7096 (noncommutative_binary_cmp0): Likewise.
7097 (noncommutative_binary_cmp0_noout): Likewise.
7098 (noncommutative_binary_comparison_result_used): Removed.
7099 (rsub_cmp0): New pattern.
7100 (rsub_cmp0_noout): Likewise.
7101 (extzvsi): Changed, keep only meaningful variants.
7102 (SQH, SEZ): New iterators.
7103 (SQH_postfix): New mode attribute.
7104 (SEZ_prefix): New code attribute.
7105 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
7106 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
7107 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
7109 (noncommutative_operator): Check the availability of barrel
7112 2018-04-09 Richard Biener <rguenther@suse.de>
7114 PR tree-optimization/85284
7115 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
7116 Only use the niter constraining form of simple_iv when the exit
7119 2018-04-09 Tom de Vries <tom@codesourcery.com>
7122 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7123 (define_expand "*memory_barrier"): New define_expand.
7124 (define_insn "memory_barrier"): New insn.
7126 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7128 PR rtl-optimization/80463
7129 PR rtl-optimization/83972
7130 PR rtl-optimization/83480
7132 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
7133 correct producer for the insn.
7134 (tidy_control_flow): Fixup seqnos in case of debug insns.
7136 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7138 PR rtl-optimization/83913
7140 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
7141 different sched-times when merging exprs.
7143 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7145 PR rtl-optimization/83962
7147 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
7148 tidy_fallthru_edge and tidy_control_flow.
7150 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
7152 PR rtl-optimization/83530
7154 * sel-sched.c (force_next_insn): New global variable.
7155 (remove_insn_for_debug): When force_next_insn is true, also leave only
7156 next insn in the ready list.
7157 (sel_sched_region): When the region wasn't scheduled, make another pass
7158 over it with force_next_insn set to 1.
7160 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7162 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
7164 * config/nds32/constants.md (unspec_volatile_element): Add enum values
7165 for interrupt control.
7166 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7167 functions for interrupt control.
7168 * config/nds32/nds32-intrinsic.md: Likewise.
7169 * config/nds32/nds32_intrinsic.h: Likewise.
7170 * config/nds32/nds32.h (nds32_builtins): Likewise.
7172 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7174 * config/nds32/nds32.c (nds32_init_machine_status,
7175 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
7176 strict_aligned_p field.
7177 (nds32_expand_to_rtl_hook): New function.
7178 (TARGET_EXPAND_TO_RTL_HOOK): Define.
7179 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
7181 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7182 Chung-Ju Wu <jasonwucj@gmail.com>
7184 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
7185 * config/nds32/nds32-n7.md: New file.
7186 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
7187 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
7189 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
7190 * config/nds32/nds32.md (pipeline_model): Add n7.
7191 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
7192 * config/nds32/pipelines.md: Include n7 settings.
7194 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7195 Chung-Ju Wu <jasonwucj@gmail.com>
7197 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
7198 * config/nds32/nds32-e8.md: New file.
7199 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
7200 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
7202 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
7203 * config/nds32/nds32.md (pipeline_model): Add e8.
7204 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
7205 * config/nds32/pipelines.md: Include e8 settings.
7207 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7208 Chung-Ju Wu <jasonwucj@gmail.com>
7210 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
7211 * config/nds32/nds32-n8.md: New file.
7212 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
7213 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
7215 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
7216 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
7217 * config/nds32/nds32.md (pipeline_model): Add n8.
7218 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
7219 * config/nds32/pipelines.md: Include n8 settings.
7221 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
7222 Chung-Ju Wu <jasonwucj@gmail.com>
7224 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
7225 * config/nds32/nds32-n9-2r1w.md: New file.
7226 * config/nds32/nds32-n9-3r2w.md: New file.
7227 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
7228 nds32_register_ports): New or modify for cpu n9.
7229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
7231 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
7232 * config/nds32/nds32-utils.c: New file.
7233 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
7234 TARGET_MUL_SLOW): Define.
7235 * config/nds32/nds32.md (pipeline_model): New attribute.
7236 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
7237 New options that support cpu n9.
7238 * config/nds32/pipelines.md: Include n9 settings.
7239 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
7241 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
7243 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
7244 information if necessary.
7245 (output_cond_branch_compare_zero): Likewise.
7246 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
7247 (nds32_target_alignment): Refine for alignment.
7248 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
7249 (FUNCTION_BOUNDARY): Modify.
7250 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
7252 * config/nds32/nds32.opt (malways-align, malign-functions): New.
7254 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
7256 * config/nds32/constants.md (unspec_volatile_element): Add values for
7257 TLB operation and data prefetch.
7258 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
7259 functions for TLB operation and data prefetch.
7260 * config/nds32/nds32-intrinsic.md: Likewise.
7261 * config/nds32/nds32_intrinsic.h: Likewise.
7262 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
7263 (nds32_print_operand): Likewise.
7264 * config/nds32/nds32.h (nds32_builtins): Likewise.
7266 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
7267 Andrew Pinski <pinsika@gcc.gnu.org>
7270 * match.pd: Use constant_boolean_node of correct type instead of
7271 boolean_true_node or boolean_false_node for simplifying
7272 pointer comparisons to zero.
7274 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7276 PR tree-optimization/80021
7277 * tree.c (verify_type_variant): Make error call in verify_variant_match
7278 translatable and remove final full stop.
7280 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7282 * config/nds32/constants.md (unspec_volatile_element): Add
7283 UNSPEC_VOLATILE_EH_RETURN.
7284 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
7285 nds32_output_stack_pop): Support dwarf exception handling process.
7286 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
7287 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
7288 exception handling process.
7289 (nds32_compute_stack_frame): Likewise.
7290 (nds32_return_addr_rtx): Likewise.
7291 (nds32_initial_elimination_offset): Likewise.
7292 (nds32_expand_prologue): Likewise.
7293 (nds32_expand_epilogue): Likewise.
7294 (nds32_dynamic_chain_address): New function.
7295 * config/nds32/nds32.h (machine_function): Add fields for dwarf
7297 (DYNAMIC_CHAIN_ADDRESS): Define.
7298 (EH_RETURN_DATA_REGNO): Define.
7299 (EH_RETURN_STACKADJ_RTX): Define.
7300 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
7301 patterns for dwarf exception handling.
7303 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7305 * config/nds32/nds32.h: Clean up obsolete macros.
7307 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7309 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7310 Add enum values for particular instructions.
7311 * config/nds32/nds32-intrinsic.c: Implementation of expanding
7312 particular intrinsic functions.
7313 * config/nds32/nds32-intrinsic.md: Likewise.
7314 * config/nds32/nds32_intrinsic.h: Likewise.
7315 * config/nds32/nds32.h (nds32_builtins): Likewise.
7316 * config/nds32/nds32.md (type): Add pbsad and pbsada.
7317 (btst, ave): New patterns for particular instructions.
7319 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7321 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7322 Add enum values for atomic load/store and memory sync.
7323 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
7325 * config/nds32/nds32-intrinsic.md: Likewise.
7326 * config/nds32/nds32_intrinsic.h: Likewise.
7327 * config/nds32/nds32.h (nds32_builtins): Likewise.
7329 2018-04-07 Jakub Jelinek <jakub@redhat.com>
7331 PR tree-optimization/85257
7332 * fold-const.c (native_encode_vector): If not all elts could fit
7333 and off is -1, return 0 rather than offset.
7334 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
7335 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
7336 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
7337 adjust buffer in native_interpret_expr call.
7339 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7341 * config/nds32/constants.md (unspec_volatile_element): Add cache
7342 control enum values.
7343 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
7344 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
7345 * config/nds32/nds32.c (nds32_cctl_names): New.
7346 (nds32_print_operand): Handle cache control register names.
7347 * config/nds32/nds32.h (nds32_builtins): New enum values.
7348 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
7350 * config/nds32/nds32.md (type): Add mmu.
7351 * config/nds32/pipelines.md (simple_insn): Add mmu.
7353 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7355 * config/nds32/nds32.md (type): Remove call.
7356 * config/nds32/pipelines.md (simple_insn): Likewise.
7358 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7360 * config/nds32/constants.md (unspec_volatile_element): Add
7361 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
7362 UNSPEC_VOLATILE_FMFCFG.
7363 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
7364 description for fmfcfg and fmfcsr.
7365 (bdesc_1arg): Add fmtcsr.
7366 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
7367 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
7368 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
7369 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
7370 unspec_fmfcfg): New patterns.
7371 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
7372 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
7373 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
7374 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
7375 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
7376 __nds32__fmfcfg): Define.
7378 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
7380 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
7381 intrinsic register names.
7382 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
7383 intrinsic register enum values and macros.
7385 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
7387 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
7388 for load/store addressing form.
7389 (nds32_print_operand_address): Likewise.
7391 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
7394 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
7395 based on LABEL_REF. Remove useless assertion.
7396 (pic_address_needs_scratch): Fix formatting.
7397 (sparc_legitimize_pic_address): Minor tweaks.
7398 (sparc_delegitimize_address): Adjust assertion accordingly.
7399 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
7400 into symbolic_operand.
7401 (movsi_high_pic_label_ref): Likewise.
7402 (movsi_lo_sum_pic_label_ref): Likewise.
7403 (movdi_pic_label_ref): Likewise.
7404 (movdi_high_pic_label_ref): Likewise.
7405 (movdi_lo_sum_pic_label_ref): Likewise.
7407 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
7409 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
7410 custom LIB_SPEC setup.
7412 2018-04-06 Ruslan Bukin <br@bsdpad.com>
7413 Kito Cheng <kito.cheng@gmail.com>
7415 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
7416 * config/riscv/freebsd.h: New.
7418 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7420 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
7421 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
7424 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7425 Kito Cheng <kito.cheng@gmail.com>
7427 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
7428 nds32_output_call, nds32_symbol_binds_local_p): New functions.
7429 * config/nds32/nds32-protos.h (nds32_output_call,
7430 nds32_output_return): Declare.
7431 * config/nds32/nds32.md: Refine all the call and return patterns.
7433 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7436 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
7437 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
7439 PR rtl-optimization/84872
7440 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
7441 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
7444 2018-04-06 Tamar Christina <tamar.christina@arm.com>
7446 * expr.c (copy_blkmode_to_reg): Revert 254862.
7447 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
7449 2018-04-06 Richard Biener <rguenther@suse.de>
7452 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
7453 after seeing a component reference with an adjacent field. Treat
7454 refs to arrays at struct end of external decls similar to
7455 refs to unconstrained commons.
7457 2018-04-06 Jakub Jelinek <jakub@redhat.com>
7460 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
7461 look through SAVE_EXPRs with non-side-effects argument. Adjust
7463 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
7466 2018-04-06 Richard Biener <rguenther@suse.de>
7469 * alias.c (find_base_term): New wrapper around find_base_term
7470 unwinding CSELIB_VAL_PTR changes.
7471 (find_base_term): Do not restore CSELIB_VAL_PTR during the
7474 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7476 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
7478 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
7479 constant definitions.
7480 ("nop"): lr 0,0 -> nopr r0
7481 ("nop_lr0", "nop_lr1"): New insn definitions.
7483 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
7485 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
7486 NDS32_V3PUSH_AVAILABLE_P macro.
7488 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
7489 Chung-Ju Wu <jasonwucj@gmail.com>
7491 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
7492 (nds32*-*-*): Add float and fpu_config into supported_defaults.
7493 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
7494 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
7495 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
7496 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
7497 * config/nds32/constraints.md: New constraints and checking for hard
7498 float configuration.
7499 * config/nds32/iterators.md: New mode iterator and attribute for hard
7500 float configuration.
7501 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
7503 * config/nds32/nds32-fpu.md: New file.
7504 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
7505 deal with hard float code generation.
7506 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
7508 (abi_type, float_reg_number): New enum type.
7509 * config/nds32/nds32-predicates.c: New predicates for hard float.
7510 * config/nds32/nds32-protos.h: Declare functions for hard float.
7511 * config/nds32/nds32.c: Implementation for hard float configuration.
7512 * config/nds32/nds32.h: Definitions for hard float configuration.
7513 * config/nds32/nds32.md: Include hard float machine description and
7514 modify patterns for hard float configuration.
7515 * config/nds32/nds32.opt: New options for hard float configuration.
7516 * config/nds32/predicates.md: New predicates for hard float
7519 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7521 * common/config/nds32/nds32-common.c
7522 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7524 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7527 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7528 CONSTRUCTOR_ELT (ctor, ...)->value.
7530 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7533 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7535 2018-04-05 Tom de Vries <tom@codesourcery.com>
7538 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7541 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7542 Kito Cheng <kito.cheng@gmail.com>
7544 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7545 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7546 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7547 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7549 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7550 Kito Cheng <kito.cheng@gmail.com>
7552 * config/nds32/constraints.md (Ufe): New memory constraint.
7553 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7554 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7555 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7557 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7558 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7560 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7562 * config/nds32/nds32.md: Use optimize_size in the condition for
7563 alu-shift instructions.
7565 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7567 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7569 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7571 * config/nds32/nds32.md (negsi2): Refine pattern.
7573 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7574 Chung-Ju Wu <jasonwucj@gmail.com>
7576 * config/nds32/iterators.md (shift_rotate): New code iterator.
7577 (shift): New code attribute.
7578 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7579 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7580 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7581 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7582 bit-wise operations.
7583 (andsi3, *andsi3): Ditto.
7584 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7585 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7586 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7587 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7588 nds32_ior_operand, nds32_xor_operand): New predicates.
7590 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7592 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7593 (addsi3, subsi3): ... this.
7595 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7597 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7599 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7601 * config/nds32/nds32.md: Adjust indention.
7603 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7605 * config/nds32/nds32.md (feature): New attribute.
7607 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7609 * config/nds32/nds32.md (subtype): New attribute.
7611 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7614 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7615 expansion to perform a bitwise AND of the argument followed by a
7616 boolean negation of the result.
7618 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7620 PR rtl-optimization/84878
7621 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7622 the basic block. Assert the use reference is not artificial and that
7623 it has an associated insn.
7625 2018-04-04 Michael Matz <matz@suse.de>
7627 * builtins.c (compute_objsize): Pass correct operand
7628 to array_at_struct_end_p.
7630 2018-04-04 Richard Biener <rguenther@suse.de>
7633 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7634 from contexts for DINFO_LEVEL_TERSE and below.
7636 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7638 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7639 resiter_operand condition.
7640 * config/nds32/nds32.md (*move<mode>): Ditto.
7642 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7643 Monk Chiang <sh.chiang04@gmail.com>
7645 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7647 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7649 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7651 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7652 Kito Cheng <kito.cheng@gmail.com>
7654 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7655 nds32_cond_code_str, output_cond_branch,
7656 output_cond_branch_compare_zero, nds32_expand_cbranch,
7657 nds32_expand_cstore, nds32_expand_movcc,
7658 nds32_output_cbranchsi4_equality_zero,
7659 nds32_output_cbranchsi4_equality_reg,
7660 nds32_output_cbranchsi4_equality_reg_or_const_int,
7661 nds32_output_cbranchsi4_greater_less_zero: New functions.
7662 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7663 nds32_expand_cstore, nds32_expand_movcc,
7664 nds32_output_cbranchsi4_equality_zero,
7665 nds32_output_cbranchsi4_equality_reg,
7666 nds32_output_cbranchsi4_equality_reg_or_const_int,
7667 nds32_output_cbranchsi4_greater_less_zero): Declare.
7668 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7669 nds32_rimm11s_operand): New predicates.
7670 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7671 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7674 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7676 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7677 * config/nds32/nds32.md: Ditto.
7678 * config/nds32/pipelines.md: Ditto.
7680 2018-04-04 Richard Biener <rguenther@suse.de>
7682 PR tree-optimization/85168
7683 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7684 propagating abnormals.
7686 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7688 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7690 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7691 Kito Cheng <kito.cheng@gmail.com>
7693 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7694 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7695 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7696 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7697 * config/nds32/nds32.md (sibcall_internal): New.
7698 (sibcall_register): Remove.
7699 (sibcall_immediate): Remove.
7700 (sibcall_value_internal): New.
7701 (sibcall_value_register): Remove.
7702 (sibcall_value_immediate): Remove.
7703 * config/nds32/predicates.md (nds32_general_register_operand): New.
7704 (nds32_call_address_operand): New.
7706 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7708 PR rtl-optimization/85167
7709 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7710 bb_defs if *split_p, instead preinitialize it to NULL.
7712 PR tree-optimization/85156
7713 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7714 evaluating the argument multiple times.
7716 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7718 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7720 (_mm_cvtpd_ps): Likewise.
7721 (_mm_cvttpd_epi32): Likewise.
7722 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7723 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7724 vector, pixel, and bool following altivec.h include.
7726 2018-04-03 Martin Sebor <msebor@redhat.com>
7728 * doc/extend.texi (Common Function Attributes): Clarify.
7729 (const attribute): Likewise.
7730 (pure attribute): Likewise.
7732 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7735 * config/i386/i386.c (ix86_expand_vector_set): Use
7736 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7738 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7740 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7741 instructions when changing rounding bits to preserve precision bits
7742 in the x87 control word.
7744 2018-04-03 Martin Liska <mliska@suse.cz>
7746 PR tree-optimization/82491
7747 * rtl.h (strip_offset_and_add): Replace += suboffset with
7748 poly_uint64 () + suboffset.
7750 2018-03-29 Martin Liska <mliska@suse.cz>
7751 Martin Jambor <mjambor@suse.cz>
7754 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7755 param_type is not an integral or pointer type.
7757 2018-04-03 Richard Biener <rguenther@suse.de>
7759 * sese.h (recompute_all_dominators): Remove.
7761 2018-04-02 Martin Sebor <msebor@redhat.com>
7763 * doc/invoke.texi (-Wrestrict): Fix typos.
7765 2018-04-02 Jim Wilson <jimw@sifive.com>
7767 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7768 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7769 (<optab>di3, <optab>si3_extend): Likewise.
7770 (<optab>si3_mask, <optab>si3_mask_1): New.
7771 (<optab>di3_mask, <optab>di3_mask_1): New.
7772 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7773 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7774 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7776 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7778 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7781 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7783 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7784 (nds32_canonicalize_comparison): New function.
7786 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7787 Kito Cheng <kito.cheng@gmail.com>
7788 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7790 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7791 * config/nds32/constants.md (unspec_volatile_element): Add
7792 UNSPEC_VOLATILE_RELAX_GROUP.
7793 * config/nds32/nds32-relax-opt.c: New file.
7794 * config/nds32/nds32-predicates.c
7795 (nds32_symbol_load_store_p): New function.
7796 * config/nds32/nds32-protos.h
7797 (nds32_symbol_load_store_p): Declare function.
7798 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7799 * config/nds32/nds32.c
7800 (nds32_register_pass): New function to register pass.
7801 (nds32_register_passes): New function to register passes.
7802 * config/nds32/nds32.md (relax_group): New pattern.
7803 * config/nds32/nds32.opt (mrelax-hint): New option.
7804 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7806 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7808 * config/nds32/t-nds32: Modify files dependency.
7810 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7812 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7813 (PROFILE_HOOK): Define its implementation.
7815 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7817 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7818 type and 32-bit size.
7820 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7823 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7824 (V_128_256): New mode iterator.
7825 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7826 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7827 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7829 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7830 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7832 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7835 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7836 NaN inputs correctly.
7838 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7841 * config/rs6000/vsx.md (??r): New mode attribute.
7842 (*vsx_mov<mode>_64bit): Use it.
7843 (*vsx_mov<mode>_32bit): Likewise.
7845 2018-03-30 Martin Sebor <msebor@redhat.com>
7847 PR tree-optimization/84818
7848 * builtins.c (check_access): Use warning_n.
7850 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7853 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7855 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7858 2018-03-30 Julia Koval <julia.koval@intel.com>
7861 * x86-tune.def (movx, partial_reg_dependency): Enable for
7864 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7867 * lra-constraints.c (process_alt_operands): Move setting
7868 this_alternative_matches below.
7870 2018-03-29 Martin Liska <mliska@suse.cz>
7873 * doc/invoke.texi: Document how LTO works with debug info.
7874 Describe auto-load support of binutils. Mention 'x86-64'
7875 as valid option value of -march option.
7877 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7879 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7882 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7883 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7884 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7887 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7890 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7891 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7892 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7893 for RS6000_BTM_POWERPC64.
7894 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7895 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7896 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7901 2018-03-28 Carl Love <cel@us.ibm.com>
7904 2017-09-27 Carl Love <cel@us.ibm.com>
7906 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7907 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7908 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7911 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7913 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7914 instead of __vector bool.
7915 (_mm_max_pu8): Likewise.
7916 (_mm_min_pi16): Likewise.
7918 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7921 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7922 (DIVWEUO): Likewise.
7924 (DIVDEUO): Likewise.
7925 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7926 DIVWEUO and DIVDEUO.
7927 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7928 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7929 (div_extend): Likewise.
7930 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7932 (__builtin_divweuo): Likewise.
7933 (__builtin_divdeo): Likewise.
7934 (__builtin_divdeuo): Likewise.
7936 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7939 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7940 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7942 PR tree-optimization/82004
7943 * gimple-match-head.c (optimize_pow_to_exp): New function.
7944 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7945 Don't fold to exp if optimize_pow_to_exp is false.
7947 2018-03-28 Martin Liska <mliska@suse.cz>
7950 * calls.c (initialize_argument_information): Fix trailing space.
7951 * common.opt: Fix typo and provide better explanation for
7952 -fsanitize-coverage option.
7953 * config/i386/i386.opt: Fix typo.
7955 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7956 Martin Liska <mliska@suse.cz>
7959 * gimplify.c (asan_poison_variable): Don't do the check for
7960 gimplify_omp_ctxp here.
7961 (gimplify_decl_expr): Do it here.
7962 (gimplify_target_expr): Likewise.
7964 2018-03-28 Martin Liska <mliska@suse.cz>
7967 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7968 chkp_type_bounds_count if MPX is not enabled.
7970 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7972 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7974 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7977 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7978 function to create the function decl for complex long double
7979 multiply and divide for -mabi=ieeelongdouble.
7980 (init_float128_ieee): Call it.
7982 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7985 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7986 -fcf-protection=branch -mibt.
7987 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7989 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7992 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7993 arm_word_relocations.
7995 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7998 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7999 extern array declarations.
8001 2018-03-27 Richard Biener <rguenther@suse.de>
8004 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
8005 explicit single_use checks.
8007 2018-03-27 Richard Biener <rguenther@suse.de>
8009 PR tree-optimization/85082
8010 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
8013 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8015 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
8016 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
8017 Turn on fasynchronous-unwind-tables and funwind-tables.
8019 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
8022 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
8023 (*bmi_blsr_<mode>_ccz): Ditto.
8025 2018-03-26 Tom de Vries <tom@codesourcery.com>
8027 PR tree-optimization/85063
8028 * omp-general.c (offloading_function_p): New function. Factor out
8030 * omp-offload.c (pass_omp_target_link::gate): ... here.
8031 * omp-general.h (offloading_function_p): Declare.
8032 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
8033 with attribute omp declare target for offloading functions.
8035 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
8037 PR tree-optimization/84005
8038 * tree-data-ref.h (get_base_for_alignment): Declare.
8039 * tree-data-ref.c (get_base_for_alignment_1): New function.
8040 (get_base_for_alignment): Likewise.
8041 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
8042 get_base_for_alignment to find a suitable base object, instead
8043 of always using drb->base_address.
8045 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8048 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
8049 known size by default.
8051 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
8054 * lra-constraints.c (process_alt_operands): Don't match BLKmode
8055 and non BLKmode operands.
8057 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8060 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
8061 Clean up attributes.
8063 2018-03-23 Richard Biener <rguenther@suse.de>
8066 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
8067 we are going to emit early debug for LTO.
8069 2018-03-23 Jakub Jelinek <jakub@redhat.com>
8072 * function.c (match_asm_constraints_1): Don't optimize if input
8073 doesn't satisfy general_operand predicate for output's mode.
8076 * alias.c (write_dependence_p): Don't require for x_canonicalized
8077 non-VOIDmode if x has VOIDmode.
8080 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
8081 just don't try to optimize it rather than assert it never happens.
8083 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8085 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
8086 macro expansions for definition of ST_INTERNAL_<mode> and
8087 LD_INTERNAL_<mode> builtins.
8088 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
8090 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
8092 (altivec_expand_st_builtin): Likewise.
8093 (altivec_expand_builtin): Remove calls to deleted functions.
8094 (rs6000_address_for_altivec): Delete this function.
8095 * config/rs6000/vector.md: Remove expands for
8096 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
8098 2018-03-22 Sudakshina Das <sudi.das@arm.com>
8101 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
8102 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
8103 re-computing once computed.
8104 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
8105 (arm_init_machine_status): Initialize
8106 machine->static_chain_stack_bytes.
8108 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8111 * doc/extend.texi: Add four new prototypes for vec_ld.
8112 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
8113 definitions for more logical presentation.
8114 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
8115 entries for V1TI variants of __builtin_altivec_ld builtin.
8116 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
8117 handling of V1TI variant of LVX icode pattern.
8118 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
8119 (rs6000_gimple_fold_builtin): Likewise.
8120 (altivec_init_builtins): Add code to define
8121 __builtin_altivec_lvx_v1ti function.
8123 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8126 * function.c (match_asm_constraints_1): Don't do the optimization
8127 if input isn't a REG, SUBREG, MEM or constant.
8129 2018-03-22 Tom de Vries <tom@codesourcery.com>
8131 PR tree-optimization/84956
8132 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
8133 bb_has_abnormal_pred.
8135 2018-03-22 Jakub Jelinek <jakub@redhat.com>
8138 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
8139 DECL_INITIAL (decl) to decl at the end.
8140 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
8143 2018-03-21 Joseph Myers <joseph@codesourcery.com>
8145 * doc/extend.texi (__builtin_tgmath): Document when complex
8146 integer types are treated as _Complex _Float64.
8148 2018-03-21 Tom de Vries <tom@codesourcery.com>
8150 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
8152 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8154 PR tree-optimization/84960
8155 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
8156 if it is ENTRY block, move them into single succ of ENTRY in that case.
8158 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
8160 PR tree-optimization/84811
8161 * poly-int.h (poly_span_traits): Remove the T3 parameter and
8162 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
8163 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
8164 (known_subrange_p): Update accordingly. Cast each value involved
8165 in the size comparison, rather than casting the result of the
8168 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8170 PR tree-optimization/84982
8171 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
8172 by flipping the least significant bit rather than all bits from
8173 bitpos to bitpos + bitsize - 1.
8175 2018-03-21 Nathan Sidwell <nathan@acm.org>
8177 * doc/extend.texi (Deprecated Features): Remove mention of
8178 long-deleted deprecations.
8180 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8183 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
8184 * configure: Regenerate.
8186 2018-03-21 Tom de Vries <tom@codesourcery.com>
8188 PR tree-optimization/83126
8189 * tree-parloops.c (num_phis): New function.
8190 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
8192 2018-03-21 Nathan Sidwell <nathan@acm.org>
8194 * doc/extend.texi (Deprecated Features): Update deprecated flags,
8195 mention anon-struct/union members and trailing attributes.
8197 2018-03-21 Bin Cheng <bin.cheng@arm.com>
8199 PR tree-optimization/84969
8200 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
8201 builtin memset partitions if they set different rhs values.
8203 2018-03-21 Jakub Jelinek <jakub@redhat.com>
8205 PR rtl-optimization/84989
8206 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
8207 VEC_DUPLICATE with scalar result mode.
8209 2018-03-21 Martin Liska <mliska@suse.cz>
8212 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
8213 not intended return statement.
8215 2018-03-21 Martin Liska <mliska@suse.cz>
8218 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
8219 (chkp_find_bound_slots_1): Limit number of iterations.
8221 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
8224 * Minor grammar fixes for x86 options.
8226 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8229 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
8230 holding REG_CFA_RESTORE notes, instead turn them into a USE.
8232 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
8235 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
8236 (altivec_lvx_<mode>_1op): Likewise.
8237 (altivec_stvx_<mode>_2op): Likewise.
8238 (altivec_stvx_<mode>_1op): Likewise.
8239 (altivec_lvx_<VM2:mode>): New define_expand.
8240 (altivec_stvx_<VM2:mode>): Likewise.
8241 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
8242 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8243 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
8244 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
8245 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
8246 (rs6000_gen_lvx): Likewise.
8247 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
8248 (altivec_expand_stv_builtin): Likewise.
8249 (altivec_expand_builtin): Likewise.
8250 * config/rs6000/vector.md: Likewise.
8252 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8255 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
8258 2018-03-20 Richard Biener <rguenther@suse.de>
8261 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
8262 sign-conversions as zero, fall back to standard scalar_stmt
8265 2018-03-20 Martin Liska <mliska@suse.cz>
8268 * predict.c (rebuild_frequencies): Handle case when we have
8269 PROFILE_ABSENT, but flag_guess_branch_prob is false.
8271 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8274 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
8275 flag_section_anchors.
8276 * varasm.c (use_blocks_for_decl_p): Remove hack for
8277 dw2_force_const_mem.
8280 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
8282 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
8283 be created, use lowpart_subreg of operands[0] rather than operands[0]
8285 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
8286 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
8287 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
8288 and n constraint instead of aarch64_shift_imm_di and Usd.
8289 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
8290 (*aarch64_<optab>_reg_minus<mode>3): ... this.
8292 2018-03-20 Sudakshina Das <sudi.das@arm.com>
8295 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
8296 to favor GPR over NEON registers.
8297 (<shift>di3_neon): Likewise.
8299 2018-03-20 Tom de Vries <tom@codesourcery.com>
8302 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
8303 (nvptx_process_pars): Emit bar.sync asap and alap.
8305 2018-03-20 Tom de Vries <tom@codesourcery.com>
8308 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
8309 seen_label if seen_label is already set.
8311 2018-03-20 Jakub Jelinek <jakub@redhat.com>
8314 * config/i386/i386.c (fold_builtin_cpu): For features above 31
8315 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
8316 Use 1U instead of 1. Formatting fixes.
8319 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
8320 instead of TREE_TYPE (s1) for the return value.
8322 2018-03-19 Jakub Jelinek <jakub@redhat.com>
8324 PR tree-optimization/84946
8325 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
8326 bitsize + bitsize in poly_uint64 rather than poly_int64.
8329 * dwarf2asm.c: Include fold-const.c.
8330 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
8331 of decl rather than decl itself.
8333 PR rtl-optimization/84643
8334 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
8336 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
8339 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
8340 calling assemble_variable.
8342 2018-03-19 Sudakshina Das <sudi.das@arm.com>
8345 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
8346 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
8348 2018-03-19 Jim Wilson <jimw@sifive.com>
8351 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
8352 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
8353 (riscv_first_stack_step): Likewise.
8354 (riscv_option_override): Use STACK_BOUNDARY instead of
8356 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
8358 (BIGGEST_ALIGNMENT): Set to 128.
8359 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
8360 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
8363 2018-03-19 Richard Biener <rguenther@suse.de>
8365 PR tree-optimization/84933
8366 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
8367 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
8369 2018-03-19 Richard Biener <rguenther@suse.de>
8371 PR tree-optimization/84859
8372 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
8373 (cond_if_else_store_replacement): Perform sinking operation on
8374 single-store BBs regardless of MAX_STORES_TO_SINK setting.
8375 Generalize what a BB with a single eligible store is.
8377 2018-03-19 Richard Biener <rguenther@suse.de>
8379 PR tree-optimization/84929
8380 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
8381 chrec_is_positive against non-chrec arg.
8383 2018-03-19 Tamar Christina <tamar.christina@arm.com>
8386 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
8388 2018-03-18 Martin Liska <mliska@suse.cz>
8390 PR rtl-optimization/84635
8391 * regrename.c (build_def_use): Use matches_mode only when
8394 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
8396 PR tree-optimization/84913
8397 * tree-vect-loop.c (vectorizable_reduction): Don't try to
8398 vectorize chains of COND_EXPRs.
8400 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8402 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
8404 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8406 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
8408 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
8410 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
8412 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8413 Kito Cheng <kito.cheng@gmail.com>
8415 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
8416 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
8417 (nds32_adjust_reg_alloc_order): New function.
8418 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
8420 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
8422 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
8423 nds32_print_operand, nds32_print_operand_address): Use
8424 HOST_WIDE_INT_PRINT_DEC instead.
8426 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
8428 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
8430 2018-03-17 Jakub Jelinek <jakub@redhat.com>
8433 * config/i386/i386.c (initial_ix86_tune_features,
8434 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
8436 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
8437 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
8438 rather than 1u << ix86_tune. Formatting fix.
8439 (ix86_option_override_internal): Change ix86_arch_mask from
8440 unsigned int to unsigned HOST_WIDE_INT, initialize to
8441 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
8442 (ix86_function_specific_restore): Likewise.
8444 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8447 * postreload.c (reload_combine_recognize_pattern): Perform
8448 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
8449 truncate_int_for_mode the result for the destination's mode.
8452 * hsa-gen.c (mem_type_for_type): Fix comment typo.
8453 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
8455 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8458 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
8461 * lra-assigns.c (lra_split_hard_reg_for): Don't use
8462 regno_allocno_class_array and sorted_pseudos.
8463 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
8464 insns where regno is used.
8466 2018-03-16 Martin Liska <mliska@suse.cz>
8469 * multiple_target.c (create_dispatcher_calls): Redirect
8470 reference in the symbol table.
8472 2018-03-16 Martin Liska <mliska@suse.cz>
8475 * multiple_target.c (create_dispatcher_calls): Redirect also
8478 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8482 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
8483 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
8484 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
8486 2018-03-16 Julia Koval <julia.koval@intel.com>
8488 * doc/invoke.texi (Skylake Server): Add CLWB.
8489 Cannonlake): Remove CLWB.
8491 2018-03-16 Jakub Jelinek <jakub@redhat.com>
8493 PR tree-optimization/84841
8494 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
8496 (FLOAT_ONE_CONST_TYPE): Define.
8497 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
8498 (sort_by_operand_rank): Put entries with higher constant_type last
8499 rather than first to match comments.
8501 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
8503 * config/nios2/nios2.md (movsi_internal): Fix thinko in
8506 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8509 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
8510 check and use address of target always.
8512 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
8515 * config/i386/i386.c (indirect_thunk_needed): Update comments.
8516 (indirect_thunk_bnd_needed): Likewise.
8517 (indirect_thunks_used): Likewise.
8518 (indirect_thunks_bnd_used): Likewise.
8519 (indirect_return_needed): New.
8520 (indirect_return_bnd_needed): Likewise.
8521 (output_indirect_thunk_function): Add a bool argument for
8523 (output_indirect_thunk_function): Don't generate alias for
8524 function return thunk.
8525 (ix86_code_end): Call output_indirect_thunk_function to generate
8526 function return thunks.
8527 (ix86_output_function_return): Set indirect_return_bnd_needed
8528 and indirect_return_needed instead of indirect_thunk_bnd_needed
8529 and indirect_thunk_needed.
8531 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8533 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8534 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8535 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8537 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8538 Paul Hua <paul.hua.gm@gmail.com>
8541 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8543 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8545 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8546 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8549 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8552 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8553 instead of GET_MODE_SIZE when comparing Units.
8555 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8558 * varasm.c (hash_section): Return an unchangeble hash value
8559 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8560 Return !aarch64_can_use_per_function_literal_pools_p ().
8562 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8565 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8566 rather than address of cmode as last argument to prepare_cmp_insn.
8568 2018-03-15 Julia Koval <julia.koval@intel.com>
8570 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8571 F_AVX512VNNI, F_AVX512BITALG): New.
8573 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8576 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8577 insn for floating-point loads and stores.
8579 2018-03-14 Carl Love <cel@us.ibm.com>
8581 * config/rs6000/rs6000-c.c: Add macro definitions for
8582 ALTIVEC_BUILTIN_VEC_PERMXOR.
8583 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8584 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8585 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8586 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8588 * config/doc/extend.texi: Add prototypes for vec_permxor.
8590 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8593 * diagnostic-show-locus.c (class layout_point): Convert m_line
8594 from int to linenum_type.
8595 (line_span::comparator): Use linenum "compare" function when
8596 comparing line numbers.
8597 (test_line_span): New function.
8598 (layout_range::contains_point): Convert param "row" from int to
8600 (layout_range::intersects_line_p): Likewise.
8601 (layout::will_show_line_p): Likewise.
8602 (layout::print_source_line): Likewise.
8603 (layout::should_print_annotation_line_p): Likewise.
8604 (layout::print_annotation_line): Likewise.
8605 (layout::print_leading_fixits): Likewise.
8606 (layout::annotation_line_showed_range_p): Likewise.
8607 (struct line_corrections): Likewise for field m_row.
8608 (line_corrections::line_corrections): Likewise for param "row".
8609 (layout::print_trailing_fixits): Likewise.
8610 (layout::get_state_at_point): Likewise.
8611 (layout::get_x_bound_for_row): Likewise.
8612 (layout::print_line): Likewise.
8613 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8614 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8615 * input.c (selftest::test_linenum_comparisons): New function.
8616 (selftest::input_c_tests): Call it.
8617 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8618 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8619 * selftest.h (ASSERT_GT): New macro.
8620 (ASSERT_GT_AT): New macro.
8621 (ASSERT_LT): New macro.
8622 (ASSERT_LT_AT): New macro.
8624 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8626 PR rtl-optimization/84780
8627 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8629 2018-03-14 Martin Liska <mliska@suse.cz>
8631 * tree.c (record_node_allocation_statistics): Use
8632 get_stats_node_kind.
8633 (get_stats_node_kind): New function extracted from
8634 record_node_allocation_statistics.
8635 (free_node): Use get_stats_node_kind.
8637 2018-03-14 Richard Biener <rguenther@suse.de>
8639 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8640 that the value-set of ANTIC_IN doesn't grow.
8643 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8645 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8646 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8648 2018-03-14 Julia Koval <julia.koval@intel.com>
8650 * config.gcc (icelake-client, icelake-server): New.
8652 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8653 (initial_ix86_arch_features): Ditto.
8654 (PTA_SKYLAKE): Add SGX.
8655 (PTA_ICELAKE): Remove.
8656 (PTA_ICELAKE_CLIENT): New.
8657 (PTA_ICELAKE_SERVER): New.
8658 (ix86_option_override_internal): Split up icelake on icelake client and
8660 (get_builtin_code_for_version): Ditto.
8661 (fold_builtin_cpu): Ditto.
8662 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8663 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8664 * config/i386/i386.h (processor_type): Ditto.
8665 * doc/invoke.texi: Ditto.
8667 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8670 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8671 INTEGER_CST offset, add it together with bitpos / 8 and
8672 sign extend based on POINTER_SIZE.
8676 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8679 * config/i386/constraints.md (Yc): New register constraint.
8680 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8681 Use Yc constraint for alternative 2 of operand 0. Remove
8682 preferred_for_speed attribute.
8684 2018-03-14 Richard Biener <rguenther@suse.de>
8686 PR tree-optimization/84830
8687 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8688 with the old one to avoid oscillations.
8690 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8693 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8695 (assign_by_spills): Return a flag of reload assignment failure.
8696 Do not process the reload assignment failures. Do not spill other
8697 reload pseudos if they has the same reg class. Update n if
8699 (lra_assign): Add a return arg. Set up from the result of
8700 assign_by_spills call.
8701 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8702 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8703 usage_insns if it is not NULL.
8704 (spill_hard_reg_in_range): New function.
8705 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8706 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8707 function prototypes.
8708 (lra_assign): Change prototype.
8709 * lra.c (lra): Add code to deal with fails by splitting hard reg
8712 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8714 * config/riscv/riscv.opt (mrelax): New option.
8715 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8716 "norelax" when riscv_mrelax is disabled.
8717 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8719 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8722 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8723 reassociation for int modes.
8725 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8727 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8728 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8730 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8731 * config/aarch64/aarch64-sve.md
8732 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8733 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8734 (*extend<mode><Vwide>2): Rename to...
8735 (aarch64_sve_extend<mode><Vwide>2): ...this.
8736 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8737 renaming the old pattern to...
8738 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8740 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8741 define_expand, renaming the old pattern to...
8742 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8743 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8744 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8745 account when deciding which SVE instruction the optab should use.
8746 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8748 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8750 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8751 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8752 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8753 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8754 (tlsdesc_small_<mode>): Turn a define_expand and use
8755 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8756 (tlsdesc_small_advsimd_<mode>): ...this.
8757 (tlsdesc_small_sve_<mode>): New pattern.
8759 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8761 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8762 (UNSPEC_UMUL_HIGHPART): New constants.
8763 (MUL_HIGHPART): New int iteraor.
8764 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8765 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8767 (*<su>mul<mode>3_highpart): New define_insn.
8769 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8772 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8775 2018-03-13 Martin Liska <mliska@suse.cz>
8778 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8780 (sem_item_optimizer::~sem_item_optimizer): Release it.
8781 (sem_item_optimizer::merge_classes): Register variable aliases.
8782 (sem_item_optimizer::fixup_pt_set): New function.
8783 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8784 * ipa-icf.h: Declare new variables and functions.
8786 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8789 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8790 integer_pow2p@2 and test integer_pow2p in condition.
8791 (A < 0 ? C : 0): Similarly for @1.
8794 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8795 characters starting at p contain '\0' character, don't look beyond
8799 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8800 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8803 * reg-stack.c (change_stack): Change update_end var from int to
8804 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8805 also call set_block_for_insn on the newly added insns and rescan.
8808 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8809 on the last operand.
8812 * tree.c (stabilize_reference_1): Return save_expr (e) for
8813 STATEMENT_LIST even if it doesn't have side-effects.
8815 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8817 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8819 2018-03-12 Renlin Li <renlin.li@arm.com>
8821 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8822 aarch64_output_scalar_simd_mov_immediate.
8824 2018-03-12 Martin Sebor <msebor@redhat.com>
8826 PR tree-optimization/83456
8827 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8828 for perfectly overlapping calls to memcpy.
8829 (gimple_fold_builtin_memory_chk): Same.
8830 (gimple_fold_builtin_strcpy): Handle no-warning.
8831 (gimple_fold_builtin_stxcpy_chk): Same.
8832 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8834 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8836 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8837 parameter. Use it for SFmode.
8838 (rs6000_function_arg_advance_1): Adjust.
8839 (rs6000_function_arg): Adjust.
8840 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8842 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8844 PR rtl-optimization/84169
8845 PR rtl-optimization/84780
8846 * combine.c (can_combine_p): Check for a 2-insn combination whether
8847 the destination register is used between the two insns, too.
8849 2018-03-12 Richard Biener <rguenther@suse.de>
8851 PR tree-optimization/84803
8852 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8853 for refs DR analysis didn't process.
8855 2018-03-12 Richard Biener <rguenther@suse.de>
8857 PR tree-optimization/84777
8858 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8859 force-vectorize loops ignore whether we are optimizing for size.
8861 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8863 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8864 (TARGET_MD_ASM_ADJUST): Define.
8866 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8867 Kito Cheng <kito.cheng@gmail.com>
8868 Chung-Ju Wu <jasonwucj@gmail.com>
8870 * config/nds32/nds32.c (nds32_compute_stack_frame,
8871 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8872 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8873 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8874 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8875 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8876 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8877 * config/nds32/nds32.md (prologue, epilogue): Use macro
8878 NDS32_V3PUSH_AVAILABLE_P to do checking.
8880 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8883 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8884 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8885 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8886 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8887 addition of most attributes on !orig_type_die or the attribute not
8888 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8890 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8891 Chung-Ju Wu <jasonwucj@gmail.com>
8893 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8895 * config/nds32/nds32.opt (mvh): New option.
8897 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8898 Chung-Ju Wu <jasonwucj@gmail.com>
8900 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8902 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8903 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8906 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8907 Chung-Ju Wu <jasonwucj@gmail.com>
8909 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8911 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8912 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8914 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8915 Kito Cheng <kito.cheng@gmail.com>
8916 Chung-Ju Wu <jasonwucj@gmail.com>
8918 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8919 UNSPEC_FFMISM and UNSPEC_FLMISM.
8920 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8921 for ffb, ffmism and flmism.
8922 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8923 (unspec_ffmism): Ditto.
8924 (unspec_flmism): Ditto.
8925 (nds32_expand_builtin_impl): Check if string extension is available.
8926 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8927 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8929 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8932 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8935 * lra-assigns.c (assign_by_spills): Return a flag of reload
8936 assignment failure. Do not process the reload assignment
8937 failures. Do not spill other reload pseudos if they has the same
8939 (lra_assign): Add a return arg. Set up from the result of
8940 assign_by_spills call.
8941 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8942 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8943 usage_insns if it is not NULL.
8944 (spill_hard_reg_in_range): New function.
8945 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8946 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8947 function prototypes.
8948 (lra_assign): Change prototype.
8949 * lra.c (lra): Add code to deal with fails by splitting hard reg
8952 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8955 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8957 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8960 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8961 (dw_val_node): Add val_symbolic_view.
8962 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8963 (symview_upper_bound): New.
8964 (new_line_info_table): Initialize symviews_since_reset.
8965 (dwarf2out_source_line): Count symviews_since_reset and set
8966 symview_upper_bound.
8967 (dw_val_equal_p): Handle symview.
8968 (add_AT_symview): New.
8969 (print_dw_val): Handle symview.
8970 (attr_checksum, attr_checksum_ordered): Likewise.
8971 (same_dw_val_p, size_of_die): Likewise.
8972 (value_format, output_die): Likewise.
8973 (add_high_low_attributes): Use add_AT_symview for entry_view.
8974 (dwarf2out_finish): Reset symview_upper_bound, clear
8977 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8980 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8981 Add strict argument and use it.
8982 (rs6000_split_multireg_move): Update for new strict argument.
8983 (mem_operand_gpr): Disallow all non-offsettable addresses.
8984 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8986 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8989 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8990 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8991 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8994 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8995 decl, use remap_type if we want to use the type.
8997 2018-03-09 Martin Sebor <msebor@redhat.com>
8999 PR tree-optimization/84526
9000 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
9002 (builtin_access::generic_overlap): Be prepared to handle non-array
9005 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
9007 PR rtl-optimization/84682
9008 * lra-constraints.c (process_address_1): Check is_address flag
9009 for address constraints.
9010 (process_alt_operands): Likewise.
9011 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
9012 preprocess_constraints.
9013 * recog.h (preprocess_constraints): Add oploc parameter.
9015 * recog.c (preprocess_constraints): Test address_operand for
9016 CT_ADDRESS constraints.
9018 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
9021 * lra-assigns.c (assign_by_spills): Return a flag of reload
9022 assignment failure. Do not process the reload assignment
9023 failures. Do not spill other reload pseudos if they has the same
9025 (lra_assign): Add a return arg. Set up from the result of
9026 assign_by_spills call.
9027 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
9028 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
9029 usage_insns if it is not NULL.
9030 (spill_hard_reg_in_range): New function.
9031 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
9032 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
9033 function prototypes.
9034 (lra_assign): Change prototype.
9035 * lra.c (lra): Add code to deal with fails by splitting hard reg
9038 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9041 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
9042 Accept complain bool parameter. Only emit errors if it is true.
9043 (arm_parse_cpu_option_name): Likewise.
9044 (arm_target_thumb_only): Adjust callers of the above.
9045 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
9046 prototype to take a default true bool parameter.
9047 (arm_parse_arch_option_name): Likewise.
9049 2018-03-09 David Malcolm <dmalcolm@redhat.com>
9050 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
9054 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
9055 * configure: Regenerate.
9056 * configure.ac ("linker --version-script option"): New.
9057 ("linker soname option"): New.
9059 2018-03-09 Richard Biener <rguenther@suse.de>
9061 PR tree-optimization/84775
9062 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
9063 immediate uses of predicate stmts and mark them modified.
9066 PR tree-optimization/84178
9067 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9069 (version_loop_for_if_conversion): Delay update_ssa call.
9070 (tree_if_conversion): Delay update_ssa until after predicate
9073 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
9076 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
9077 when the function accesses prior frames.
9079 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9082 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
9083 gen_llsym, otherwise call maybe_gen_llsym.
9086 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9087 has ',' character inside of it.
9089 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9092 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
9093 as clobbering CC_REGNUM.
9095 2018-03-08 Richard Biener <rguenther@suse.de>
9098 * tree-scalar-evolution.c: Include tree-into-ssa.h.
9099 (follow_copies_to_constant): Do not follow SSA names registered
9102 2018-03-08 Richard Biener <rguenther@suse.de>
9104 PR tree-optimization/84178
9105 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
9107 (version_loop_for_if_conversion): Delay update_ssa call.
9108 (tree_if_conversion): Delay update_ssa until after predicate
9111 2018-03-08 David Malcolm <dmalcolm@redhat.com>
9113 PR tree-optimization/84178
9114 * tree-if-conv.c (release_bb_predicate): Remove the
9115 the assertion that the stmts have NULL use_ops.
9116 Discard the statements, asserting that they haven't
9117 yet been added to a BB.
9119 2018-03-08 Richard Biener <rguenther@suse.de>
9121 PR tree-optimization/84746
9122 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
9123 (phi_translate): Pass in destination ANTIC_OUT set.
9124 (phi_translate_1): Likewise. For a simplified result lookup
9125 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
9126 (phi_translate_set): Adjust.
9127 (do_pre_regular_insertion): Likewise.
9128 (do_pre_partial_partial_insertion): Likewise.
9130 2018-03-08 Martin Liska <mliska@suse.cz>
9132 PR gcov-profile/84735
9133 * doc/gcov.texi: Document usage of profile files.
9134 * gcov-io.h: Document changes in the format.
9136 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
9140 * dwarf2out.c (struct dw_line_info_table): Update comments for
9142 (FORCE_RESET_NEXT_VIEW): New.
9143 (FORCE_RESETTING_VIEW_P): New.
9144 (RESETTING_VIEW_P): Check for -1 too.
9145 (ZERO_VIEW_P): Likewise.
9146 (new_line_info_table): Force-reset next view.
9147 (dwarf2out_begin_function): Likewise.
9148 (dwarf2out_source_line): Simplify zero_view_p initialization.
9149 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
9150 view directly. Omit view when omitting .loc at line 0.
9152 2018-03-08 Jakub Jelinek <jakub@redhat.com>
9154 PR tree-optimization/84740
9155 * tree-switch-conversion.c (process_switch): Call build_constructors
9156 only if info.phi_count is non-zero.
9158 PR tree-optimization/84739
9159 * tree-tailcall.c (find_tail_calls): Check call arguments against
9160 DECL_ARGUMENTS (current_function_decl) rather than
9161 DECL_ARGUMENTS (func) when checking for tail recursion.
9163 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9165 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
9166 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
9167 Volker Reichelt's entry and add entries for people that perform
9168 GCC fuzzy testing and report numerous bugs.
9170 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
9173 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
9174 readonly data in sdata, if that is disabled.
9175 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
9176 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
9177 -mreadonly-in-sdata option.
9179 2018-03-07 Martin Sebor <msebor@redhat.com>
9181 PR tree-optimization/84468
9182 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
9183 basic block when looking for nul assignment.
9185 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
9188 * except.h (output_function_exception_table): Adjust prototype.
9189 * except.c (output_function_exception_table): Remove FNNAME parameter
9190 and add SECTION parameter. Ouput one part of the table at a time.
9191 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
9192 the first part of the exception table and emit unwind directives.
9193 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
9194 (i386_pe_seh_cold_init): Likewise.
9195 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
9196 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
9197 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
9198 (ix86_output_call_insn): Emit a nop in one more case for SEH.
9199 * config/i386/winnt.c: Include except.h.
9200 (struct seh_frame_state): Add reg_offset, after_prologue and
9201 in_cold_section fields.
9202 (i386_pe_seh_end_prologue): Set seh->after_prologue.
9203 (i386_pe_seh_cold_init): New function.
9204 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
9205 to seh->in_cold_section.
9206 (seh_emit_push): Record the offset of the push.
9207 (seh_emit_save): Record the offet of the save.
9208 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
9209 Test seh->after_prologue to disregard the epilogue.
9210 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
9211 (i386_pe_end_cold_function): New function.
9213 2018-03-07 Jakub Jelinek <jakub@redhat.com>
9216 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
9217 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
9220 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
9222 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
9223 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
9226 * multiple_target.c: Include tree-inline.h and intl.h.
9227 (expand_target_clones): Diagnose and fail if node->definition and
9228 !tree_versionable_function_p (node->decl).
9230 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
9232 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
9234 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
9235 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
9236 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
9238 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9241 * combine.c (try_combine): Use reg_or_subregno instead of handling
9242 just paradoxical SUBREGs and REGs.
9244 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
9246 * config/arc/arc.c (arc_finalize_pic): Remove function.
9247 (arc_must_save_register): We use single base PIC register, remove
9248 checks to save/restore the PIC register.
9249 (arc_expand_prologue): Likewise.
9250 * config/arc/arc-protos.h (arc_set_default_type_attributes):
9252 (arc_verify_short): Likewise.
9253 (arc_attr_type): Likewise.
9254 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
9255 (walk_stores): Likewise.
9256 (arc_address_cost): Make it static.
9257 (arc_verify_short): Likewise.
9258 (branch_dest): Likewise.
9259 (arc_attr_type): Likewise.
9260 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
9261 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
9262 (arc_final_prescan_insn): Remove inserting the nops due to
9263 hardware hazards. It is done in reorg step.
9264 (insn_length_variant_t): Remove.
9265 (insn_length_parameters_t): Likewise.
9266 (arc_insn_length_parameters): Likewise.
9267 (arc_get_insn_variants): Likewise.
9268 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
9270 2018-03-06 Jakub Jelinek <jakub@redhat.com>
9273 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
9276 PR tree-optimization/84687
9277 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
9279 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
9281 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9283 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
9284 Rename to ppc_speculation_barrier.
9285 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
9286 __builtin_ppc_speculation_barrier.
9288 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9291 * combine.c (combine_simplify_rtx): Don't try to simplify if
9292 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
9295 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
9297 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
9298 to 32 bytes when compiling for POWER9.
9300 2018-03-05 Jakub Jelinek <jakub@redhat.com>
9303 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
9304 regparm >= 3 with no arg reg available also for calls with
9305 flag_force_indirect_call. Pass decl to ix86_function_regparm.
9308 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
9310 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
9312 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
9315 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
9317 2018-03-05 Richard Biener <rguenther@suse.de>
9319 PR tree-optimization/84486
9320 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
9321 When inserting a __builtin_assume_aligned call set the LHS
9322 SSA name alignment info accordingly.
9324 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
9326 PR tree-optimization/84114
9327 * config/aarch64/aarch64.c (aarch64_reassociation_width)
9328 Avoid reassociation of FLOAT_MODE addition.
9330 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
9332 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
9333 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
9334 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
9335 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
9336 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
9337 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
9338 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
9340 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
9341 __builtin_ia32_wbinvd): New builtins.
9342 (SPECIAL_ARGS2): New.
9343 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
9344 (SPECIAL_ARGS2): New.
9345 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
9346 (ix86_valid_target_attribute_inner_p): Ditto.
9347 (ix86_init_mmx_sse_builtins): Add special_args2.
9348 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
9349 TARGET_WBNOINVD_P): New.
9350 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
9351 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
9352 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
9353 * config/i386/immintrin.h (_wbinvd): New intrinsic.
9354 * config/i386/pconfigintrin.h: New file.
9355 * config/i386/wbnoinvdintrin.h: Ditto.
9356 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
9358 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
9360 2018-03-05 Richard Biener <rguenther@suse.de>
9362 PR tree-optimization/84670
9363 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
9365 (BB_VISITED_WITH_VISITED_SUCCS): New define.
9366 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
9367 (compute_antic_aux): Only assert the number of values in ANTIC_IN
9368 doesn't grow if all successors (recursively) were visited at least
9371 2018-03-05 Richard Biener <rguenther@suse.de>
9373 PR tree-optimization/84650
9374 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
9375 if executed in the loop pipeline.
9377 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
9379 * doc/configfiles.texi (Configuration Files): Move info about
9380 conditionalizing $target-protos.h to...
9381 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
9382 differs from $target-protos.h.
9384 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
9385 Chung-Ju Wu <jasonwucj@gmail.com>
9387 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
9388 * config/nds32/nds32-multiple.md (setmemsi): Define.
9389 * config/nds32/nds32-memory-manipulation.c
9390 (nds32_gen_dup_4_byte_to_word_value): New.
9391 (emit_setmem_word_loop): New.
9392 (emit_setmem_byte_loop): New.
9393 (nds32_expand_setmem_loop): New.
9394 (nds32_expand_setmem_loop_v3m): New.
9395 (nds32_expand_setmem_unroll): New.
9396 (nds32_expand_setmem): New.
9398 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9399 Chung-Ju Wu <jasonwucj@gmail.com>
9401 * config/nds32/nds32-memory-manipulation.c
9402 (nds32_emit_load_store): New.
9403 (nds32_emit_post_inc_load_store): New.
9404 (nds32_emit_mem_move): New.
9405 (nds32_emit_mem_move_block): New.
9406 (nds32_expand_movmemsi_loop_unknown_size): New.
9407 (nds32_expand_movmemsi_loop_known_size): New.
9408 (nds32_expand_movmemsi_loop): New.
9409 (nds32_expand_movmemsi_unroll): New.
9410 (nds32_expand_movmemqi): Rename ...
9411 (nds32_expand_movmemsi): ... to this.
9412 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
9413 (movmemsi): ... to this.
9414 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
9415 (nds32_expand_movmemsi): ... to this.
9417 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9418 Monk Chiang <sh.chiang04@gmail.com>
9419 Chung-Ju Wu <jasonwucj@gmail.com>
9421 * config/nds32/nds32-protos.h
9422 (nds32_expand_load_multiple): New arguments.
9423 (nds32_expand_store_multiple): Ditto.
9424 (nds32_valid_multiple_load_store): Rename ...
9425 (nds32_valid_multiple_load_store_p): ... to this.
9426 * config/nds32/nds32-memory-manipulation.c
9427 (nds32_expand_load_multiple): Refine implementation.
9428 (nds32_expand_store_multiple): Ditto.
9429 * config/nds32/nds32-multiple.md
9430 (load_multiple): Update nds32_expand_load_multiple interface.
9431 (store_multiple): Update nds32_expand_store_multiple interface.
9432 * config/nds32/nds32-predicates.c
9433 (nds32_valid_multiple_load_store): Rename ...
9434 (nds32_valid_multiple_load_store_p): ... to this and refine
9436 * config/nds32/predicates.md
9437 (nds32_load_multiple_and_update_address_operation): New predicate.
9438 (nds32_store_multiple_and_update_address_operation): New predicate.
9440 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
9441 Chung-Ju Wu <jasonwucj@gmail.com>
9443 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
9444 (combo): New attribute.
9445 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
9447 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
9449 * config/nds32/nds32.opt: Change -mcmodel= default value.
9451 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
9452 Monk Chiang <sh.chiang04@gmail.com>
9453 Chung-Ju Wu <jasonwucj@gmail.com>
9455 * config/nds32/constants.md (unspec_element): New enum.
9456 * config/nds32/constraints.md (Umw): New constraint.
9457 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
9458 * config/nds32/nds32-intrinsic.md: Likewise.
9459 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
9460 (nds32_valid_smw_lwm_base_p): New.
9461 (nds32_output_smw_single_word): New.
9462 (nds32_output_lmw_single_word): New.
9463 (nds32_expand_unaligned_load): New.
9464 (nds32_expand_unaligned_store): New.
9465 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
9466 (nds32_output_smw_single_word): Declare.
9467 (nds32_output_lmw_single_word): Declare.
9468 (nds32_expand_unaligned_load): Declare.
9469 (nds32_expand_unaligned_store): Declare.
9470 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
9471 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
9472 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
9473 NDS32_BUILTIN_UASTORE_DW.
9474 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
9477 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
9478 Kito Cheng <kito.cheng@gmail.com>
9479 Chung-Ju Wu <jasonwucj@gmail.com>
9481 * config/nds32/nds32-intrinsic.c
9482 (nds32_expand_builtin_null_ftype_reg): Delete.
9483 (nds32_expand_builtin_reg_ftype_imm): Ditto.
9484 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
9485 (nds32_read_argument): New.
9486 (nds32_legitimize_target): Ditto.
9487 (nds32_legitimize_argument): Ditto.
9488 (nds32_check_constant_argument): Ditto.
9489 (nds32_expand_unop_builtin): Ditto.
9490 (nds32_expand_unopimm_builtin): Ditto.
9491 (nds32_expand_binop_builtin): Ditto.
9492 (nds32_builtin_decl_impl): Ditto.
9493 (builtin_description): Ditto.
9494 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
9495 (nds32_init_builtins_impl): Ditto.
9496 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
9497 (nds32_builtin_decl): New.
9498 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
9499 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
9501 2018-03-02 Jeff Law <law@redhat.com>
9503 * reorg.c (stop_search_p): Handle DEBUG_INSN.
9504 (redundant_insn, fill_simple_delay_slots): Likewise.
9505 (fill_slots_from_thread): Likewise.
9506 * resource.c (mark_referenced_resources): Likewise.
9507 (mark_set_resources, find_dead_or_set_registers): Likewise.
9509 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9511 * substring-locations.h (format_warning_va): Formatting fix for
9513 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
9515 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
9516 * substring-locations.c: Include intl.h.
9517 (format_warning_va): Turned into small wrapper around
9518 format_warning_n_va, renamed to ...
9519 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
9520 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
9522 (format_warning_at_substring_n): New function.
9523 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9524 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9525 format_warning_at_substring with just a shorter name instead of
9526 const function pointer.
9527 (fmtwarn_n): New function.
9528 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9529 appropriate, get rid of all the fmtstr temporaries, move conditionals
9530 with G_() wrapped string literals directly into fmtwarn arguments,
9531 cast dir.len to (int), formatting fixes.
9533 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9535 * doc/invoke.texi: Remove "Cilk Plus" references.
9537 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9538 Richard Biener <rguenther@suse.de>
9541 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9542 for error or warning attributes if CALL_FROM_THUNK_P is set.
9545 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9548 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9549 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9552 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9553 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9555 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9556 instead of -1U in last predictors element's probability member.
9558 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9561 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9562 arguments if they are comparable.
9564 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9566 PR tree-optimization/84634
9567 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9568 masks and masked_loop_p with a single loop_masks, making sure it's
9569 null for bb vectorization.
9571 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9573 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9574 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9575 loop->force_vectorize to check whether there is no alias.
9577 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9580 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9582 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9584 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9585 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9586 instead of a loop around prev_real_insn.
9587 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9591 * config/i386/i386.c (ix86_print_operand): Use conditional
9592 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9595 2018-03-02 Richard Biener <rguenther@suse.de>
9597 PR tree-optimization/84427
9598 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9599 (bitmap_set_subtract_values): Rewrite to handle multiple
9602 (prune_clobbered_mems): Likewise.
9603 (phi_translate): Take edge instead of pred/phiblock.
9604 (phi_translate_1): Likewise.
9605 (phi_translate_set): Likewise. Insert all translated
9606 exprs for a value into the set, keeping possibly multiple
9607 expressions per value.
9608 (compute_antic_aux): Adjust for phi_translate changes.
9609 When intersecting union the expressions and prune those
9610 not in the final value set, keeping possibly multiple
9611 expressions per value. Do not use value-insertion
9612 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9613 all expressions. Add verification that the value-sets
9614 only shrink during iteration.
9615 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9616 (do_pre_regular_insertion): Likewise.
9617 (do_pre_partial_partial_insertion): Likewise.
9619 2018-03-02 Richard Biener <rguenther@suse.de>
9622 * config/darwin.c (saved_debug_info_level): New static global.
9623 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9624 (darwin_asm_lto_end): Restore debug info generation settings.
9626 2018-03-01 Martin Liska <mliska@suse.cz>
9629 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9632 2018-03-01 Richard Biener <rguenther@suse.de>
9635 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9636 type annotation with LTO when debug was disabled at compile-time.
9638 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9640 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9642 (mips_final_postscan_insn): Likewise.
9644 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9646 PR rtl-optimization/84528
9647 * alias.c (init_alias_target): Add commentary.
9648 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9649 a unique base value if the frame pointer is not eliminated
9650 to the stack pointer.
9652 2018-03-01 Tom de Vries <tom@codesourcery.com>
9654 PR rtl-optimization/83327
9655 * lra-int.h (hard_regs_spilled_into): Declare.
9656 * lra.c (hard_regs_spilled_into): Define.
9657 (init_reg_info): Init hard_regs_spilled_into.
9658 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9659 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9660 (process_bb_lives): Handle hard_regs_spilled_into.
9661 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9662 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9664 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9666 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9667 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9668 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9669 * config/rs6000/aix72.h: New file.
9671 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9673 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9674 instead of warning_at with conditional singular and plural messages
9678 * stor-layout.c (update_alignment_for_field): For
9679 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9680 && !DECL_PACKED (field), do the alignment update, just use
9681 only desired_align instead of MAX (type_align, desired_align)
9683 (place_field): Don't do known_align < desired_align handling
9684 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9685 is non-NULL, instead do it after rli->prev_field handling and
9686 only if not within a bitfield word. For DECL_PACKED (field)
9687 use type_align of BITS_PER_UNIT.
9689 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9691 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9692 superfluous parentheses and trailing spaces.
9694 2018-02-28 Richard Biener <rguenther@suse.de>
9696 PR tree-optimization/84584
9697 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9698 SCoPs with fake exit edge.
9700 2018-02-28 Martin Liska <mliska@suse.cz>
9703 * timevar.c (timer::print): Fix format to properly print 100%
9706 2018-02-28 Richard Biener <rguenther@suse.de>
9709 * genmatch.c (capture_info::walk_match): Do not mark
9710 captured expressions without operands as expr_p given
9711 they act more like predicates and should be subject to
9712 "lost tail" side-effect preserving.
9714 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9716 PR rtl-optimization/81611
9717 * auto-inc-dec.c (attempt_change): Move dead note from
9718 mem_insn if it's the next use of regno
9719 (find_address): Take address use of reg holding
9720 non-incremented value. Add parm to limit search to the named
9722 (merge_in_block): Attempt to use a mem insn that is the next
9723 use of the original regno.
9725 2018-02-27 Martin Sebor <msebor@redhat.com>
9728 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9729 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9731 2018-02-27 Martin Sebor <msebor@redhat.com>
9733 PR translation/84207
9734 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9735 n argument to unsigned HOST_WIDE_INT.
9736 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9737 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9738 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9739 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9741 2018-02-27 Richard Biener <rguenther@suse.de>
9743 PR tree-optimization/84512
9744 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9745 Do not use the estimate returned from record_stmt_cost for
9746 the scalar iteration cost but sum properly using add_stmt_cost.
9748 2018-02-27 Richard Biener <rguenther@suse.de>
9750 PR tree-optimization/84466
9751 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9752 Adjust last change to less strictly validate use operands.
9754 2018-02-27 Martin Liska <mliska@suse.cz>
9756 PR gcov-profile/84548
9757 * gcov.c (process_file): Allow partial overlap and consider it
9758 also as group functions.
9759 (output_lines): Properly calculate range of lines for a group.
9761 2018-02-27 Martin Liska <mliska@suse.cz>
9763 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9764 'ggc' suffixes. Change first column width.
9765 (timer::print): Fix formatting of the column.
9767 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9769 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9770 preserve inline entry blocks for the sake of debug inline
9771 entry point markers alone.
9772 (remove_unused_locals): Suggest in comments a better place to
9773 force the preservation of inline entry blocks that are
9774 otherwise unused, but do not preserve them.
9776 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9778 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9780 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9783 * config/i386/constraints.md (Bs): Replace
9784 ix86_indirect_branch_register with
9785 TARGET_INDIRECT_BRANCH_REGISTER.
9787 * config/i386/i386.md (indirect_jump): Likewise.
9788 (tablejump): Likewise.
9789 (*sibcall_memory): Likewise.
9790 (*sibcall_value_memory): Likewise.
9791 Peepholes of indirect call and jump via memory: Likewise.
9792 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9793 (*sibcall_value_GOT_32): Likewise.
9794 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9795 (GOT_memory_operand): Likewise.
9796 (call_insn_operand): Likewise.
9797 (sibcall_insn_operand): Likewise.
9798 (GOT32_symbol_operand): Likewise.
9799 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9801 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9803 PR rtl-optimization/83496
9804 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9805 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9806 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9807 redundant insn, if any.
9808 (relax_delay_slots): Likewise.
9809 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9811 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9813 PR tree-optimization/83965
9814 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9815 that grouped statements are part of a reduction chain. Return
9816 true if the statement is not marked as a reduction itself but
9818 (vect_recog_dot_prod_pattern): Don't check whether the statement
9819 is part of a group here.
9820 (vect_recog_sad_pattern): Likewise.
9821 (vect_recog_widen_sum_pattern): Likewise.
9823 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9826 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9828 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9831 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9833 (ix86_output_indirect_function_return): New prototype.
9834 (ix86_split_simple_return_pop_internal): Likewise.
9835 * config/i386/i386.c (indirect_return_via_cx): New.
9836 (indirect_return_via_cx_bnd): Likewise.
9837 (indirect_thunk_name): Handle return va CX_REG.
9838 (output_indirect_thunk_function): Create alias for
9839 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9840 (ix86_output_indirect_jmp): Remove the bool argument.
9841 (ix86_output_indirect_function_return): New function.
9842 (ix86_split_simple_return_pop_internal): Likewise.
9843 * config/i386/i386.md (*indirect_jump): Don't pass false
9844 to ix86_output_indirect_jmp.
9845 (*tablejump_1): Likewise.
9846 (simple_return_pop_internal): Change it to define_insn_and_split.
9847 Call ix86_split_simple_return_pop_internal to split it for
9849 (simple_return_indirect_internal): Call
9850 ix86_output_indirect_function_return instead of
9851 ix86_output_indirect_jmp.
9853 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9856 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9857 memset and value initialization afterwards.
9859 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9861 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9863 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9866 * common/config/aarch64/aarch64-common.c
9867 (aarch_option_optimization_table[]): Switch
9868 off fomit-frame-pointer
9870 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9871 Chung-Ju Wu <jasonwucj@gmail.com>
9873 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9875 (store_multiple): Ditto.
9877 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9879 * config.gcc: Add --with-cpu support for nds32 target.
9880 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9881 * config/nds32/nds32.opt: Add -mcpu= option.
9883 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9885 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9886 isel=yes): Warn for these deprecated options.
9888 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9890 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9891 ISA_2_5_MASKS_EMBEDDED.
9893 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9895 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9896 p->max as pointers rather than using iterative_hash_expr.
9898 2018-02-23 Carl Love <cel@us.ibm.com>
9900 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9901 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9903 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9904 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9905 P8V_BUILTIN_VEC_VUNSIGNED2.
9907 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9910 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9911 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9913 * lra-constraints.c (curr_insn_transform): Set up
9914 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9916 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9917 * lra-eliminations.c (spill_pseudos): Ditto.
9918 (process_insn_for_elimination): Ditto.
9919 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9920 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9921 LRA_NON_CLOBBERED_ALT.
9923 2018-02-22 Martin Sebor <msebor@redhat.com>
9925 PR tree-optimization/84480
9926 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9927 to maybe_diag_stxncpy_trunc. Call it.
9928 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9929 from gimple_fold_builtin_strcpy. Print inlining stack.
9930 (handle_builtin_stxncpy): Print inlining stack.
9931 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9933 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9936 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9937 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9938 and -fcheck-pointer-bounds are used together.
9939 (indirect_thunk_prefix): New enum.
9940 (indirect_thunk_need_prefix): New function.
9941 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9942 "_nt" instead of "_bnd" for NOTRACK prefix.
9943 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9944 (output_indirect_thunk_function): Likewise.
9946 (ix86_code_end): Update output_indirect_thunk_function calls.
9947 (ix86_output_indirect_branch_via_reg): Replace
9948 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9949 (ix86_output_indirect_branch_via_push): Likewise.
9950 (ix86_output_function_return): Likewise.
9951 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9952 incompatible with -fcf-protection=branch and
9953 -fcheck-pointer-bounds.
9955 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9958 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9959 Change gcc_assert call to output_operand_lossage.
9961 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9963 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9965 2018-02-22 DJ Delorie <dj@redhat.com>
9966 Sebastian Perta <sebastian.perta@renesas.com>
9967 Oleg Endo <olegendo@gcc.gnu.org>
9969 * config/rx/rx.c (rx_rtx_costs): New function.
9970 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9972 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9974 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9976 2018-02-22 Martin Liska <mliska@suse.cz>
9979 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9980 Add "native" as a possible value.
9982 2018-02-22 Martin Liska <mliska@suse.cz>
9985 * config/i386/i386.c (ix86_option_override_internal):
9986 Add "native" as a possible value for -march and -mtune.
9988 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9991 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9992 to all type variants.
9994 PR tree-optimization/84503
9995 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9996 width as info->bitpos + info->bitsize - start.
9997 (merged_store_group::merge_overlapping): Simplify width computation.
9998 (check_no_overlap): New function.
9999 (imm_store_chain_info::try_coalesce_bswap): Compute expected
10000 start + width and last_order of the group, fail if check_no_overlap
10002 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
10003 to group if check_no_overlap fails.
10005 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10007 * config/rs6000/altivec.md: Delete contraint arguments to
10008 define_expand, define_split, and define_peephole2, and in
10009 define_insn_and_split if always unused.
10010 * config/rs6000/darwin.md: Ditto.
10011 * config/rs6000/dfp.md: Ditto.
10012 * config/rs6000/rs6000.md: Ditto.
10013 * config/rs6000/sync.md: Ditto.
10014 * config/rs6000/vector.md: Ditto.
10015 * config/rs6000/vsx.md: Ditto.
10017 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
10019 * config/rs6000/altivec.md: Write output control strings as braced
10020 blocks instead of double-quoted strings.
10021 * config/rs6000/darwin.md: Ditto.
10022 * config/rs6000/rs6000.md: Ditto.
10023 * config/rs6000/vector.md: Ditto.
10024 * config/rs6000/vsx.md: Ditto.
10026 2018-02-21 Jason Merrill <jason@redhat.com>
10028 PR c++/84314 - ICE with templates and fastcall attribute.
10029 * attribs.c (build_type_attribute_qual_variant): Remove assert.
10031 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10033 * ipa-cp.c (determine_versionability): Fix comment typos.
10035 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
10038 * ipa-cp.c (determine_versionability): Do not version functions caling
10041 2018-02-21 Martin Liska <mliska@suse.cz>
10044 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
10045 Add "native" as a possible value.
10046 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
10047 the macro when native cpu detection is available.
10049 2018-02-21 Martin Liska <mliska@suse.cz>
10052 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
10053 Add "native" as a possible value.
10054 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
10055 when native cpu detection is available.
10057 2018-02-21 Jakub Jelinek <jakub@redhat.com>
10058 Martin Sebor <msebor@redhat.com>
10060 PR tree-optimization/84478
10061 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
10063 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
10064 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
10065 support which is conservatively correct, for 2 only stay conservative
10066 for maxlen. Formatting and comment capitalization fixes. Add STRICT
10067 argument to the 2 argument get_range_strlen, adjust 6 arg
10068 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
10070 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
10071 (gimple_fold_builtin_strlen): Pass true as last argument to
10074 2018-02-20 Martin Sebor <msebor@redhat.com>
10076 PR middle-end/84095
10077 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
10078 (builtin_memref::set_base_and_offset): Same. Handle inner references.
10079 (builtin_memref::builtin_memref): Factor out parts into
10080 set_base_and_offset and call it.
10082 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10084 PR middle-end/84406
10085 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
10086 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
10087 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
10088 search at the associated MODE_INT.
10090 2018-02-20 Jeff Law <law@redhat.com>
10092 PR middle-end/82123
10093 PR tree-optimization/81592
10094 PR middle-end/79257
10095 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
10096 for range data rather than using global data.
10097 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
10098 range data rather than using global data.
10099 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
10100 pass it to children as needed.
10101 (struct directive::fmtresult): Similarly.
10102 (struct directive::set_width): Similarly.
10103 (struct directive::set_precision): Similarly.
10104 (format_integer, format_directive, parse_directive): Similarly.
10105 (format_none): Accept unnamed vr_values parameter.
10106 (format_percent, format_floating, format_character): Similarly.
10107 (format_string, format_plain): Similarly.
10108 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
10109 the EVRP range analyzer for range data rather than using global data.
10110 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
10111 gimple-ssa-evrp-analyze.h
10112 (class sprintf_dom_walker): Add after_dom_children member function.
10113 Add evrp_range_analyzer member.
10114 (sprintf_dom_walker::before_dom_children): Call into the EVRP
10115 range analyzer as needed.
10116 (sprintf_dom_walker::after_dom_children): New member function.
10117 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
10119 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
10120 (evrp_range_analyzer::pop_to_marker): Likewise.
10122 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
10124 PR tree-optimization/84419
10125 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
10126 with the required type if its current type is compatible but
10129 2018-02-20 Jakub Jelinek <jakub@redhat.com>
10131 PR middle-end/82004
10132 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
10133 after vectorization.
10135 2018-02-20 Martin Liska <mliska@suse.cz>
10138 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
10139 possible values if we don't have a hint.
10141 2018-02-20 Martin Liska <mliska@suse.cz>
10145 * final.c (shorten_branches): Build align_tab array with one
10147 * opts.c (finish_options): Add alignment option limit check.
10148 (MAX_CODE_ALIGN): Likewise.
10149 (MAX_CODE_ALIGN_VALUE): Likewise.
10150 * doc/invoke.texi: Document maximum allowed option value for
10151 all -falign-* options.
10153 2018-02-19 Jakub Jelinek <jakub@redhat.com>
10156 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
10157 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
10158 * var-tracking.c (emit_note_insn_var_location): Remove all references
10159 to NOTE_INSN_CALL_ARG_LOCATION.
10160 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
10161 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
10162 Use copy_rtx_if_shared.
10163 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
10164 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
10165 (dwarf2out_var_location): Remove handling of
10166 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
10168 * final.c (final_scan_insn): Remove all references to
10169 NOTE_INSN_CALL_ARG_LOCATION.
10170 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
10171 before dumping final insns.
10172 * except.c (emit_note_eh_region_end): Remove all references to
10173 NOTE_INSN_CALL_ARG_LOCATION.
10174 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
10175 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
10176 * config/arc/arc.c (hwloop_optimize): Likewise.
10177 * config/arm/arm.c (create_fix_barrier): Likewise.
10178 * config/s390/s390.c (s390_chunkify_start): Likewise.
10179 * config/sh/sh.c (find_barrier): Likewise.
10180 * config/i386/i386.c (rest_of_insert_endbranch,
10181 ix86_seh_fixup_eh_fallthru): Likewise.
10182 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
10183 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
10184 * config/frv/frv.c (frv_function_prologue): Likewise.
10185 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
10187 (note_outside_basic_block_p): Remove all references to
10188 NOTE_INSN_CALL_ARG_LOCATION.
10189 * gengtype.c (adjust_field_rtx_def): Likewise.
10190 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
10192 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
10193 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
10196 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
10199 PR tree-optimization/84452
10200 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
10201 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
10204 2018-02-19 Martin Liska <mliska@suse.cz>
10207 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
10209 2018-02-19 Martin Liska <mliska@suse.cz>
10210 Richard Sandiford <richard.sandiford@linaro.org>
10212 PR tree-optimization/82491
10213 * gimple-fold.c (get_base_constructor): Make earlier bail out
10216 2018-02-19 Carl Love <cel@us.ibm.com>
10218 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
10219 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
10221 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
10222 P8V_BUILTIN_VEC_NEG.
10224 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
10226 * config/rl78/rl78.md (movdf): New define expand.
10228 2018-02-19 Martin Liska <mliska@suse.cz>
10231 * doc/invoke.texi: Fix typo.
10232 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
10234 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
10236 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
10237 handle rs6000_single_float and rs6000_double_float specially for
10240 2018-02-16 Jeff Law <law@redhat.com>
10242 * config/rx/rx.c (add_pop_cfi_notes): New function.;
10243 (pop_regs): Use it.
10245 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10248 * ipa-inline.c (inline_small_functions): Fix a typo.
10250 2018-02-16 Nathan Sidwell <nathan@acm.org>
10252 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
10254 2018-02-16 Carl Love <cel@us.ibm.com>
10256 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
10257 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
10258 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
10259 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
10260 expansion to P8V_BUILTIN_VEC_FLOAT2.
10262 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
10264 PR rtl-optimization/70023
10265 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
10266 src_regno into account.
10268 2018-02-16 Carl Love <cel@us.ibm.com>
10270 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
10271 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
10272 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
10273 * config/rs6000/rs6000.c: Remove case statements for
10274 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
10275 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
10276 and P9V_BUILTIN_VEC_VINSERT4B.
10277 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
10278 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
10279 * config/rs6000/vsx.md:
10280 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
10283 2018-02-16 Carl Love <cel@us.ibm.com>
10285 * config/rs6000/altivec.h: Add builtin names vec_extract4b
10287 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
10289 * config/rs6000/rs6000-c.c: Add the definitions for
10290 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
10291 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
10292 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
10293 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
10294 definition for insert4b and define insn *insert3b_internal.
10295 * doc/extend.texi: Add documentation for vec_extract4b.
10297 2018-02-16 Nathan Sidwell <nathan@acm.org>
10299 * doc/extend.texi (Backwards Compatibility): Mention friend
10300 injection. Note for-scope is deprecated.
10301 * doc/invoke.texi (-ffriend-injection): Deprecate.
10303 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
10305 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
10306 that moved to I2, also allow destinations that are a paradoxical
10307 subreg (instead of a normal reg).
10309 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
10312 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
10315 2018-02-16 Richard Biener <rguenther@suse.de>
10317 PR tree-optimization/84037
10318 PR tree-optimization/84016
10320 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10321 Adjust vec_construct for the fact we need additional higher latency
10322 128bit inserts for AVX256 and AVX512 vector builds.
10323 (ix86_add_stmt_cost): Scale vector construction cost for
10326 2018-02-16 Richard Biener <rguenther@suse.de>
10328 PR tree-optimization/84417
10329 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
10330 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
10331 (non_rewritable_lvalue_p): Likewise, use poly-ints.
10333 2018-02-16 Martin Liska <mliska@suse.cz>
10336 * internal-fn.def (ASAN_CHECK): Set proper flags.
10337 (ASAN_MARK): Likewise.
10339 2018-02-16 Julia Koval <julia.koval@intel.com>
10341 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
10342 from PTA_CANNONLAKE.
10344 2018-02-16 Jakub Jelinek <jakub@redhat.com>
10347 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
10348 Use ++iter rather than iter++ for std::list iterators.
10349 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
10350 defer deleting them until all nodes in the forest are processed. Do
10351 free even leaf nodes. Change to_process into auto_vec.
10354 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
10355 * vec.h (vec_default_construct): Use memset instead of placement new
10356 if BROKEN_VALUE_INITIALIZATION is defined.
10357 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
10358 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
10361 PR rtl-optimization/83723
10362 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
10363 * lra.c (lra_substitute_pseudo): Likewise. If true, use
10364 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
10366 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
10368 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
10370 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
10372 PR rtl-optimization/81443
10373 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
10374 from inner REGs to paradoxical SUBREGs.
10376 2018-02-16 Richard Biener <rguenther@suse.de>
10378 PR tree-optimization/84399
10379 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
10380 For operands we can analyze at their definition make sure we can
10381 analyze them at each use as well.
10383 2018-02-16 Richard Biener <rguenther@suse.de>
10385 PR tree-optimization/84190
10386 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
10387 volatile accesses if the decl isn't volatile.
10389 2018-02-15 Jason Merrill <jason@redhat.com>
10391 PR c++/84314 - ICE with templates and fastcall attribute.
10392 * attribs.c (build_type_attribute_qual_variant): Don't clobber
10393 TYPE_CANONICAL on an existing type.
10395 2018-02-15 Jakub Jelinek <jakub@redhat.com>
10397 PR tree-optimization/84383
10398 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
10399 dstoff nor call operand_equal_p if dstbase is NULL.
10401 PR tree-optimization/84334
10402 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
10403 also a CONSTANT_CLASS_P, punt.
10405 2018-02-14 Jim Wilson <jimw@sifive.com>
10407 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
10408 first SMALL_OPERAND check. New local min_second_step. Move assert
10409 to where locals are set. Add TARGET_RVC support.
10410 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
10412 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
10414 * doc/invoke.texi: Correct -Wformat-overflow code sample.
10416 2018-02-14 Martin Sebor <msebor@redhat.com>
10418 PR tree-optimization/83698
10419 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
10420 arrays constrain the offset range to their bounds.
10421 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
10422 (builtin_access::overlap): Avoid setting the size of overlap if it's
10424 (maybe_diag_overlap): Also consider arrays when deciding what values
10425 of offsets to include in diagnostics.
10427 2018-02-14 Martin Sebor <msebor@redhat.com>
10430 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
10431 that correspond to the kind of a declaration.
10433 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
10436 * config/pa/pa.md: Load address of PIC label using the linkage table
10437 if the label is nonlocal.
10439 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
10441 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
10442 warning message if user requests -maltivec=be.
10443 * doc/invoke.texi: Document deprecation of -maltivec=be.
10445 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
10448 * config/rs6000/rs6000-c.c: Update definitions for
10449 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
10450 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
10452 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10455 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
10456 add _get_ssp intrinsics. Remove argument from
10457 __builtin_ia32_rdssp[d|q].
10458 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
10459 * config/i386/i386-builtin.def: Remove argument from
10460 __builtin_ia32_rdssp[d|q].
10461 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
10462 ix86_expand_special_args_builtin for _rdssp[d|q].
10463 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
10464 Clear register before usage.
10465 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
10466 Add documentation for new _get_ssp and _inc_ssp intrinsics.
10468 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
10470 PR tree-optimization/84357
10471 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
10472 operand 1 of an ARRAY_REF too.
10474 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
10477 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
10478 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10480 (set_of_reg): New struct.
10481 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
10482 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
10483 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
10485 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
10486 Split into bitclr, bitset, bitinvert patterns if appropriate.
10487 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
10488 use rx_fuse_in_memory_bitop.
10489 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
10490 to named insn, correct maximum insn length.
10492 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
10495 * machmode.def: Define a complex mode for PARTIAL_INT.
10496 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
10498 * doc/rtl.texi: Document CSPImode.
10499 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
10501 (msp430_hard_regno_nregs_with_padding): Likewise.
10503 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
10506 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
10508 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
10510 PR rtl-optimization/84169
10511 * combine.c (try_combine): New variable split_i2i3. Set it to true if
10512 we generated a parallel as new i3 and we split that to new i2 and i3
10513 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
10514 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
10515 those to i2, not i1. Partially rewrite this scan code.
10517 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10520 * stor-layout.c (place_field): For variable length fields, adjust
10521 offset_align afterwards not just based on the field's alignment,
10522 but also on the size.
10524 PR middle-end/84309
10525 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10526 of exps and logs in the use_exp2 case.
10528 2018-02-13 Jeff Law <law@redhat.com>
10530 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10531 entry for "vector".
10533 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10536 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10540 * common.opt (gas-loc-support, gas-locview-support): New.
10541 (ginline-points, ginternal-reset-location-views): New.
10542 * doc/invoke.texi: Document them. Use @itemx where intended.
10543 (gvariable-location-views): Adjust.
10544 * target.def (reset_location_view): New.
10545 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10546 (TARGET_RESET_LOCATION_VIEW): New.
10547 * doc/tm.texi: Rebuilt.
10548 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10549 (dwarf2out_default_as_locview_support): New.
10550 (output_asm_line_debug_info): Use option variables.
10551 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10552 (output_loc_list): Likewise.
10553 (add_high_low_attributes): Check option variables.
10554 Don't output entry view attribute in strict mode.
10555 (gen_inlined_subroutine_die): Check option variables.
10556 (dwarf2out_inline_entry): Likewise.
10557 (init_sections_and_labels): Likewise.
10558 (dwarf2out_early_finish): Likewise.
10559 (maybe_reset_location_view): New, from...
10560 (dwarf2out_var_location): ... here. Call it.
10561 * debug.h (dwarf2out_default_as_loc_support): Declare.
10562 (dwarf2out_default_as_locview_support): Declare.
10563 * hooks.c (hook_int_rtx_insn_0): New.
10564 * hooks.h (hook_int_rtx_insn_0): Declare.
10565 * toplev.c (process_options): Take -gas-loc-support and
10566 -gas-locview-support from dwarf2out. Enable
10567 -gvariable-location-views by default only with locview
10568 assembler support. Enable -ginternal-reset-location-views by
10569 default only if the target defines the corresponding hook.
10570 Enable -ginline-points by default if location views are
10571 enabled; force it disabled if statement frontiers are
10573 * tree-inline.c (expand_call_inline): Check option variables.
10574 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10576 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10578 PR tree-optimization/84321
10579 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10580 handling. Also check whether the anti-range contains any values
10581 that satisfy the mask; switch to a VR_RANGE if not.
10583 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10586 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10588 2018-02-13 Martin Jambor <mjambor@suse.cz>
10591 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10592 of call statements, also set location of a load to a temporary.
10594 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10596 * config/rl78/rl78.c (add_vector_labels): New function.
10597 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10598 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10599 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10600 which checks that no arguments are passed.
10601 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10602 * doc/extend.texi: Documentation for the new attribute.
10604 2018-02-13 Andreas Schwab <schwab@suse.de>
10606 * config/riscv/linux.h (CPP_SPEC): Define.
10608 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10611 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10612 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10613 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10614 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10615 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10616 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10617 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10618 temporarily for AES and PCLMUL builtins.
10620 PR tree-optimization/84339
10621 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10622 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10625 PR middle-end/84309
10626 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10627 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10628 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10630 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10632 * omp-simd-clone.h: New file.
10633 * omp-simd-clone.c: Include omp-simd-clone.h.
10634 (expand_simd_clones): No longer static.
10635 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10636 cgraph.h and omp-simd-clone.h.
10637 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10638 (vect_recog_widen_shift_pattern): Formatting fix.
10639 (vect_pattern_recog_1): Don't check optab for calls.
10642 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10643 operands[2] into a REG before using gen_lowpart on it.
10645 2018-02-12 Jeff Law <law@redhat.com>
10648 * config/sh/sh.c (find_barrier): Consider a sibling call
10651 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10652 successfully back substituting a reg.
10654 2018-02-12 Richard Biener <rguenther@suse.de>
10656 PR tree-optimization/84037
10657 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10658 parameter, move visited init to caller.
10659 (vect_slp_analyze_operations): Separate cost from validity
10660 check, initialize visited once for all instances.
10661 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10663 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10665 (vect_model_promotion_demotion_cost): Likewise.
10666 (vectorizable_bswap): Guard cost modeling with !slp_node
10667 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10669 (vectorizable_call): Likewise.
10670 (vectorizable_conversion): Likewise.
10671 (vectorizable_assignment): Likewise.
10672 (vectorizable_shift): Likewise.
10673 (vectorizable_operation): Likewise.
10674 (vectorizable_store): Likewise.
10675 (vectorizable_load): Likewise.
10676 (vectorizable_condition): Likewise.
10677 (vectorizable_comparison): Likewise.
10679 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10682 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10683 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10684 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10686 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10688 PR middle-end/83665
10689 * params.def (inline-min-speedup): Increase from 8 to 15.
10690 (max-inline-insns-auto): Decrease from 40 to 30.
10691 * ipa-split.c (consider_split): Add some buffer for function to
10692 be considered inlining candidate.
10693 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10696 2018-02-12 Richard Biener <rguenther@suse.de>
10698 PR tree-optimization/84037
10699 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10700 matched stmts if we cannot swap the non-matched ones.
10702 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10704 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10705 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10706 _mm_maskz_scalef_round_ss): New intrinsics.
10707 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10708 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10709 __builtin_ia32_scalefss_round): Remove.
10710 (__builtin_ia32_scalefsd_mask_round,
10711 __builtin_ia32_scalefss_mask_round): New intrinsics.
10712 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10713 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10714 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10715 "<round_constraint>")): Changed to ...
10716 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10717 "<round_scalar_constraint>")): ... this.
10718 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10719 %0, %1, %2<round_op3>}"): Changed to ...
10720 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10721 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10722 %2<round_scalar_mask_op3>}"): ... this.
10723 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10725 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10727 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10728 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10729 (_mm_maskz_sqrt_round_ss): New intrinsics.
10730 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10731 (__builtin_ia32_sqrtsd_mask_round)
10732 (__builtin_ia32_sqrtss_mask_round): New builtins.
10733 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10734 (__builtin_ia32_sqrtss_round): Remove.
10735 (__builtin_ia32_sqrtsd_mask_round)
10736 (__builtin_ia32_sqrtss_mask_round): New builtins.
10737 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10738 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10739 ((match_operand:VF_128 1 "vector_operand"
10740 "xBm,<round_constraint>")): Changed to ...
10741 ((match_operand:VF_128 1 "vector_operand"
10742 "xBm,<round_scalar_constraint>")): ... this.
10743 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10744 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10745 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10746 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10747 %<iptr>1<round_scalar_mask_op3>}): ... this.
10748 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10749 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10751 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10754 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10755 Cast vec_cmpeq result to correct type.
10756 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10757 Cast vec_cmpgt result to correct type.
10759 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10761 * final.c (final_scan_insn_1): Renamed from...
10762 (final_scan_insn): ... this. New wrapper, to recover
10763 seen from the outermost call in recursive ones.
10764 * config/sparc/sparc.c (output_return): Drop seen from call.
10765 (output_sibcall): Likewise.
10766 * config/visium/visium.c (output_branch): Likewise.
10768 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10770 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10773 2018-02-10 Alan Modra <amodra@gmail.com>
10776 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10777 Specify LR as an input.
10779 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10782 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10783 remove_member_access_dummy_vars): New functions.
10784 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10785 lower_omp_1, execute_lower_omp): Use them.
10787 PR rtl-optimization/84308
10788 * shrink-wrap.c (spread_components): Release todo vector.
10790 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10792 PR rtl-optimization/57193
10793 * ira-color.c (struct allocno_color_data): Add member
10794 conflict_allocno_hard_prefs.
10795 (update_conflict_allocno_hard_prefs): New.
10796 (bucket_allocno_compare_func): Add a preference based on
10797 conflict_allocno_hard_prefs.
10798 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10799 (color_allocnos): Remove a dead code. Initiate
10800 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10802 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10805 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10806 constraint from =wa to wa. Avoid a subreg on the output operand,
10807 instead use a pseudo and subreg it in a move.
10808 (p9_xxbrd_<mode>): Changed to ...
10809 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10810 (p9_xxbrd_v2df): New expander.
10811 (p9_xxbrw_<mode>): Changed to ...
10812 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10813 (p9_xxbrw_v4sf): New expander.
10815 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10817 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10819 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10822 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10823 multiply in 32-bit mode.
10824 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10825 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10828 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10830 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10831 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10832 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10833 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10835 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10838 * dwarf2out.c (is_trivial_indirect_ref): New function.
10839 (dwarf2out_late_global_decl): Do not generate a location
10840 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10841 and that are not defined in the current unit.
10843 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10845 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10846 instead of a libcall for UNORDERED.
10848 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10851 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10852 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10854 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10857 * config/s390/s390.c (s390_set_current_function): Invoke
10858 s390_indirect_branch_settings also if fndecl didn't change.
10860 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10862 * config/rs6000/rs6000.md (blockage): Set length to zero.
10864 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10866 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10868 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10871 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10872 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10876 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10877 PARALLEL incoming that failed vt_get_decl_and_offset check.
10879 PR middle-end/84237
10880 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10881 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10883 (get_variable_section): For decls in named .bss* sections pass true as
10884 second argument to bss_initializer_p.
10886 2018-02-09 Marek Polacek <polacek@redhat.com>
10887 Jakub Jelinek <jakub@redhat.com>
10890 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10891 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10892 Sync some changes from cxx_fold_indirect_ref.
10894 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10896 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10898 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10899 (BLOCK_INLINE_ENTRY_LABEL): New.
10900 (dwarf2out_var_location): Disregard inline entry markers.
10901 (inline_entry_data): New struct.
10902 (inline_entry_data_hasher): New hashtable type.
10903 (inline_entry_data_hasher::hash): New.
10904 (inline_entry_data_hasher::equal): New.
10905 (inline_entry_data_table): New variable.
10906 (add_high_low_attributes): Add DW_AT_entry_pc and
10907 DW_AT_GNU_entry_view attributes if a pending entry is found
10908 in inline_entry_data_table. Add old entry_pc attribute only
10909 if debug nonbinding markers are disabled.
10910 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10911 markers are enabled.
10912 (block_within_block_p, dwarf2out_inline_entry): New.
10913 (dwarf2out_finish): Check that no entries remained in
10914 inline_entry_data_table.
10915 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10916 (final_scan_insn, notice_source_line): Likewise.
10917 (rest_of_clean_state): Skip inline entry markers.
10918 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10920 * gimple.c (gimple_build_debug_inline_entry): New.
10921 * gimple.h (enum gimple_debug_subcode): Add
10922 GIMPLE_DEBUG_INLINE_ENTRY.
10923 (gimple_build_debug_inline_entry): Declare.
10924 (gimple_debug_inline_entry_p): New.
10925 (gimple_debug_nonbind_marker_p): Adjust.
10926 * insn-notes.def (INLINE_ENTRY): New.
10927 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10928 inline entry marker notes.
10929 (print_insn): Likewise.
10930 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10931 (INSN_DEBUG_MARKER_KIND): Likewise.
10932 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10933 * tree-inline.c (expand_call_inline): Build and insert
10934 debug_inline_entry stmt.
10935 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10936 inline entry blocks early, if nonbind markers are enabled.
10937 (dump_scope_block): Dump fragment info.
10938 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10939 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10940 (gimple_build_debug_inline_entry): New.
10941 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10942 Enable/disable inline entry points too.
10943 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10944 (DEBUG_INSN): Describe inline entry markers.
10946 * common.opt (gvariable-location-views): New.
10947 (gvariable-location-views=incompat5): New.
10948 * config.in: Rebuilt.
10949 * configure: Rebuilt.
10950 * configure.ac: Test assembler for view support.
10951 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10952 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10953 * dwarf2out.c (var_loc_view): New typedef.
10954 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10955 (dwarf2out_locviews_in_attribute): New.
10956 (dwarf2out_locviews_in_loclist): New.
10957 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10958 (enum dw_line_info_opcode): Add LI_adv_address.
10959 (struct dw_line_info_table): Add view.
10960 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10961 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10962 (zero_view_p): New variable.
10963 (ZERO_VIEW_P): New macro.
10964 (output_asm_line_debug_info): New.
10965 (struct var_loc_node): Add view.
10966 (add_AT_view_list, AT_loc_list): New.
10967 (add_var_loc_to_decl): Add view param. Test it against last.
10968 (new_loc_list): Add view params. Record them.
10969 (AT_loc_list_ptr): Handle loc and view lists.
10970 (view_list_to_loc_list_val_node): New.
10971 (print_dw_val): Handle dw_val_class_view_list.
10972 (size_of_die): Likewise.
10973 (value_format): Likewise.
10974 (loc_list_has_views): New.
10975 (gen_llsym): Set vl_symbol too.
10976 (maybe_gen_llsym, skip_loc_list_entry): New.
10977 (dwarf2out_maybe_output_loclist_view_pair): New.
10978 (output_loc_list): Output view list or entries too.
10979 (output_view_list_offset): New.
10980 (output_die): Handle dw_val_class_view_list.
10981 (output_dwarf_version): New.
10982 (output_compilation_unit_header): Use it.
10983 (output_skeleton_debug_sections): Likewise.
10984 (output_rnglists, output_line_info): Likewise.
10985 (output_pubnames, output_aranges): Update version comments.
10986 (output_one_line_info_table): Output view numbers in asm comments.
10987 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10988 Call maybe_gen_llsym.
10989 (loc_list_from_tree_1): Adjust.
10990 (add_AT_location_description): Create view list attribute if
10991 needed, check it's absent otherwise.
10992 (convert_cfa_to_fb_loc_list): Adjust.
10993 (maybe_emit_file): Call output_asm_line_debug_info for test.
10994 (dwarf2out_var_location): Reset views as needed. Precompute
10995 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10996 attribute. Set view.
10997 (new_line_info_table): Reset next view.
10998 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10999 (dwarf2out_source_line): Likewise. Output view resets and labels to
11000 the assembler, or select appropriate line info opcodes.
11001 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
11002 (optimize_string_length): Catch it. Adjust.
11003 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
11004 dw_val_class_view_list, and remove it if no longer needed.
11005 (hash_loc_list): Hash view numbers.
11006 (loc_list_hasher::equal): Compare them.
11007 (optimize_location_lists): Check whether a view list symbol is
11008 needed, and whether the locview attribute is present, and
11009 whether they match. Remove the locview attribute if no longer
11011 (index_location_lists): Call skip_loc_list_entry for test.
11012 (dwarf2out_finish): Call output_asm_line_debug_info for test.
11013 Use output_dwarf_version.
11014 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
11015 (struct dw_val_node): Add val_view_list.
11016 * final.c (SEEN_NEXT_VIEW): New.
11017 (set_next_view_needed): New.
11018 (clear_next_view_needed): New.
11019 (maybe_output_next_view): New.
11020 (final_start_function): Rename to...
11021 (final_start_function_1): ... this. Take pointer to FIRST,
11022 add SEEN parameter. Emit param bindings in the initial view.
11023 (final_start_function): Reintroduce SEEN-less interface.
11024 (final): Rename to...
11025 (final_1): ... this. Take SEEN parameter. Output final pending
11026 next view at the end.
11027 (final): Reintroduce seen-less interface.
11028 (final_scan_insn): Output pending next view before switching
11029 sections or ending a block. Mark the next view as needed when
11030 outputting variable locations. Notify debug backend of section
11031 changes, and of location view changes.
11032 (rest_of_handle_final): Adjust.
11033 * toplev.c (process_options): Autodetect value for debug variable
11034 location views option. Warn on incompat5 without -gdwarf-5.
11035 * doc/invoke.texi (gvariable-location-views): New.
11036 (gvariable-location-views=incompat5): New.
11037 (gno-variable-location-views): New.
11039 2018-02-08 David Malcolm <dmalcolm@redhat.com>
11041 PR tree-optimization/84136
11042 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
11043 that the result of find_edge is non-NULL.
11045 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11048 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
11049 storing integer register in SImode. Fix cost of 256 and 512
11050 byte aligned SSE register store.
11052 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
11054 * config/i386/i386.c (ix86_multiplication_cost): Fix
11055 multiplication cost for TARGET_AVX512DQ.
11057 2018-02-08 Marek Polacek <polacek@redhat.com>
11059 PR tree-optimization/84238
11060 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
11063 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11065 PR tree-optimization/84265
11066 * tree-vect-stmts.c (vectorizable_store): Don't treat
11067 VMAT_CONTIGUOUS accesses as grouped.
11068 (vectorizable_load): Likewise.
11070 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11072 PR tree-optimization/81635
11073 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
11074 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
11075 (test_round_for_mask): New functions.
11076 (wide_int_cc_tests): Call test_round_for_mask.
11077 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
11078 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
11079 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
11080 range returned by get_range_info.
11082 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
11085 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
11086 * symtab.c: Include builtins.h
11087 (symtab_node::output_to_lto_symbol_table_p): Move here
11088 from lto-streamer-out.c:output_symbol_p.
11089 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
11090 (output_symbol_p): Move all logic to symtab.c
11091 (produce_symtab): Update.
11093 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11095 * config/s390/s390-opts.h (enum indirect_branch): Define.
11096 * config/s390/s390-protos.h (s390_return_addr_from_memory)
11097 (s390_indirect_branch_via_thunk)
11098 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
11099 (enum s390_indirect_branch_type): Define.
11100 * config/s390/s390.c (struct s390_frame_layout, struct
11101 machine_function): Remove.
11102 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
11103 (indirect_branch_table_label_no, indirect_branch_table_name):
11105 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
11106 (enum s390_indirect_branch_option): Define.
11107 (s390_return_addr_from_memory): New function.
11108 (s390_handle_string_attribute): New function.
11109 (s390_attribute_table): Add new attribute handler.
11110 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
11111 (s390_indirect_branch_via_thunk): New function.
11112 (s390_indirect_branch_via_inline_thunk): New function.
11113 (s390_function_ok_for_sibcall): When jumping via thunk disallow
11114 sibling call optimization for non z10 compiles.
11115 (s390_emit_call): Force indirect branch target to be a single
11116 register. Add r1 clobber for non-z10 compiles.
11117 (s390_emit_epilogue): Emit return jump via return_use expander.
11118 (s390_reorg): Handle JUMP_INSNs as execute targets.
11119 (s390_option_override_internal): Perform validity checks for the
11120 new command line options.
11121 (s390_indirect_branch_attrvalue): New function.
11122 (s390_indirect_branch_settings): New function.
11123 (s390_set_current_function): Invoke s390_indirect_branch_settings.
11124 (s390_output_indirect_thunk_function): New function.
11125 (s390_code_end): Implement target hook.
11126 (s390_case_values_threshold): Implement target hook.
11127 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
11129 * config/s390/s390.h (struct s390_frame_layout)
11130 (struct machine_function): Move here from s390.c.
11131 (TARGET_INDIRECT_BRANCH_NOBP_RET)
11132 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
11133 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
11134 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
11135 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
11136 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
11137 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
11138 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
11139 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
11140 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
11141 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
11142 (mnemonic attribute): Add values which aren't recognized
11144 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
11145 pattern for branch conversion. Fix mnemonic attribute.
11146 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
11147 indirect branch via thunk if requested.
11148 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
11149 ("*indirect_jump"): Disable for branch conversion using out of
11151 ("indirect_jump_via_thunk<mode>_z10")
11152 ("indirect_jump_via_thunk<mode>")
11153 ("indirect_jump_via_inlinethunk<mode>_z10")
11154 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
11155 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
11156 ("casesi_jump_via_inlinethunk<mode>_z10")
11157 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
11158 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
11159 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
11160 ("*indirect2_jump"): Disable for branch conversion.
11161 ("casesi_jump"): Turn into expander and expand patterns for branch
11163 ("return_use"): New expander.
11164 ("*return"): Emit return via thunk and rename it to ...
11165 ("*return<mode>"): ... this one.
11166 * config/s390/s390.opt: Add new options and and enum for the
11169 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
11171 * lra-constraints.c (match_reload): Unconditionally use
11172 gen_lowpart_SUBREG, rather than selecting between that
11173 and equivalent gen_rtx_SUBREG code.
11175 2018-02-08 Richard Biener <rguenther@suse.de>
11177 PR tree-optimization/84233
11178 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
11179 changed flag instead of boguously re-using phi_inserted.
11181 2018-02-08 Martin Jambor <mjambor@suse.cz>
11183 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
11184 static local variables.
11186 2018-02-08 Richard Biener <rguenther@suse.de>
11188 PR tree-optimization/84278
11189 * tree-vect-stmts.c (vectorizable_store): When looking for
11190 smaller vector types to perform grouped strided loads/stores
11191 make sure the mode is supported by the target.
11192 (vectorizable_load): Likewise.
11194 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11196 * config/aarch64/aarch64.c (aarch64_components_for_bb):
11197 Increase LDP/STP opportunities by adding adjacent callee-saves.
11199 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
11201 PR rtl-optimization/84068
11202 PR rtl-optimization/83459
11203 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
11205 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
11207 PR tree-optimization/84224
11208 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
11209 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
11210 non-zero arguments.
11212 2018-02-07 Iain Sandoe <iain@codesourcery.com>
11215 * config/rs6000/altivec.md (*restore_world): Remove LR use.
11216 * config/rs6000/predicates.md (restore_world_operation): Adjust op
11217 count, remove one USE.
11219 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
11221 * doc/install.texi (Configuration): Document the
11222 --with-long-double-format={ibm,ieee} PowerPC configuration
11226 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
11227 Convert from define_expand to be define_insn_and_split. Rework
11228 float/double/_Float128 conversions to QI/HI/SImode to work with
11229 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
11230 conversions to QI/HImode types did a store and then a load to
11231 truncate the value. For conversions to VSX registers, don't split
11232 the insn, instead emit the code directly. Use the code iterator
11233 any_fix to combine signed and unsigned conversions.
11234 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
11235 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
11236 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11237 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
11238 (fix_<mode>di2_hw): Likewise.
11239 (fixuns_<mode>di2_hw): Likewise.
11240 (fix_<mode>si2_hw): Likewise.
11241 (fixuns_<mode>si2_hw): Likewise.
11242 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
11243 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
11244 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
11245 fix<uns>_trunc<SFDF:mode>si2_p8.
11246 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
11247 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
11248 (fix<uns>_<mode>_mem): Likewise.
11249 (fctiw<u>z_<mode>_mem): Likewise.
11250 (fix<uns>_<mode>_mem): Likewise.
11251 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
11252 the register allocator from doing a direct move to the GPRs to do
11253 a store, and instead use the ISA 3.0 store byte/half-word from
11254 vector register instruction. For IEEE 128-bit floating point,
11255 also optimize stores of 32-bit ints.
11256 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
11258 2018-02-07 Alan Hayward <alan.hayward@arm.com>
11260 * genextract.c (push_pathstr_operand): New function to support
11262 (walk_rtx): Call push_pathstr_operand.
11263 (print_path): Support [a-zA-Z].
11265 2018-02-07 Richard Biener <rguenther@suse.de>
11267 PR tree-optimization/84037
11268 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
11269 (cse_and_gimplify_to_preheader): Declare.
11270 (vect_get_place_in_interleaving_chain): Likewise.
11271 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11273 (_loop_vec_info::~_loop_vec_info): Delete it.
11274 (cse_and_gimplify_to_preheader): New function.
11275 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
11276 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
11277 (vectorizable_load): Likewise. For grouped stores always base
11278 the IV on the first element.
11279 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
11280 condition before gimplifying.
11282 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11284 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
11285 *DIV_EXPR and *MOD_EXPR.
11287 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
11290 * config/i386/i386.c (ix86_option_override_internal): Mask out
11291 the CF_SET bit when checking -fcf-protection.
11293 2018-02-07 Tom de Vries <tom@codesourcery.com>
11296 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
11299 2018-02-07 Richard Biener <rguenther@suse.de>
11301 PR tree-optimization/84204
11302 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
11305 PR tree-optimization/84205
11306 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
11307 special-case isl_ast_op_zdiv_r.
11309 PR tree-optimization/84223
11310 * graphite-scop-detection.c (gather_bbs::before_dom_children):
11311 Only add conditions from within the region.
11312 (gather_bbs::after_dom_children): Adjust.
11314 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
11317 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
11318 * config/avr/avr.md: Only post-reload split REG-REG moves if
11319 either register is GENERAL_REG_P.
11321 2018-02-07 Jakub Jelinek <jakub@redhat.com>
11323 PR tree-optimization/84235
11324 * tree-ssa-scopedtables.c
11325 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
11326 if the subtraction is performed in floating point type where NaNs are
11327 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
11328 build 1. Formatting fix.
11330 2018-02-06 Jakub Jelinek <jakub@redhat.com>
11333 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
11334 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
11335 and skip it regardless of bb boundaries. Use CALL_P macro,
11336 don't test INSN_P (insn) together with CALL_P or JUMP_P check
11337 unnecessarily, formatting fix.
11339 2018-02-06 Michael Collison <michael.collison@arm.com>
11341 * config/arm/thumb2.md:
11342 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
11343 (*thumb_mov_notscc): Ditto.
11345 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
11348 * config/rs6000/rs6000.md (su code attribute): Use "u" for
11349 unsigned_fix, not "s".
11351 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11353 * configure.ac (gcc_fn_eh_frame_ro): New function.
11354 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
11355 correct .eh_frame permissions.
11356 * configure: Regenerate.
11358 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
11360 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
11361 irrelevant options.
11363 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11365 * config/rs6000/rs6000.c (rs6000_option_override_internal):
11366 Display warning message for -mno-speculate-indirect-jumps.
11368 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
11370 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
11372 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
11374 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
11376 PR tree-optimization/84225
11377 * tree-eh.c (find_trapping_overflow): Only call
11378 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
11380 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11383 * config/i386/i386.c: Reimplement the check of possible options
11384 -mibt/-mshstk conbination. Change error messages.
11385 * doc/invoke.texi: Fix a typo: remove extra '='.
11387 2018-02-06 Marek Polacek <polacek@redhat.com>
11389 PR tree-optimization/84228
11390 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
11392 2018-02-06 Tamar Christina <tamar.christina@arm.com>
11395 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
11396 emitted arch directives.
11397 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
11398 __ARM_FEATURE_COPROC before changing architectures.
11400 2018-02-06 Richard Biener <rguenther@suse.de>
11402 * config/i386/i386.c (print_reg): Fix typo.
11403 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
11405 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
11407 * configure: Regenerate.
11409 2018-02-05 Martin Sebor <msebor@redhat.com>
11411 PR tree-optimization/83369
11412 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
11415 2018-02-05 Martin Liska <mliska@suse.cz>
11417 * doc/invoke.texi: Cherry-pick upstream r323995.
11419 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
11421 * ira.c (ira_init_register_move_cost): Adjust comment.
11423 2018-02-05 Martin Liska <mliska@suse.cz>
11425 PR gcov-profile/84137
11426 * doc/gcov.texi: Fix typo in documentation.
11428 2018-02-05 Martin Liska <mliska@suse.cz>
11430 PR gcov-profile/83879
11431 * doc/gcov.texi: Document necessity of --dynamic-list-data when
11432 using dlopen functionality.
11434 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
11436 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
11437 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
11438 _mm_maskz_range_ss, _mm_mask_range_round_ss,
11439 _mm_maskz_range_round_ss): New intrinsics.
11440 (__builtin_ia32_rangesd128_round)
11441 (__builtin_ia32_rangess128_round): Remove.
11442 (__builtin_ia32_rangesd128_mask_round,
11443 __builtin_ia32_rangess128_mask_round): New builtins.
11444 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
11445 __builtin_ia32_rangess128_round): Remove.
11446 (__builtin_ia32_rangesd128_mask_round,
11447 __builtin_ia32_rangess128_mask_round): New builtins.
11448 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
11449 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
11450 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
11451 "<round_saeonly_constraint>")): Changed to ...
11452 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
11453 "<round_saeonly_scalar_constraint>")): ... this.
11454 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
11455 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
11456 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
11457 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
11458 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
11460 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
11462 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
11464 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
11465 Remove all values except native, 8540 and 8548.
11467 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
11469 * config/i386/i386.c (ix86_output_function_return): Pass
11470 INVALID_REGNUM, instead of -1, as invalid register number to
11471 indirect_thunk_name and output_indirect_thunk.
11473 2018-02-02 Julia Koval <julia.koval@intel.com>
11475 * config.gcc: Add -march=icelake.
11476 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
11477 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
11478 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
11479 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
11480 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
11481 (processor_target_table): Add icelake.
11482 (ix86_option_override_internal): Handle new PTAs.
11483 (get_builtin_code_for_version): Handle icelake.
11484 (M_INTEL_COREI7_ICELAKE): New.
11485 (fold_builtin_cpu): Handle icelake.
11486 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
11487 * doc/invoke.texi: Add -march=icelake.
11489 2018-02-02 Julia Koval <julia.koval@intel.com>
11491 * config/i386/i386.c (ix86_option_override_internal): Change flags type
11492 to wide_int_bitmask.
11493 * wide-int-bitmask.h: New.
11495 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
11498 * config/i386/i386.md: Replace Pmode with word_mode in
11499 builtin_setjmp_setup and builtin_longjmp to support x32.
11501 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
11505 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
11507 (rs6000_supported_cpu_names): New static variable.
11508 (linux_cpu_translation_table): Likewise.
11509 (elf_platform) <cpu>: Define new static variable and use it.
11510 Translate kernel AT_PLATFORM name to canonical name if needed.
11511 Error if platform name is unknown.
11513 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
11516 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
11518 2018-02-01 Jeff Law <law@redhat.com>
11521 * config/i386/i386.c (release_scratch_register_on_entry): Add new
11522 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11523 the scratch if RELEASE_VIA_POP is false.
11524 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11525 If we have to save a temporary register, decrement SIZE appropriately.
11526 Pass new arguments to release_scratch_register_on_entry.
11527 (ix86_adjust_stack_and_probe): Likewise.
11528 (ix86_emit_probe_stack_range): Pass new arguments to
11529 release_scratch_register_on_entry.
11531 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11533 PR rtl-optimization/84157
11534 * combine.c (change_zero_ext): Use REG_P predicate in
11535 front of HARD_REGISTER_P predicate.
11537 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11539 * config/avr/avr.c (avr_option_override): Move disabling of
11540 -fdelete-null-pointer-checks to...
11541 * common/config/avr/avr-common.c (avr_option_optimization_table):
11544 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11546 PR tree-optimization/81635
11547 * tree-data-ref.c (split_constant_offset_1): For types that
11548 wrap on overflow, try to use range info to prove that wrapping
11551 2018-02-01 Renlin Li <renlin.li@arm.com>
11554 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11555 TAILCALL_ADDR_REGS.
11556 (aarch64_register_move_cost): Likewise.
11557 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11558 TAILCALL_ADDR_REGS.
11559 (REG_CLASS_NAMES): Likewise.
11560 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11561 TAILCALL_ADDR_REGS. Remove IP registers.
11562 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11564 2018-02-01 Richard Biener <rguenther@suse.de>
11566 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11567 for specifying RPO order and allow NULL for that.
11568 * domwalk.c (dom_walker::dom_walker): Likewise.
11569 (dom_walker::walk): Handle NULL RPO order.
11570 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11572 (rewrite_update_dom_walker): Likewise.
11573 (mark_def_dom_walker): Likewise.
11575 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11577 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11578 (aarch64_maybe_expand_sve_subreg_move): Declare.
11579 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11580 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11582 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11583 that are semantically a reverse operation.
11584 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11585 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11586 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11588 (aarch64_can_change_mode_class): For big-endian, forbid changes
11589 between two SVE modes if they have different element sizes.
11591 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11593 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11594 the TImode handling for big-endian targets.
11596 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11598 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11599 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11601 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11602 Remove BSWAP handing for big-endian targets and use the form of
11603 LD1RQ appropariate for the mode.
11605 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11607 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11608 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11609 duplicated element.
11611 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11614 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11615 check for operands that need to go through aarch64_sve_reload_be.
11617 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11619 PR tree-optimization/81661
11620 PR tree-optimization/84117
11621 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11622 * tree-eh.c: Include gimplify.h.
11623 (find_trapping_overflow, replace_trapping_overflow,
11624 rewrite_to_non_trapping_overflow): New functions.
11625 * tree-vect-loop.c: Include tree-eh.h.
11626 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11627 * tree-data-ref.c: Include tree-eh.h.
11628 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11630 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11632 PR rtl-optimization/84123
11633 * combine.c (change_zero_ext): Check if hard register satisfies
11634 can_change_dest_mode before calling gen_lowpart_SUBREG.
11636 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11639 * ira.c (ira_init_register_move_cost): Remove assert.
11641 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11643 PR rtl-optimization/84071
11644 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11645 * doc/tm.texi: Regenerate.
11647 2018-01-31 Richard Biener <rguenther@suse.de>
11649 PR tree-optimization/84132
11650 * tree-data-ref.c (analyze_miv_subscript): Properly
11651 check whether evolution_function_is_affine_multivariate_p
11652 before calling gcd_of_steps_may_divide_p.
11654 2018-01-31 Julia Koval <julia.koval@intel.com>
11657 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11658 * config/i386/i386.md (rdpid_rex64) New.
11659 (rdpid): Make 32bit only.
11661 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11664 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11665 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11667 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11670 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11672 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11674 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11676 PR rtl-optimization/84071
11677 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11678 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11680 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11682 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11683 (arc_attribute_table): Add 'aux' attribute.
11684 (arc_in_small_data_p): Consider aux like variables.
11685 (arc_is_aux_reg_p): New function.
11686 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11687 (arc_get_aux_arg): New function.
11688 (prepare_move_operands): Handle aux-register access.
11689 (arc_handle_aux_attribute): New function.
11690 * doc/extend.texi (ARC Variable attributes): Add subsection.
11692 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11694 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11695 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11696 (arc_attribute_table): Add 'uncached' attribute.
11697 (arc_print_operand): Print '.di' flag for uncached memory
11699 (arc_in_small_data_p): Do not consider for small data the uncached
11701 (arc_is_uncached_mem_p): New function.
11702 * config/arc/predicates.md (compact_store_memory_operand): Check
11703 for uncached memory accesses.
11704 (nonvol_nonimm_operand): Likewise.
11705 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11707 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11710 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11711 falign-loops=): Add Optimization flag.
11713 2018-01-30 Jeff Law <law@redhat.com>
11716 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11717 INT_REGISTERS_SAVED. Check it prior to calling
11718 get_scratch_register_on_entry.
11719 (ix86_adjust_stack_and_probe): Similarly.
11720 (ix86_emit_probe_stack_range): Similarly.
11721 (ix86_expand_prologue): Corresponding changes.
11723 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11726 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11727 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11729 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11732 * lra-constraints.c (curr_insn_transform): Process AND in the
11735 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11737 PR rtl-optimization/83986
11738 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11739 dependence against last_pending_memory_flush in addition to
11740 pending_jump_insns.
11742 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11744 PR tree-optimization/81611
11745 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11748 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11751 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11754 2018-01-30 Richard Biener <rguenther@suse.de>
11755 Jakub Jelinek <jakub@redhat.com>
11757 PR tree-optimization/84111
11758 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11759 inner loops added during recursion, as they don't have up-to-date
11762 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11765 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11766 (can_inline_edge_by_limits_p): ... here.
11767 (can_early_inline_edge_p, check_callers,
11768 update_caller_keys, update_callee_keys, recursive_inlining,
11769 add_new_edges_to_heap, speculation_useful_p,
11770 inline_small_functions,
11771 inline_small_functions, flatten_function,
11772 inline_to_all_callers_1): Update.
11774 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11776 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11779 2018-01-30 Richard Biener <rguenther@suse.de>
11781 PR tree-optimization/83008
11782 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11783 invariant and constant vector uses in stmts when they need
11784 more than one stmt.
11786 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11789 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11790 * configure: Regenerate.
11792 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11794 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11796 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11797 Use gen_rtx_REG rather than gen_lowpart.
11799 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11801 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11802 rather than 0 when creating partial subregs.
11804 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11806 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11809 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11812 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11813 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11814 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11815 flags. This restores the settings used before the 2017-07-24.
11816 Turning off pre increment/decrement/modify allows IVOPTS to
11817 optimize DF/SF loops where the index is an int.
11819 2018-01-29 Richard Biener <rguenther@suse.de>
11820 Kelvin Nilsen <kelvin@gcc.gnu.org>
11823 * tree-vect-stmts.c (vectorizable_call): Don't call
11824 targetm.vectorize_builtin_md_vectorized_function if callee is
11827 2018-01-22 Carl Love <cel@us.ibm.com>
11829 * doc/extend.tex: Fix typo in second arg in
11830 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11832 2018-01-29 Richard Biener <rguenther@suse.de>
11834 PR tree-optimization/84086
11835 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11836 (flush_ssaname_freelist): When SSA names were released reset
11837 the SCEV hash table.
11839 2018-01-29 Richard Biener <rguenther@suse.de>
11841 PR tree-optimization/84057
11842 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11843 removed paths when removing edges.
11845 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11847 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11848 -mfunction-return=@var{choice}.
11850 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11852 PR diagnostic/84034
11853 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11854 Handle CR like TAB.
11855 (layout::print_source_line): Likewise.
11856 (test_get_line_width_without_trailing_whitespace): Add test cases.
11858 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11860 PR middle-end/84040
11861 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11864 2018-01-26 Jim Wilson <jimw@sifive.com>
11866 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11868 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11871 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11873 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11874 and CMP + SUB-immediate -> SUBS.
11876 2018-01-26 Martin Sebor <msebor@redhat.com>
11878 PR tree-optimization/83896
11879 * tree-ssa-strlen.c (get_string_len): Rename...
11880 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11881 Avoid assuming length is constant.
11882 (handle_char_store): Use HOST_WIDE_INT for string length.
11884 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11887 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11888 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11890 2018-01-26 Richard Biener <rguenther@suse.de>
11892 PR rtl-optimization/84003
11893 * dse.c (record_store): Only record redundant stores when
11894 the earlier store aliases at least all accesses the later one does.
11896 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11898 PR rtl-optimization/83985
11899 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11900 REG_CFA_RESTORE insns.
11901 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11902 REG_CFA_RESTORE insns here.
11905 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11906 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11908 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11910 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11911 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11912 (arc_init): Likewise.
11913 (arc_override_options): Likewise.
11914 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11916 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11918 * config/arc/arc.h (TARGET_DBNZ): Define.
11919 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11920 properly set the tune attribute.
11921 (dbnz): Use TARGET_DBNZ guard.
11922 * config/arc/arc.opt (mtune): Add core3 option.
11924 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11926 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11927 recognize new pic like addresses.
11928 (arc_delegitimize_address): Clean up.
11930 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11932 * config/arc/arc-arches.def: Option mrf16 valid for all
11934 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11935 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11936 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11937 * config/arc/arc-tables.opt: Regenerate.
11938 * config/arc/arc.c (arc_conditional_register_usage): Handle
11939 reduced register file case.
11940 (arc_file_start): Set must have build attributes.
11941 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11942 mrf16 option value.
11943 * config/arc/arc.opt (mrf16): Add new option.
11944 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11945 * config/arc/genmultilib.awk: Handle new mrf16 option.
11946 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11947 * config/arc/t-multilib: Regenerate.
11948 * doc/invoke.texi (ARC Options): Document mrf16 option.
11950 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11952 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11953 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11954 (arc_attribute_table): Add 'secure_call' attribute.
11955 (arc_print_operand): Print secure call operand.
11956 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11958 (arc_is_secure_call_p): New function. * config/arc/arc.md
11959 (call_i): Add support for sjli instruction.
11960 (call_value_i): Likewise.
11961 * config/arc/constraints.md (Csc): New constraint.
11963 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11964 John Eric Martin <John.Martin@emmicro-us.com>
11966 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11967 * config/arc/arc.c (_arc_jli_section): New struct.
11968 (arc_jli_section): New type.
11969 (rc_jli_sections): New static variable.
11970 (arc_handle_jli_attribute): New function.
11971 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11972 (arc_file_end): New function.
11973 (TARGET_ASM_FILE_END): Define.
11974 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11975 (arc_add_jli_section): New function.
11976 (jli_call_scan): Likewise.
11977 (arc_reorg): Call jli_call_scan.
11978 (arc_output_addsi): Remove 'S' from printing asm operand.
11979 (arc_is_jli_call_p): New function.
11980 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11982 (movhi_insn): Likewise.
11983 (movsi_insn): Likewise.
11984 (movsi_set_cc_insn): Likewise.
11985 (loadqi_update): Likewise.
11986 (load_zeroextendqisi_update): Likewise.
11987 (load_signextendqisi_update): Likewise.
11988 (loadhi_update): Likewise.
11989 (load_zeroextendhisi_update): Likewise.
11990 (load_signextendhisi_update): Likewise.
11991 (loadsi_update): Likewise.
11992 (loadsf_update): Likewise.
11993 (movsicc_insn): Likewise.
11994 (bset_insn): Likewise.
11995 (bxor_insn): Likewise.
11996 (bclr_insn): Likewise.
11997 (bmsk_insn): Likewise.
11998 (bicsi3_insn): Likewise.
11999 (cmpsi_cc_c_insn): Likewise.
12000 (movsi_ne): Likewise.
12001 (movsi_cond_exec): Likewise.
12002 (clrsbsi2): Likewise.
12003 (norm_f): Likewise.
12013 (call_i): Remove 'S' asm letter, add jli instruction.
12014 (call_value_i): Likewise.
12015 * config/arc/arc.op (mjli-always): New option.
12016 * config/arc/constraints.md (Cji): New constraint.
12017 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
12019 (subsf3_fpx): Likewise.
12020 (mulsf3_fpx): Likewise.
12021 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
12023 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
12024 function attrbutes.
12025 * doc/invoke.texi (ARC): Document mjli-always option.
12027 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
12029 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
12030 avoid addition with 0 and use incw and decw where possible.
12032 2018-01-26 Richard Biener <rguenther@suse.de>
12034 PR tree-optimization/81082
12035 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
12036 association if it requires casting to unsigned.
12037 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
12038 from fold_plusminus_mult_expr to catch important cases late when
12039 range info is available.
12041 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12043 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
12044 * configure.ac (hidden_linkonce): New test.
12045 * configure: Regenerate.
12046 * config.in: Regenerate.
12048 2018-01-26 Julia Koval <julia.koval@intel.com>
12050 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
12051 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
12052 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
12053 _mm_mask_bitshuffle_epi64_mask): Fix type.
12054 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
12055 USI_FTYPE_V4DI_V4DI_USI): Remove.
12056 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
12057 __builtin_ia32_vpshufbitqmb256_mask,
12058 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
12059 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
12060 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
12062 2018-01-26 Alan Modra <amodra@gmail.com>
12065 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
12066 UNSPEC_VBPERMQ. Sort other unspecs.
12068 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
12070 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
12072 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
12074 PR middle-end/83055
12075 * predict.c (drop_profile): Do not push/pop cfun; update also
12077 (handle_missing_profiles): Fix logic looking for zero profiles.
12079 2018-01-25 Jakub Jelinek <jakub@redhat.com>
12081 PR middle-end/83977
12082 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
12083 on functions with #pragma omp declare simd or functions with simd
12085 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
12086 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
12087 Remove trailing \n from warning_at calls.
12089 2018-01-25 Tom de Vries <tom@codesourcery.com>
12092 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12093 for neutered workers.
12095 2018-01-24 Joseph Myers <joseph@codesourcery.com>
12098 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
12099 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
12101 2018-01-24 Jeff Law <law@redhat.com>
12104 * i386.c (get_probe_interval): Move to earlier point.
12105 (ix86_compute_frame_layout): If -fstack-clash-protection and
12106 the frame is larger than the probe interval, then use pushes
12107 to save registers rather than reg->mem moves.
12108 (ix86_expand_prologue): Remove conditional for int_registers_saved
12111 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
12114 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
12115 min/max for never referenced object.
12117 2018-01-24 Jakub Jelinek <jakub@redhat.com>
12119 PR middle-end/83977
12120 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
12122 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
12123 attributes from DECL_ATTRIBUTES (decl) without affecting
12124 DECL_ATTRIBUTES (current_function_decl).
12125 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
12126 functions with non-NULL DECL_ABSTRACT_ORIGIN.
12128 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
12130 PR tree-optimization/83979
12131 * fold-const.c (fold_comparison): Use constant_boolean_node
12132 instead of boolean_{true,false}_node.
12134 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
12136 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
12139 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12141 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12142 Simplify the clause that sets the length attribute.
12143 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12144 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
12145 clause that sets the length attribute.
12146 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12148 2018-01-24 Tom de Vries <tom@codesourcery.com>
12151 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
12152 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
12153 Add strict parameter.
12154 (prevent_branch_around_nothing): Insert dummy insn between branch to
12155 label and label with no ptx insn inbetween.
12156 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
12158 2018-01-24 Tom de Vries <tom@codesourcery.com>
12161 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
12162 for neutered threads in warp.
12163 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
12165 2018-01-24 Richard Biener <rguenther@suse.de>
12167 PR tree-optimization/83176
12168 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
12171 2018-01-24 Richard Biener <rguenther@suse.de>
12173 PR tree-optimization/82819
12174 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
12175 code generating pluses that are no-ops in the target precision.
12177 2018-01-24 Richard Biener <rguenther@suse.de>
12179 PR middle-end/84000
12180 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
12182 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12184 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
12185 to merge probabilities.
12186 * predict.c (probably_never_executed): Also mark as cold functions
12187 with global 0 profile and guessed local profile.
12188 * profile-count.c (profile_probability::combine_with_count): New
12190 * profile-count.h (profile_probability::operator*,
12191 profile_probability::operator*=, profile_probability::operator/,
12192 profile_probability::operator/=): Reduce precision to adjusted
12193 and set value to guessed on contradictory divisions.
12194 (profile_probability::combine_with_freq): Remove.
12195 (profile_probability::combine_wiht_count): Declare.
12196 (profile_count::force_nonzero):: Set to adjusted.
12197 (profile_count::probability_in):: Set quality to adjusted.
12198 * tree-ssa-tail-merge.c (replace_block_by): Use
12199 combine_with_count.
12201 2018-01-23 Andrew Waterman <andrew@sifive.com>
12202 Jim Wilson <jimw@sifive.com>
12204 * config/riscv/riscv.c (riscv_stack_boundary): New.
12205 (riscv_option_override): Set riscv_stack_boundary. Handle
12206 riscv_preferred_stack_boundary_arg.
12207 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
12208 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
12209 (STACK_BOUNDARY): Set to riscv_stack_boundary.
12210 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
12211 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
12212 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
12214 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
12217 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
12218 of struct ix86_frame.
12219 (ix86_expand_epilogue): Likewise. Add a local variable for
12220 the reg_save_offset field in struct ix86_frame.
12222 2018-01-23 Bin Cheng <bin.cheng@arm.com>
12224 PR tree-optimization/82604
12225 * tree-loop-distribution.c (enum partition_kind): New enum item
12226 PKIND_PARTIAL_MEMSET.
12227 (partition_builtin_p): Support above new enum item.
12228 (generate_code_for_partition): Ditto.
12229 (compute_access_range): Differentiate cases that equality can be
12230 proven at all loops, the innermost loops or no loops.
12231 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
12232 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
12233 (finalize_partitions, distribute_loop): Don't fuse partition of
12234 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
12235 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
12236 parloop is enabled.
12238 2018-01-23 Martin Liska <mliska@suse.cz>
12240 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
12241 order to ignore the predictor.
12242 (PRED_POLYMORPHIC_CALL): Likewise.
12243 (PRED_RECURSIVE_CALL): Likewise.
12245 2018-01-23 Martin Liska <mliska@suse.cz>
12247 * tree-profile.c (tree_profiling): Print function header to
12248 aware reader which function we are working on.
12249 * value-prof.c (gimple_find_values_to_profile): Do not print
12250 not interesting value histograms.
12252 2018-01-23 Martin Liska <mliska@suse.cz>
12254 * profile-count.h (enum profile_quality): Add
12255 profile_uninitialized as the first value. Do not number values
12256 as they are zero based.
12257 (profile_count::verify): Update sanity check.
12258 (profile_probability::verify): Likewise.
12260 2018-01-23 Nathan Sidwell <nathan@acm.org>
12262 * doc/invoke.texi (ffor-scope): Deprecate.
12264 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12266 PR tree-optimization/83510
12267 * domwalk.c (set_all_edges_as_executable): New function.
12268 (dom_walker::dom_walker): Convert bool param
12269 "skip_unreachable_blocks" to enum reachability. Move setup of
12270 edge flags to set_all_edges_as_executable and only do it when
12271 reachability is REACHABLE_BLOCKS.
12272 * domwalk.h (enum dom_walker::reachability): New enum.
12273 (dom_walker::dom_walker): Convert bool param
12274 "skip_unreachable_blocks" to enum reachability.
12275 (set_all_edges_as_executable): New decl.
12276 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
12277 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
12279 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
12280 but converting true to REACHABLE_BLOCKS.
12281 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
12283 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
12284 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
12285 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
12287 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
12288 if check_all_array_refs will be called.
12290 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12292 * tree.c (selftest::test_location_wrappers): Add more test
12295 2018-01-23 David Malcolm <dmalcolm@redhat.com>
12297 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
12298 (selftest::test_bit_in_range): Likewise.
12300 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12303 * doc/sourcebuild.texi (vect_float): Say that the selector
12304 only describes the situation when -funsafe-math-optimizations is on.
12305 (vect_float_strict): Document.
12307 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
12309 PR tree-optimization/83965
12310 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
12311 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
12312 instead of checking only for a reduction.
12313 (vect_recog_widen_sum_pattern): Likewise.
12315 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12317 * predict.c (probably_never_executed): Only use precise profile info.
12318 (compute_function_frequency): Skip after inlining hack since we now
12319 have quality checking.
12321 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
12323 * profile-count.h (profile_probability::very_unlikely,
12324 profile_probability::unlikely, profile_probability::even): Set
12325 precision to guessed.
12327 2018-01-23 Richard Biener <rguenther@suse.de>
12329 PR tree-optimization/83963
12330 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
12331 Properly terminate dominator walk when crossing the exit edge not
12332 when visiting its source block.
12334 2018-01-23 Jakub Jelinek <jakub@redhat.com>
12337 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
12338 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
12340 2018-01-22 Jakub Jelinek <jakub@redhat.com>
12342 PR tree-optimization/83957
12343 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
12344 semicolon after for body surrounded by braces.
12346 PR tree-optimization/83081
12347 * profile-count.h (profile_probability::split): New method.
12348 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
12349 Use profile_probability::split.
12350 (do_compare_rtx_and_jump): Fix adjustment of probabilities
12351 when splitting a single conditional jump into 2.
12353 2018-01-22 David Malcolm <dmalcolm@redhat.com>
12355 PR tree-optimization/69452
12356 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
12359 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12361 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
12362 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
12363 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
12365 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12367 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
12369 * config/rl78/rl78.md (movdi): New define_expand.
12370 * config/rl78/rl78.c (rl78_split_movdi): New function.
12372 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
12375 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
12377 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
12378 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
12379 128-bit to produce an UNSPEC move to get the double word with the
12380 signbit and then a shift directly to do signbit.
12381 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
12382 implementation with a new version that just does either a direct
12383 move or a regular move. Move memory interface to separate insns.
12384 Move insns so they are next to the expander.
12385 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
12386 with signbit move. Split big and little endian case.
12387 (signbit<mode>2_dm_mem_le): Likewise.
12388 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
12389 (signbit<mode>2_dm2): Likewise.
12391 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12393 * config/rl78/rl78.md (anddi3): New define_expand.
12395 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12397 * config/rl78/rl78.md (umindi3): New define_expand.
12399 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12401 * config/rl78/rl78.md (smindi3): New define_expand.
12403 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12405 * config/rl78/rl78.md (smaxdi3): New define_expand.
12407 2018-01-22 Carl Love <cel@us.ibm.com>
12409 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
12410 LVX_V1TI): Add macro expansion.
12411 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
12412 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
12413 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
12414 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
12415 Change check to determine if the instruction is a byte reversing
12416 entry. Fix typo in comment.
12417 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
12418 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
12419 Add def_builtin calls for new builtins.
12420 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
12421 Add define_insn expansion.
12423 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12425 * config/rl78/rl78.md (umaxdi3): New define_expand.
12427 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
12429 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
12430 for non-QImode registers.
12432 2018-01-22 Richard Biener <rguenther@suse.de>
12434 PR tree-optimization/83963
12435 * graphite-scop-detection.c (scop_detection::get_sese): Delay
12436 including the loop exit block.
12437 (scop_detection::merge_sese): Likewise.
12438 (scop_detection::add_scop): Do it here instead.
12440 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12442 * doc/sourcebuild.texi (arm_softfloat): Document.
12444 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
12447 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
12448 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
12449 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
12451 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12452 David Edelsohn <dje.gcc@gmail.com>
12455 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12456 Change "crset eq" to "crset 2".
12457 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12458 (*call_indirect_aix<mode>_nospec): Likewise.
12459 (*call_value_indirect_aix<mode>_nospec): Likewise.
12460 (*call_indirect_elfv2<mode>_nospec): Likewise.
12461 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
12462 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
12463 change assembly output from . to $.
12464 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12465 (indirect_jump<mode>_nospec): Change assembly output from . to $.
12466 (*tablejump<mode>_internal1_nospec): Likewise.
12468 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
12471 * config/sh/sh_optimize_sett_clrt.cc:
12472 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
12474 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
12476 PR tree-optimization/83940
12477 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
12478 offset_dt to vect_constant_def rather than vect_unknown_def_type.
12479 (vect_check_load_store_mask): Add a mask_dt_out parameter and
12480 use it to pass back the definition type.
12481 (vect_check_store_rhs): Likewise rhs_dt_out.
12482 (vect_build_gather_load_calls): Add a mask_dt argument and use
12483 it instead of a call to vect_is_simple_use.
12484 (vectorizable_store): Update calls to vect_check_load_store_mask
12485 and vect_check_store_rhs. Use the dt returned by the latter instead
12486 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
12487 instead of calls to vect_is_simple_use. Pass the scalar rather
12488 than the vector operand to vect_is_simple_use when handling
12489 second and subsequent copies of an rhs value.
12490 (vectorizable_load): Update calls to vect_check_load_store_mask
12491 and vect_build_gather_load_calls. Use the cached mask_dt and
12492 gs_info.offset_dt instead of calls to vect_is_simple_use.
12494 2018-01-20 Jakub Jelinek <jakub@redhat.com>
12496 PR middle-end/83945
12497 * tree-emutls.c: Include gimplify.h.
12498 (lower_emutls_2): New function.
12499 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
12500 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
12501 it before further processing.
12504 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
12505 UINTVAL (trueop1) instead of INTVAL (op1).
12507 2018-01-19 Jakub Jelinek <jakub@redhat.com>
12511 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
12512 INCOMING_FRAME_SP_OFFSET if not defined.
12513 (scan_trace): Add ENTRY argument. If true and
12514 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
12515 emit a note to adjust the CFA offset.
12516 (create_cfi_notes): Adjust scan_trace callers.
12517 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
12518 INCOMING_FRAME_SP_OFFSET in the CIE.
12519 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
12520 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
12522 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12523 * doc/tm.texi: Regenerated.
12525 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12527 PR rtl-optimization/83147
12528 * lra-constraints.c (remove_inheritance_pseudos): Use
12529 lra_substitute_pseudo_within_insn.
12531 2018-01-19 Tom de Vries <tom@codesourcery.com>
12532 Cesar Philippidis <cesar@codesourcery.com>
12535 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12537 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12540 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12541 spaces for function labels.
12543 2018-01-19 Martin Liska <mliska@suse.cz>
12545 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12546 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12547 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12548 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12549 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12550 (PRED_CONST_RETURN): Change from 69 to 65.
12551 (PRED_NULL_RETURN): Change from 91 to 71.
12552 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12553 (PRED_LOOP_GUARD): Change from 66 to 73.
12555 2018-01-19 Martin Liska <mliska@suse.cz>
12557 * predict.c (predict_insn_def): Add new assert.
12558 (struct branch_predictor): Change type to signed integer.
12559 (test_prediction_value_range): Amend test to cover
12560 PROB_UNINITIALIZED.
12561 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12562 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12563 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12564 (PRED_LOOP_IV_COMPARE): Likewise.
12565 * predict.h (PROB_UNINITIALIZED): Define new constant.
12567 2018-01-19 Martin Liska <mliska@suse.cz>
12569 * predict.c (dump_prediction): Add new format for
12570 analyze_brprob.py script which is enabled with -details
12572 * profile-count.h (precise_p): New function.
12574 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12576 PR tree-optimization/83922
12577 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12578 there are no statements that need masking.
12579 (vect_active_double_reduction_p): New function.
12580 (vect_analyze_loop_operations): Use it when handling phis that
12581 are not in the loop header.
12583 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12585 PR tree-optimization/83914
12586 * tree-vect-loop.c (vectorizable_induction): Don't convert
12587 init_expr or apply the peeling adjustment for inductions
12588 that are nested within the vectorized loop.
12590 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12592 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12595 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12599 * function.h (gimplify_parameters): Add gimple_seq * argument.
12600 * function.c: Include gimple.h and options.h.
12601 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12602 for the added local temporaries if needed.
12603 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12604 if there are any parameter cleanups, wrap whole body into a
12605 try/finally with the cleanups.
12607 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12610 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12611 Use GET_MODE_CLASS for scalar floating point.
12613 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12617 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12618 Fix call of call_cgraph_insertion_hooks.
12620 2018-01-18 Martin Sebor <msebor@redhat.com>
12622 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12624 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12627 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12630 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12633 * common.opt: (-ffile-prefix-map): New option.
12634 * opts.c (common_handle_option): Defer it.
12635 * opts-global.c (handle_common_deferred_options): Handle it.
12636 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12637 * file-prefix-map.h: New file.
12638 (remap_debug_filename, add_debug_prefix_map): ...here.
12639 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12640 * final.c (debug_prefix_map, add_debug_prefix_map
12641 remap_debug_filename): Move to...
12642 * file-prefix-map.c: New file.
12643 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12644 generalize, get rid of alloca(), use strrchr() instead of strchr().
12645 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12646 Implement in terms of add_prefix_map().
12647 (remap_macro_filename, remap_debug_filename): Implement in term of
12649 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12650 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12651 * dbxout.c: Include file-prefix-map.h.
12652 * varasm.c: Likewise.
12653 * vmsdbgout.c: Likewise.
12654 * xcoffout.c: Likewise.
12655 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12656 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12657 * doc/invoke.texi (-ffile-prefix-map): Document.
12658 (-fdebug-prefix-map): Update description.
12660 2018-01-18 Martin Liska <mliska@suse.cz>
12662 * config/i386/i386.c (indirect_thunk_name): Document that also
12664 (output_indirect_thunk): Document why both instructions
12665 (pause and lfence) are generated.
12667 2018-01-18 Richard Biener <rguenther@suse.de>
12669 PR tree-optimization/83887
12670 * graphite-scop-detection.c
12671 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12672 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12673 (scop_detection::merge_sese): Re-implement with a flood-fill
12674 algorithm that properly finds a SESE region if it exists.
12676 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12679 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12680 pointer_diff optimizations use view_convert instead of convert.
12682 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12684 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12685 Generate different code for -mno-speculate-indirect-jumps.
12686 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12687 (*call_indirect_aix<mode>): Disable for
12688 -mno-speculate-indirect-jumps.
12689 (*call_indirect_aix<mode>_nospec): New define_insn.
12690 (*call_value_indirect_aix<mode>): Disable for
12691 -mno-speculate-indirect-jumps.
12692 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12693 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12694 -mno-speculate-indirect-jumps.
12695 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12697 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12699 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12700 long double type, set the flags for noting the default long double
12701 type, even if we don't pass or return a long double type.
12703 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12706 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12709 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12711 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12712 support for merge[hl].
12713 (fold_mergehl_helper): New helper function.
12714 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12715 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12716 (altivec_vmrglw_direct): Add xxmrglw insn.
12718 2018-01-17 Andrew Waterman <andrew@sifive.com>
12720 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12721 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12723 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12726 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12727 call the lto_location_cache before reading the
12728 DECL_SOURCE_LOCATION of the types.
12730 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12731 Richard Sandiford <richard.sandiford@linaro.org>
12733 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12734 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12735 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12736 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12737 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12739 * config/aarch64/constraints.md (aarch64_movti_operand):
12741 * config/aarch64/predicates.md (Uti): Add new constraint.
12743 2018-01-17 Carl Love <cel@us.ibm.com>
12745 * config/rs6000/vsx.md (define_expand xl_len_r,
12746 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12747 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12749 (define_expand, define_insn): Move the shift left from the
12750 define_insn to the define_expand for lxvl and stxvl instructions.
12751 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12752 and XL_LEN_R definitions to PURE.
12754 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12756 * config/i386/i386.c (indirect_thunk_name): Declare regno
12757 as unsigned int. Compare regno with INVALID_REGNUM.
12758 (output_indirect_thunk): Ditto.
12759 (output_indirect_thunk_function): Ditto.
12760 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12761 in the call to output_indirect_thunk_function.
12763 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12765 PR middle-end/83884
12766 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12767 rather than the size of inner_type to determine the stack slot size
12768 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12770 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12773 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12776 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12778 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12779 endian Linux systems to optionally enable multilibs for selecting
12780 the long double type if the user configured an explicit type.
12781 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12782 have no long double multilibs if not defined.
12783 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12784 warn if the user used -mabi={ieee,ibm}longdouble and we built
12785 multilibs for long double.
12786 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12787 appropriate multilib option.
12788 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12790 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12791 for building long double multilibs.
12792 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12794 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12796 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12799 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12801 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12804 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12807 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12810 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12812 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12813 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12815 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12817 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12818 different rtl trees depending on TARGET_64BIT.
12819 (rs6000_gen_lvx): Likewise.
12821 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12823 * config/visium/visium.md (nop): Tweak comment.
12824 (hazard_nop): Likewise.
12826 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12828 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12829 -mspeculate-indirect-jumps.
12830 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12831 for -mno-speculate-indirect-jumps.
12832 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12833 (*call_value_indirect_elfv2<mode>): Disable for
12834 -mno-speculate-indirect-jumps.
12835 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12836 (indirect_jump): Emit different RTL for
12837 -mno-speculate-indirect-jumps.
12838 (*indirect_jump<mode>): Disable for
12839 -mno-speculate-indirect-jumps.
12840 (*indirect_jump<mode>_nospec): New define_insn.
12841 (tablejump): Emit different RTL for
12842 -mno-speculate-indirect-jumps.
12843 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12844 (tablejumpsi_nospec): New define_expand.
12845 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12846 (tablejumpdi_nospec): New define_expand.
12847 (*tablejump<mode>_internal1): Disable for
12848 -mno-speculate-indirect-jumps.
12849 (*tablejump<mode>_internal1_nospec): New define_insn.
12850 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12853 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12855 * caller-save.c (insert_save): Drop unnecessary parameter. All
12858 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12859 Richard Biener <rguenth@suse.de>
12862 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12863 return early, inline manually is_gimple_sizepos. Make sure if we
12864 call gimplify_expr we don't end up with a gimple constant.
12865 * tree.c (variably_modified_type_p): Don't return true for
12866 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12867 * gimplify.h (is_gimple_sizepos): Remove.
12869 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12871 PR tree-optimization/83857
12872 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12873 vectorizable_live_operation for pure SLP statements.
12874 (vectorizable_live_operation): Handle PHIs.
12876 2018-01-16 Richard Biener <rguenther@suse.de>
12878 PR tree-optimization/83867
12879 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12880 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12882 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12885 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12886 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12887 If off is not INTEGER_CST, issue a may not be aligned warning
12888 rather than isn't aligned. Use isn%'t rather than isn't.
12889 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12891 <case MULT_EXPR>: Improve the case when bottom and one of the
12892 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12893 operand, in that case check if the other operand is multiple of
12894 bottom divided by the INTEGER_CST operand.
12896 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12899 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12900 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12901 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12902 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12903 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12904 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12905 (pa_function_arg_size): New function.
12907 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12909 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12910 in a separate statement.
12912 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12914 PR tree-optimization/83847
12915 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12916 group gathers and scatters.
12918 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12920 PR rtl-optimization/86620
12921 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12923 PR rtl-optimization/83213
12924 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12925 to last if both are JUMP_INSNs.
12927 PR tree-optimization/83843
12928 * gimple-ssa-store-merging.c
12929 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12930 store_immediate_info for bswap/nop orig_stores.
12932 2018-01-15 Andrew Waterman <andrew@sifive.com>
12934 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12936 <UDIV>: Increase cost if !TARGET_DIV.
12938 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12940 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12941 (define_attr "cr_logical_3op"): New.
12942 (cceq_ior_compare): Adjust.
12943 (cceq_ior_compare_complement): Adjust.
12944 (*cceq_rev_compare): Adjust.
12945 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12946 (is_cracked_insn): Adjust.
12947 (insn_must_be_first_in_group): Adjust.
12948 * config/rs6000/40x.md: Adjust.
12949 * config/rs6000/440.md: Adjust.
12950 * config/rs6000/476.md: Adjust.
12951 * config/rs6000/601.md: Adjust.
12952 * config/rs6000/603.md: Adjust.
12953 * config/rs6000/6xx.md: Adjust.
12954 * config/rs6000/7450.md: Adjust.
12955 * config/rs6000/7xx.md: Adjust.
12956 * config/rs6000/8540.md: Adjust.
12957 * config/rs6000/cell.md: Adjust.
12958 * config/rs6000/e300c2c3.md: Adjust.
12959 * config/rs6000/e500mc.md: Adjust.
12960 * config/rs6000/e500mc64.md: Adjust.
12961 * config/rs6000/e5500.md: Adjust.
12962 * config/rs6000/e6500.md: Adjust.
12963 * config/rs6000/mpc.md: Adjust.
12964 * config/rs6000/power4.md: Adjust.
12965 * config/rs6000/power5.md: Adjust.
12966 * config/rs6000/power6.md: Adjust.
12967 * config/rs6000/power7.md: Adjust.
12968 * config/rs6000/power8.md: Adjust.
12969 * config/rs6000/power9.md: Adjust.
12970 * config/rs6000/rs64.md: Adjust.
12971 * config/rs6000/titan.md: Adjust.
12973 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12975 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12976 ix86_indirect_branch_register logic.
12978 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12980 * config/i386/constraints.md (Bs): Update
12981 ix86_indirect_branch_register check. Don't check
12982 ix86_indirect_branch_register with GOT_memory_operand.
12984 * config/i386/predicates.md (GOT_memory_operand): Don't check
12985 ix86_indirect_branch_register here.
12986 (GOT32_symbol_operand): Likewise.
12988 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12990 * config/i386/predicates.md (constant_call_address_operand):
12991 Rewrite ix86_indirect_branch_register logic.
12992 (sibcall_insn_operand): Likewise.
12994 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12996 * config/i386/constraints.md (Bs): Replace
12997 ix86_indirect_branch_thunk_register with
12998 ix86_indirect_branch_register.
13000 * config/i386/i386.md (indirect_jump): Likewise.
13001 (tablejump): Likewise.
13002 (*sibcall_memory): Likewise.
13003 (*sibcall_value_memory): Likewise.
13004 Peepholes of indirect call and jump via memory: Likewise.
13005 * config/i386/i386.opt: Likewise.
13006 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13007 (GOT_memory_operand): Likewise.
13008 (call_insn_operand): Likewise.
13009 (sibcall_insn_operand): Likewise.
13010 (GOT32_symbol_operand): Likewise.
13012 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13014 PR middle-end/83837
13015 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
13016 type rather than type addr's type points to.
13017 (expand_omp_atomic_mutex): Likewise.
13018 (expand_omp_atomic): Likewise.
13020 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
13023 * config/i386/i386.c (output_indirect_thunk_function): Use
13024 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
13025 for __x86_return_thunk.
13027 2018-01-15 Richard Biener <rguenther@suse.de>
13029 PR middle-end/83850
13030 * expmed.c (extract_bit_field_1): Fix typo.
13032 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13035 * config/arm/iterators.md (VF): New mode iterator.
13036 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
13037 Remove integer-related logic from pattern.
13038 (neon_vabd<mode>_3): Likewise.
13040 2018-01-15 Jakub Jelinek <jakub@redhat.com>
13042 PR middle-end/82694
13043 * common.opt (fstrict-overflow): No longer an alias.
13044 (fwrapv-pointer): New option.
13045 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
13046 also for pointer types based on flag_wrapv_pointer.
13047 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
13048 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
13049 opts->x_flag_wrapv got set.
13050 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
13051 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
13052 POINTER_TYPE_OVERFLOW_UNDEFINED.
13053 * match.pd: Likewise in address comparison pattern.
13054 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
13056 2018-01-15 Richard Biener <rguenther@suse.de>
13059 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
13060 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
13061 Reset type names to their identifier if their TYPE_DECL doesn't
13062 have linkage (and thus is used for ODR and devirt).
13063 (save_debug_info_for_decl): Remove.
13064 (save_debug_info_for_type): Likewise.
13065 (add_tree_to_fld_list): Adjust.
13066 * tree-pretty-print.c (dump_generic_node): Make dumping of
13067 type names more robust.
13069 2018-01-15 Richard Biener <rguenther@suse.de>
13071 * BASE-VER: Bump to 8.0.1.
13073 2018-01-14 Martin Sebor <msebor@redhat.com>
13076 * builtins.c (check_access): Avoid warning when the no-warning bit
13079 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
13081 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
13082 * ira-color (allocno_hard_regs_compare): Likewise.
13084 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
13087 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
13088 Use .pushsection/.popsection.
13090 2018-01-14 Martin Sebor <msebor@redhat.com>
13093 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
13095 2018-01-14 Jakub Jelinek <jakub@redhat.com>
13097 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
13098 entry from extra_headers.
13099 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
13100 extra_headers, make the list bitwise identical to the i?86-*-* one.
13102 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13104 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
13105 -mcmodel=large with -mindirect-branch=thunk,
13106 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
13107 -mfunction-return=thunk-extern.
13108 * doc/invoke.texi: Document -mcmodel=large is incompatible with
13109 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
13110 -mfunction-return=thunk and -mfunction-return=thunk-extern.
13112 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13114 * config/i386/i386.c (print_reg): Print the name of the full
13115 integer register without '%'.
13116 (ix86_print_operand): Handle 'V'.
13117 * doc/extend.texi: Document 'V' modifier.
13119 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13121 * config/i386/constraints.md (Bs): Disallow memory operand for
13122 -mindirect-branch-register.
13124 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13125 (GOT_memory_operand): Likewise.
13126 (call_insn_operand): Likewise.
13127 (sibcall_insn_operand): Likewise.
13128 (GOT32_symbol_operand): Likewise.
13129 * config/i386/i386.md (indirect_jump): Call convert_memory_address
13130 for -mindirect-branch-register.
13131 (tablejump): Likewise.
13132 (*sibcall_memory): Likewise.
13133 (*sibcall_value_memory): Likewise.
13134 Disallow peepholes of indirect call and jump via memory for
13135 -mindirect-branch-register.
13136 (*call_pop): Replace m with Bw.
13137 (*call_value_pop): Likewise.
13138 (*sibcall_pop_memory): Replace m with Bs.
13139 * config/i386/i386.opt (mindirect-branch-register): New option.
13140 * doc/invoke.texi: Document -mindirect-branch-register option.
13142 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13144 * config/i386/i386-protos.h (ix86_output_function_return): New.
13145 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
13146 set function_return_type.
13147 (indirect_thunk_name): Add ret_p to indicate thunk for function
13149 (output_indirect_thunk_function): Pass false to
13150 indirect_thunk_name.
13151 (ix86_output_indirect_branch_via_reg): Likewise.
13152 (ix86_output_indirect_branch_via_push): Likewise.
13153 (output_indirect_thunk_function): Create alias for function
13154 return thunk if regno < 0.
13155 (ix86_output_function_return): New function.
13156 (ix86_handle_fndecl_attribute): Handle function_return.
13157 (ix86_attribute_table): Add function_return.
13158 * config/i386/i386.h (machine_function): Add
13159 function_return_type.
13160 * config/i386/i386.md (simple_return_internal): Use
13161 ix86_output_function_return.
13162 (simple_return_internal_long): Likewise.
13163 * config/i386/i386.opt (mfunction-return=): New option.
13164 (indirect_branch): Mention -mfunction-return=.
13165 * doc/extend.texi: Document function_return function attribute.
13166 * doc/invoke.texi: Document -mfunction-return= option.
13168 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
13170 * config/i386/i386-opts.h (indirect_branch): New.
13171 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
13172 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
13173 with local indirect jump when converting indirect call and jump.
13174 (ix86_set_indirect_branch_type): New.
13175 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
13176 (indirectlabelno): New.
13177 (indirect_thunk_needed): Likewise.
13178 (indirect_thunk_bnd_needed): Likewise.
13179 (indirect_thunks_used): Likewise.
13180 (indirect_thunks_bnd_used): Likewise.
13181 (INDIRECT_LABEL): Likewise.
13182 (indirect_thunk_name): Likewise.
13183 (output_indirect_thunk): Likewise.
13184 (output_indirect_thunk_function): Likewise.
13185 (ix86_output_indirect_branch_via_reg): Likewise.
13186 (ix86_output_indirect_branch_via_push): Likewise.
13187 (ix86_output_indirect_branch): Likewise.
13188 (ix86_output_indirect_jmp): Likewise.
13189 (ix86_code_end): Call output_indirect_thunk_function if needed.
13190 (ix86_output_call_insn): Call ix86_output_indirect_branch if
13192 (ix86_handle_fndecl_attribute): Handle indirect_branch.
13193 (ix86_attribute_table): Add indirect_branch.
13194 * config/i386/i386.h (machine_function): Add indirect_branch_type
13195 and has_local_indirect_jump.
13196 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
13198 (tablejump): Likewise.
13199 (*indirect_jump): Use ix86_output_indirect_jmp.
13200 (*tablejump_1): Likewise.
13201 (simple_return_indirect_internal): Likewise.
13202 * config/i386/i386.opt (mindirect-branch=): New option.
13203 (indirect_branch): New.
13206 (thunk-inline): Likewise.
13207 (thunk-extern): Likewise.
13208 * doc/extend.texi: Document indirect_branch function attribute.
13209 * doc/invoke.texi: Document -mindirect-branch= option.
13211 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
13214 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
13216 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
13218 * ipa-inline.c (want_inline_small_function_p): Return false if
13219 inlining has already failed with CIF_FINAL_ERROR.
13220 (update_caller_keys): Call want_inline_small_function_p before
13222 (update_callee_keys): Likewise.
13224 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13226 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
13228 (rs6000_quadword_masked_address_p): Likewise.
13229 (quad_aligned_load_p): Likewise.
13230 (quad_aligned_store_p): Likewise.
13231 (const_load_sequence_p): Add comment to describe the outer-most loop.
13232 (mimic_memory_attributes_and_flags): New function.
13233 (rs6000_gen_stvx): Likewise.
13234 (replace_swapped_aligned_store): Likewise.
13235 (rs6000_gen_lvx): Likewise.
13236 (replace_swapped_aligned_load): Likewise.
13237 (replace_swapped_load_constant): Capitalize argument name in
13238 comment describing this function.
13239 (rs6000_analyze_swaps): Add a third pass to search for vector loads
13240 and stores that access quad-word aligned addresses and replace
13241 with stvx or lvx instructions when appropriate.
13242 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
13243 New function prototype.
13244 (rs6000_quadword_masked_address_p): Likewise.
13245 (rs6000_gen_lvx): Likewise.
13246 (rs6000_gen_stvx): Likewise.
13247 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
13248 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
13249 when memory address is aligned.
13250 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
13251 this split to select lvx instruction when memory address is aligned.
13252 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
13253 instruction when memory address is aligned.
13254 (*vsx_le_perm_load_v16qi): Likewise.
13255 (four unnamed splitters): Modify to select the stvx instruction
13256 when memory is aligned.
13258 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
13260 * predict.c (determine_unlikely_bbs): Handle correctly BBs
13261 which appears in the queue multiple times.
13263 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13264 Alan Hayward <alan.hayward@arm.com>
13265 David Sherwood <david.sherwood@arm.com>
13267 * tree-vectorizer.h (vec_lower_bound): New structure.
13268 (_loop_vec_info): Add check_nonzero and lower_bounds.
13269 (LOOP_VINFO_CHECK_NONZERO): New macro.
13270 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
13271 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
13272 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
13273 fields. Make seg_len the distance travelled, not including the
13275 (dr_direction_indicator): Declare.
13276 (dr_zero_step_indicator): Likewise.
13277 (dr_known_forward_stride_p): Likewise.
13278 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
13280 (runtime_alias_check_p): Allow runtime alias checks with
13282 (operator ==): Compare access_size and align.
13283 (prune_runtime_alias_test_list): Rework for new distinction between
13284 the access_size and seg_len.
13285 (create_intersect_range_checks_index): Likewise. Cope with polynomial
13287 (get_segment_min_max): New function.
13288 (create_intersect_range_checks): Use it.
13289 (dr_step_indicator): New function.
13290 (dr_direction_indicator): Likewise.
13291 (dr_zero_step_indicator): Likewise.
13292 (dr_known_forward_stride_p): Likewise.
13293 * tree-loop-distribution.c (data_ref_segment_size): Return
13294 DR_STEP * (niters - 1).
13295 (compute_alias_check_pairs): Update call to the dr_with_seg_len
13297 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
13298 (vect_preserves_scalar_order_p): New function, split out from...
13299 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
13300 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
13301 (vect_vfa_access_size): New function.
13302 (vect_vfa_align): Likewise.
13303 (vect_compile_time_alias): Take access_size_a and access_b arguments.
13304 (dump_lower_bound): New function.
13305 (vect_check_lower_bound): Likewise.
13306 (vect_small_gap_p): Likewise.
13307 (vectorizable_with_step_bound_p): Likewise.
13308 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
13309 depencies if the vectorization factor is 1. Convert the checks
13310 for nonzero steps into checks on the bounds of DR_STEP. Try using
13311 a bunds check for variable steps if the minimum required step is
13312 relatively small. Update calls to the dr_with_seg_len
13313 constructor and to vect_compile_time_alias.
13314 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
13316 (vect_loop_versioning): Call it.
13317 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
13319 (vect_estimate_min_profitable_iters): Account for any bounds checks.
13321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13322 Alan Hayward <alan.hayward@arm.com>
13323 David Sherwood <david.sherwood@arm.com>
13325 * doc/sourcebuild.texi (vect_scatter_store): Document.
13326 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
13328 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
13330 * genopinit.c (main): Add supports_vec_scatter_store and
13331 supports_vec_scatter_store_cached to target_optabs.
13332 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
13333 IFN_MASK_SCATTER_STORE.
13334 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
13336 * internal-fn.h (internal_store_fn_p): Declare.
13337 (internal_fn_stored_value_index): Likewise.
13338 * internal-fn.c (scatter_store_direct): New macro.
13339 (expand_scatter_store_optab_fn): New function.
13340 (direct_scatter_store_optab_supported_p): New macro.
13341 (internal_store_fn_p): New function.
13342 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
13343 IFN_MASK_SCATTER_STORE.
13344 (internal_fn_mask_index): Likewise.
13345 (internal_fn_stored_value_index): New function.
13346 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
13347 for scatter stores.
13348 * optabs-query.h (supports_vec_scatter_store_p): Declare.
13349 * optabs-query.c (supports_vec_scatter_store_p): New function.
13350 * tree-vectorizer.h (vect_get_store_rhs): Declare.
13351 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
13352 true for scatter stores.
13353 (vect_gather_scatter_fn_p): Handle scatter stores too.
13354 (vect_check_gather_scatter): Consider using scatter stores if
13355 supports_vec_scatter_store_p.
13356 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
13357 scatter stores too.
13358 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13359 internal_fn_stored_value_index.
13360 (check_load_store_masking): Handle scatter stores too.
13361 (vect_get_store_rhs): Make public.
13362 (vectorizable_call): Use internal_store_fn_p.
13363 (vectorizable_store): Handle scatter store internal functions.
13364 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
13365 when deciding whether the end of the group has been reached.
13366 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
13367 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
13368 (mask_scatter_store<mode>): New insns.
13370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13371 Alan Hayward <alan.hayward@arm.com>
13372 David Sherwood <david.sherwood@arm.com>
13374 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
13375 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
13376 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
13378 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
13379 Use vect_truncate_gather_scatter_offset if we can't treat the
13380 operation as a normal gather load or scatter store.
13381 (get_group_load_store_type): Take the gather_scatter_info
13382 as argument. Try using a gather load or scatter store for
13383 single-element groups.
13384 (get_load_store_type): Update calls to get_group_load_store_type
13385 and vect_use_strided_gather_scatters_p.
13387 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13388 Alan Hayward <alan.hayward@arm.com>
13389 David Sherwood <david.sherwood@arm.com>
13391 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
13392 optional tree argument.
13393 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
13395 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
13396 but continue to use the current value as a fallback.
13397 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
13398 to compare the updates.
13399 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
13400 (get_load_store_type): Use it when handling a strided access.
13401 (vect_get_strided_load_store_ops): New function.
13402 (vect_get_data_ptr_increment): Likewise.
13403 (vectorizable_load): Handle strided gather loads. Always pass
13404 a step to vect_create_data_ref_ptr and bump_vector_ptr.
13406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13407 Alan Hayward <alan.hayward@arm.com>
13408 David Sherwood <david.sherwood@arm.com>
13410 * doc/md.texi (gather_load@var{m}): Document.
13411 (mask_gather_load@var{m}): Likewise.
13412 * genopinit.c (main): Add supports_vec_gather_load and
13413 supports_vec_gather_load_cached to target_optabs.
13414 * optabs-tree.c (init_tree_optimization_optabs): Use
13415 ggc_cleared_alloc to allocate target_optabs.
13416 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
13417 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
13419 * internal-fn.h (internal_load_fn_p): Declare.
13420 (internal_gather_scatter_fn_p): Likewise.
13421 (internal_fn_mask_index): Likewise.
13422 (internal_gather_scatter_fn_supported_p): Likewise.
13423 * internal-fn.c (gather_load_direct): New macro.
13424 (expand_gather_load_optab_fn): New function.
13425 (direct_gather_load_optab_supported_p): New macro.
13426 (direct_internal_fn_optab): New function.
13427 (internal_load_fn_p): Likewise.
13428 (internal_gather_scatter_fn_p): Likewise.
13429 (internal_fn_mask_index): Likewise.
13430 (internal_gather_scatter_fn_supported_p): Likewise.
13431 * optabs-query.c (supports_at_least_one_mode_p): New function.
13432 (supports_vec_gather_load_p): Likewise.
13433 * optabs-query.h (supports_vec_gather_load_p): Declare.
13434 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
13435 and memory_type field.
13436 (NUM_PATTERNS): Bump to 15.
13437 * tree-vect-data-refs.c: Include internal-fn.h.
13438 (vect_gather_scatter_fn_p): New function.
13439 (vect_describe_gather_scatter_call): Likewise.
13440 (vect_check_gather_scatter): Try using internal functions for
13441 gather loads. Recognize existing calls to a gather load function.
13442 (vect_analyze_data_refs): Consider using gather loads if
13443 supports_vec_gather_load_p.
13444 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
13445 (vect_get_gather_scatter_offset_type): Likewise.
13446 (vect_convert_mask_for_vectype): Likewise.
13447 (vect_add_conversion_to_patterm): Likewise.
13448 (vect_try_gather_scatter_pattern): Likewise.
13449 (vect_recog_gather_scatter_pattern): New pattern recognizer.
13450 (vect_vect_recog_func_ptrs): Add it.
13451 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
13452 internal_fn_mask_index and internal_gather_scatter_fn_p.
13453 (check_load_store_masking): Take the gather_scatter_info as an
13454 argument and handle gather loads.
13455 (vect_get_gather_scatter_ops): New function.
13456 (vectorizable_call): Check internal_load_fn_p.
13457 (vectorizable_load): Likewise. Handle gather load internal
13459 (vectorizable_store): Update call to check_load_store_masking.
13460 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
13461 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
13462 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
13463 (aarch64_gather_scale_operand_d): New predicates.
13464 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
13465 (mask_gather_load<mode>): New insns.
13467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13468 Alan Hayward <alan.hayward@arm.com>
13469 David Sherwood <david.sherwood@arm.com>
13471 * optabs.def (fold_left_plus_optab): New optab.
13472 * doc/md.texi (fold_left_plus_@var{m}): Document.
13473 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
13474 * internal-fn.c (fold_left_direct): Define.
13475 (expand_fold_left_optab_fn): Likewise.
13476 (direct_fold_left_optab_supported_p): Likewise.
13477 * fold-const-call.c (fold_const_fold_left): New function.
13478 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
13479 * tree-parloops.c (valid_reduction_p): New function.
13480 (gather_scalar_reductions): Use it.
13481 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
13482 (vect_finish_replace_stmt): Declare.
13483 * tree-vect-loop.c (fold_left_reduction_fn): New function.
13484 (needs_fold_left_reduction_p): New function, split out from...
13485 (vect_is_simple_reduction): ...here. Accept reductions that
13486 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
13487 (vect_force_simple_reduction): Also store the reduction type in
13488 the assignment's STMT_VINFO_REDUC_TYPE.
13489 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
13490 (merge_with_identity): New function.
13491 (vect_expand_fold_left): Likewise.
13492 (vectorize_fold_left_reduction): Likewise.
13493 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
13494 scalar phi in place for it. Check for target support and reject
13495 cases that would reassociate the operation. Defer the transform
13496 phase to vectorize_fold_left_reduction.
13497 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
13498 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
13499 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
13501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13503 * tree-if-conv.c (predicate_mem_writes): Remove redundant
13504 call to ifc_temp_var.
13506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13507 Alan Hayward <alan.hayward@arm.com>
13508 David Sherwood <david.sherwood@arm.com>
13510 * target.def (legitimize_address_displacement): Take the original
13511 offset as a poly_int.
13512 * targhooks.h (default_legitimize_address_displacement): Update
13514 * targhooks.c (default_legitimize_address_displacement): Likewise.
13515 * doc/tm.texi: Regenerate.
13516 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
13517 as an argument, moving assert of ad->disp == ad->disp_term to...
13518 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
13519 Try calling targetm.legitimize_address_displacement before expanding
13520 the address rather than afterwards, and adjust for the new interface.
13521 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
13522 Match the new hook interface. Handle SVE addresses.
13523 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13524 new hook interface.
13526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13528 * Makefile.in (OBJS): Add early-remat.o.
13529 * target.def (select_early_remat_modes): New hook.
13530 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13531 * doc/tm.texi: Regenerate.
13532 * targhooks.h (default_select_early_remat_modes): Declare.
13533 * targhooks.c (default_select_early_remat_modes): New function.
13534 * timevar.def (TV_EARLY_REMAT): New timevar.
13535 * passes.def (pass_early_remat): New pass.
13536 * tree-pass.h (make_pass_early_remat): Declare.
13537 * early-remat.c: New file.
13538 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13540 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13543 Alan Hayward <alan.hayward@arm.com>
13544 David Sherwood <david.sherwood@arm.com>
13546 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13547 vfm1 with a bound_epilog parameter.
13548 (vect_do_peeling): Update calls accordingly, and move the prologue
13549 call earlier in the function. Treat the base bound_epilog as 0 for
13550 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13551 this base when peeling for gaps.
13552 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13553 with fully-masked loops.
13554 (vect_estimate_min_profitable_iters): Handle the single peeled
13555 iteration in that case.
13557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13558 Alan Hayward <alan.hayward@arm.com>
13559 David Sherwood <david.sherwood@arm.com>
13561 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13562 single-element interleaving even if the size is not a power of 2.
13563 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13564 accesses for single-element interleaving if the group size is
13567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13568 Alan Hayward <alan.hayward@arm.com>
13569 David Sherwood <david.sherwood@arm.com>
13571 * doc/md.texi (fold_extract_last_@var{m}): Document.
13572 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13573 * optabs.def (fold_extract_last_optab): New optab.
13574 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13575 * internal-fn.c (fold_extract_direct): New macro.
13576 (expand_fold_extract_optab_fn): Likewise.
13577 (direct_fold_extract_optab_supported_p): Likewise.
13578 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13579 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13580 EXTRACT_LAST_REDUCTION.
13581 (get_initial_def_for_reduction): Do not create an initial vector
13582 for EXTRACT_LAST_REDUCTION reductions.
13583 (vectorizable_reduction): Leave the scalar phi in place for
13584 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13585 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13586 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13587 transform phase to vectorizable_condition.
13588 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13590 (vect_finish_stmt_generation): ...here.
13591 (vect_finish_replace_stmt): New function.
13592 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13593 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13595 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13598 Alan Hayward <alan.hayward@arm.com>
13599 David Sherwood <david.sherwood@arm.com>
13601 * doc/md.texi (extract_last_@var{m}): Document.
13602 * optabs.def (extract_last_optab): New optab.
13603 * internal-fn.def (EXTRACT_LAST): New internal function.
13604 * internal-fn.c (cond_unary_direct): New macro.
13605 (expand_cond_unary_optab_fn): Likewise.
13606 (direct_cond_unary_optab_supported_p): Likewise.
13607 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13608 loops using EXTRACT_LAST.
13609 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13610 (extract_last_<mode>): ...this optab.
13611 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
13618 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13619 * doc/tm.texi: Regenerate.
13620 * targhooks.h (default_empty_mask_is_expensive): Declare.
13621 * targhooks.c (default_empty_mask_is_expensive): New function.
13622 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13623 if the target says that empty masks are expensive.
13624 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13626 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13629 Alan Hayward <alan.hayward@arm.com>
13630 David Sherwood <david.sherwood@arm.com>
13632 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13633 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13634 (vect_use_loop_mask_for_alignment_p): New function.
13635 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13636 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13637 niters_skip argument. Make sure that the first niters_skip elements
13638 of the first iteration are inactive.
13639 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13640 Update call to vect_set_loop_masks_directly.
13641 (get_misalign_in_elems): New function, split out from...
13642 (vect_gen_prolog_loop_niters): ...here.
13643 (vect_update_init_of_dr): Take a code argument that specifies whether
13644 the adjustment should be added or subtracted.
13645 (vect_update_init_of_drs): Likewise.
13646 (vect_prepare_for_masked_peels): New function.
13647 (vect_do_peeling): Skip prologue peeling if we're using a mask
13648 instead. Update call to vect_update_inits_of_drs.
13649 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13651 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13652 alignment. Do not include the number of peeled iterations in
13653 the minimum threshold in that case.
13654 (vectorizable_induction): Adjust the start value down by
13655 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13656 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13657 Take the number of skipped iterations into account when calculating
13659 * tree-vect-stmts.c (vect_gen_while_not): New function.
13661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13662 Alan Hayward <alan.hayward@arm.com>
13663 David Sherwood <david.sherwood@arm.com>
13665 * doc/sourcebuild.texi (vect_fully_masked): Document.
13666 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13667 default value to 0.
13668 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13670 (vect_analyze_loop_2): ...here. Don't check the vectorization
13671 factor against the number of loop iterations if the loop is
13674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13675 Alan Hayward <alan.hayward@arm.com>
13676 David Sherwood <david.sherwood@arm.com>
13678 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13679 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13680 (dump_groups): Update accordingly.
13681 (iv_use::mem_type): New member variable.
13682 (address_p): New function.
13683 (record_use): Add a mem_type argument and initialize the new
13685 (record_group_use): Add a mem_type argument. Use address_p.
13686 Remove obsolete null checks of base_object. Update call to record_use.
13687 (find_interesting_uses_op): Update call to record_group_use.
13688 (find_interesting_uses_cond): Likewise.
13689 (find_interesting_uses_address): Likewise.
13690 (get_mem_type_for_internal_fn): New function.
13691 (find_address_like_use): Likewise.
13692 (find_interesting_uses_stmt): Try find_address_like_use before
13693 calling find_interesting_uses_op.
13694 (addr_offset_valid_p): Use the iv mem_type field as the type
13695 of the addressed memory.
13696 (add_autoinc_candidates): Likewise.
13697 (get_address_cost): Likewise.
13698 (split_small_address_groups_p): Use address_p.
13699 (split_address_groups): Likewise.
13700 (add_iv_candidate_for_use): Likewise.
13701 (autoinc_possible_for_pair): Likewise.
13702 (rewrite_groups): Likewise.
13703 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13704 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13705 (get_alias_ptr_type_for_ptr_address): New function.
13706 (rewrite_use_address): Rewrite address uses in calls that were
13707 identified by find_address_like_use.
13709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13710 Alan Hayward <alan.hayward@arm.com>
13711 David Sherwood <david.sherwood@arm.com>
13713 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13715 * gimple-expr.h (is_gimple_addressable: Likewise.
13716 * gimple-expr.c (is_gimple_address): Likewise.
13717 * internal-fn.c (expand_call_mem_ref): New function.
13718 (expand_mask_load_optab_fn): Use it.
13719 (expand_mask_store_optab_fn): Likewise.
13721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13722 Alan Hayward <alan.hayward@arm.com>
13723 David Sherwood <david.sherwood@arm.com>
13725 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13726 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13727 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13728 (cond_umax@var{mode}): Document.
13729 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13730 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13731 (cond_umin_optab, cond_umax_optab): New optabs.
13732 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13733 (COND_IOR, COND_XOR): New internal functions.
13734 * internal-fn.h (get_conditional_internal_fn): Declare.
13735 * internal-fn.c (cond_binary_direct): New macro.
13736 (expand_cond_binary_optab_fn): Likewise.
13737 (direct_cond_binary_optab_supported_p): Likewise.
13738 (get_conditional_internal_fn): New function.
13739 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13740 Cope with reduction statements that are vectorized as calls rather
13742 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13743 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13744 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13745 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13746 (UNSPEC_COND_EOR): New unspecs.
13747 (optab): Add mappings for them.
13748 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13749 (sve_int_op, sve_fp_op): New int attributes.
13751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13752 Alan Hayward <alan.hayward@arm.com>
13753 David Sherwood <david.sherwood@arm.com>
13755 * optabs.def (while_ult_optab): New optab.
13756 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13757 * internal-fn.def (WHILE_ULT): New internal function.
13758 * internal-fn.h (direct_internal_fn_supported_p): New override
13759 that takes two types as argument.
13760 * internal-fn.c (while_direct): New macro.
13761 (expand_while_optab_fn): New function.
13762 (convert_optab_supported_p): Likewise.
13763 (direct_while_optab_supported_p): New macro.
13764 * wide-int.h (wi::udiv_ceil): New function.
13765 * tree-vectorizer.h (rgroup_masks): New structure.
13766 (vec_loop_masks): New typedef.
13767 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13768 and fully_masked_p.
13769 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13770 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13771 (vect_max_vf): New function.
13772 (slpeel_make_loop_iterate_ntimes): Delete.
13773 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13774 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13775 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13776 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13777 internal-fn.h, stor-layout.h and optabs-query.h.
13778 (vect_set_loop_mask): New function.
13779 (add_preheader_seq): Likewise.
13780 (add_header_seq): Likewise.
13781 (interleave_supported_p): Likewise.
13782 (vect_maybe_permute_loop_masks): Likewise.
13783 (vect_set_loop_masks_directly): Likewise.
13784 (vect_set_loop_condition_masked): Likewise.
13785 (vect_set_loop_condition_unmasked): New function, split out from
13786 slpeel_make_loop_iterate_ntimes.
13787 (slpeel_make_loop_iterate_ntimes): Rename to..
13788 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13789 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13790 (vect_do_peeling): Update call accordingly.
13791 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13793 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13794 mask_compare_type, can_fully_mask_p and fully_masked_p.
13795 (release_vec_loop_masks): New function.
13796 (_loop_vec_info): Use it to free the loop masks.
13797 (can_produce_all_loop_masks_p): New function.
13798 (vect_get_max_nscalars_per_iter): Likewise.
13799 (vect_verify_full_masking): Likewise.
13800 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13801 retries, and free the mask rgroups before retrying. Check loop-wide
13802 reasons for disallowing fully-masked loops. Make the final decision
13803 about whether use a fully-masked loop or not.
13804 (vect_estimate_min_profitable_iters): Do not assume that peeling
13805 for the number of iterations will be needed for fully-masked loops.
13806 (vectorizable_reduction): Disable fully-masked loops.
13807 (vectorizable_live_operation): Likewise.
13808 (vect_halve_mask_nunits): New function.
13809 (vect_double_mask_nunits): Likewise.
13810 (vect_record_loop_mask): Likewise.
13811 (vect_get_loop_mask): Likewise.
13812 (vect_transform_loop): Handle the case in which the final loop
13813 iteration might handle a partial vector. Call vect_set_loop_condition
13814 instead of slpeel_make_loop_iterate_ntimes.
13815 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13816 (check_load_store_masking): New function.
13817 (prepare_load_store_mask): Likewise.
13818 (vectorizable_store): Handle fully-masked loops.
13819 (vectorizable_load): Likewise.
13820 (supportable_widening_operation): Use vect_halve_mask_nunits for
13822 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13823 (vect_gen_while): New function.
13824 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13825 (aarch64_uqdec<mode>): New insn.
13827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13828 Alan Hayward <alan.hayward@arm.com>
13829 David Sherwood <david.sherwood@arm.com>
13831 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13832 (reduc_xor_scal_optab): New optabs.
13833 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13834 (reduc_xor_scal_@var{m}): Document.
13835 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13836 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13837 internal functions.
13838 * fold-const-call.c (fold_const_call): Handle them.
13839 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13840 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13841 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13842 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13843 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13844 (UNSPEC_XORV): New unspecs.
13845 (optab): Add entries for them.
13846 (BITWISEV): New int iterator.
13847 (bit_reduc_op): New int attributes.
13849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13850 Alan Hayward <alan.hayward@arm.com>
13851 David Sherwood <david.sherwood@arm.com>
13853 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13854 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13855 * optabs.def (vec_shl_insert_optab): New optab.
13856 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13857 (duplicate_and_interleave): Likewise.
13858 * tree-vect-loop.c: Include internal-fn.h.
13859 (neutral_op_for_slp_reduction): New function, split out from
13860 get_initial_defs_for_reduction.
13861 (get_initial_def_for_reduction): Handle option 2 for variable-length
13862 vectors by loading the neutral value into a vector and then shifting
13863 the initial value into element 0.
13864 (get_initial_defs_for_reduction): Replace the code argument with
13865 the neutral value calculated by neutral_op_for_slp_reduction.
13866 Use gimple_build_vector for constant-length vectors.
13867 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13868 but the first group_size elements have a neutral value.
13869 Use duplicate_and_interleave otherwise.
13870 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13871 Update call to get_initial_defs_for_reduction. Handle SLP
13872 reductions for variable-length vectors by creating one vector
13873 result for each scalar result, with the elements associated
13874 with other scalar results stubbed out with the neutral value.
13875 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13876 Require IFN_VEC_SHL_INSERT for double reductions on
13877 variable-length vectors, or SLP reductions that have
13878 a neutral value. Require can_duplicate_and_interleave_p
13879 support for variable-length unchained SLP reductions if there
13880 is no neutral value, such as for MIN/MAX reductions. Also require
13881 the number of vector elements to be a multiple of the number of
13882 SLP statements when doing variable-length unchained SLP reductions.
13883 Update call to vect_create_epilog_for_reduction.
13884 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13885 and remove initial values.
13886 (duplicate_and_interleave): Make public.
13887 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13888 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13891 Alan Hayward <alan.hayward@arm.com>
13892 David Sherwood <david.sherwood@arm.com>
13894 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13895 (can_duplicate_and_interleave_p): New function.
13896 (vect_get_and_check_slp_defs): Take the vector of statements
13897 rather than just the current one. Remove excess parentheses.
13898 Restriction rejectinon of vect_constant_def and vect_external_def
13899 for variable-length vectors to boolean types, or types for which
13900 can_duplicate_and_interleave_p is false.
13901 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13902 (duplicate_and_interleave): New function.
13903 (vect_get_constant_vectors): Use gimple_build_vector for
13904 constant-length vectors and suitable variable-length constant
13905 vectors. Use duplicate_and_interleave for other variable-length
13906 vectors. Don't defer the update when inserting new statements.
13908 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13909 Alan Hayward <alan.hayward@arm.com>
13910 David Sherwood <david.sherwood@arm.com>
13912 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13913 min_profitable_iters doesn't go negative.
13915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13916 Alan Hayward <alan.hayward@arm.com>
13917 David Sherwood <david.sherwood@arm.com>
13919 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13920 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13921 * optabs.def (vec_mask_load_lanes_optab): New optab.
13922 (vec_mask_store_lanes_optab): Likewise.
13923 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13924 (MASK_STORE_LANES): Likewise.
13925 * internal-fn.c (mask_load_lanes_direct): New macro.
13926 (mask_store_lanes_direct): Likewise.
13927 (expand_mask_load_optab_fn): Handle masked operations.
13928 (expand_mask_load_lanes_optab_fn): New macro.
13929 (expand_mask_store_optab_fn): Handle masked operations.
13930 (expand_mask_store_lanes_optab_fn): New macro.
13931 (direct_mask_load_lanes_optab_supported_p): Likewise.
13932 (direct_mask_store_lanes_optab_supported_p): Likewise.
13933 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13935 (vect_load_lanes_supported): Likewise.
13936 * tree-vect-data-refs.c (strip_conversion): New function.
13937 (can_group_stmts_p): Likewise.
13938 (vect_analyze_data_ref_accesses): Use it instead of checking
13939 for a pair of assignments.
13940 (vect_store_lanes_supported): Take a masked_p parameter.
13941 (vect_load_lanes_supported): Likewise.
13942 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13943 vect_store_lanes_supported and vect_load_lanes_supported.
13944 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13945 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13946 parameter. Don't allow gaps for masked accesses.
13947 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13948 and vect_load_lanes_supported.
13949 (get_load_store_type): Take a masked_p parameter and update
13950 call to get_group_load_store_type.
13951 (vectorizable_store): Update call to get_load_store_type.
13952 Handle IFN_MASK_STORE_LANES.
13953 (vectorizable_load): Update call to get_load_store_type.
13954 Handle IFN_MASK_LOAD_LANES.
13956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13957 Alan Hayward <alan.hayward@arm.com>
13958 David Sherwood <david.sherwood@arm.com>
13960 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13962 * config/aarch64/aarch64-protos.h
13963 (aarch64_sve_struct_memory_operand_p): Declare.
13964 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13965 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13966 (VPRED, vpred): Handle SVE structure modes.
13967 * config/aarch64/constraints.md (Utx): New constraint.
13968 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13969 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13970 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13971 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13972 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13973 structure modes. Split into pieces after RA.
13974 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13975 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13977 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13978 SVE structure modes.
13979 (aarch64_classify_address): Likewise.
13980 (sizetochar): Move earlier in file.
13981 (aarch64_print_operand): Handle SVE register lists.
13982 (aarch64_array_mode): New function.
13983 (aarch64_sve_struct_memory_operand_p): Likewise.
13984 (TARGET_ARRAY_MODE): Redefine.
13986 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13987 Alan Hayward <alan.hayward@arm.com>
13988 David Sherwood <david.sherwood@arm.com>
13990 * target.def (array_mode): New target hook.
13991 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13992 * doc/tm.texi: Regenerate.
13993 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13994 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13995 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13996 targetm.array_mode.
13997 * stor-layout.c (mode_for_array): Likewise. Support polynomial
14000 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14001 Alan Hayward <alan.hayward@arm.com>
14002 David Sherwood <david.sherwood@arm.com>
14004 * fold-const.c (fold_binary_loc): Check the argument types
14005 rather than the result type when testing for a vector operation.
14007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14009 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
14010 * doc/tm.texi: Regenerate.
14012 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14013 Alan Hayward <alan.hayward@arm.com>
14014 David Sherwood <david.sherwood@arm.com>
14016 * doc/invoke.texi (-msve-vector-bits=): Document new option.
14017 (sve): Document new AArch64 extension.
14018 * doc/md.texi (w): Extend the description of the AArch64
14019 constraint to include SVE vectors.
14020 (Upl, Upa): Document new AArch64 predicate constraints.
14021 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
14023 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
14024 (msve-vector-bits=): New option.
14025 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
14026 SVE when these are disabled.
14027 (sve): New extension.
14028 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
14029 modes. Adjust their number of units based on aarch64_sve_vg.
14030 (MAX_BITSIZE_MODE_ANY_MODE): Define.
14031 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
14032 aarch64_addr_query_type.
14033 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
14034 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
14035 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
14036 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
14037 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
14038 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
14039 (aarch64_simd_imm_zero_p): Delete.
14040 (aarch64_check_zero_based_sve_index_immediate): Declare.
14041 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14042 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14043 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14044 (aarch64_sve_float_mul_immediate_p): Likewise.
14045 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14046 rather than an rtx.
14047 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
14048 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
14049 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
14050 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
14051 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
14052 (aarch64_regmode_natural_size): Likewise.
14053 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
14054 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
14056 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
14057 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
14058 for VG and the SVE predicate registers.
14059 (V_ALIASES): Add a "z"-prefixed alias.
14060 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
14061 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
14062 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
14063 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
14064 (REG_CLASS_NAMES): Add entries for them.
14065 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
14066 and the predicate registers.
14067 (aarch64_sve_vg): Declare.
14068 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
14069 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
14070 (REGMODE_NATURAL_SIZE): Define.
14071 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
14073 * config/aarch64/aarch64.c: Include cfgrtl.h.
14074 (simd_immediate_info): Add a constructor for series vectors,
14075 and an associated step field.
14076 (aarch64_sve_vg): New variable.
14077 (aarch64_dbx_register_number): Handle VG and the predicate registers.
14078 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
14079 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
14080 (VEC_ANY_DATA, VEC_STRUCT): New constants.
14081 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
14082 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
14083 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
14084 (aarch64_get_mask_mode): New functions.
14085 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
14086 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
14087 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
14088 predicate modes and predicate registers. Explicitly restrict
14089 GPRs to modes of 16 bytes or smaller. Only allow FP registers
14090 to store a vector mode if it is recognized by
14091 aarch64_classify_vector_mode.
14092 (aarch64_regmode_natural_size): New function.
14093 (aarch64_hard_regno_caller_save_mode): Return the original mode
14095 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
14096 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
14097 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
14098 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
14100 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
14101 does not overlap dest if the function is frame-related. Handle
14103 (aarch64_split_add_offset): New function.
14104 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
14105 them aarch64_add_offset.
14106 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
14107 and update call to aarch64_sub_sp.
14108 (aarch64_add_cfa_expression): New function.
14109 (aarch64_expand_prologue): Pass extra temporary registers to the
14110 functions above. Handle the case in which we need to emit new
14111 DW_CFA_expressions for registers that were originally saved
14112 relative to the stack pointer, but now have to be expressed
14113 relative to the frame pointer.
14114 (aarch64_output_mi_thunk): Pass extra temporary registers to the
14116 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
14117 IP0 and IP1 values for SVE frames.
14118 (aarch64_expand_vec_series): New function.
14119 (aarch64_expand_sve_widened_duplicate): Likewise.
14120 (aarch64_expand_sve_const_vector): Likewise.
14121 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
14122 Handle SVE constants. Use emit_move_insn to move a force_const_mem
14123 into the register, rather than emitting a SET directly.
14124 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
14125 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
14126 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
14127 (offset_9bit_signed_scaled_p): New functions.
14128 (aarch64_replicate_bitmask_imm): New function.
14129 (aarch64_bitmask_imm): Use it.
14130 (aarch64_cannot_force_const_mem): Reject expressions involving
14131 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
14132 (aarch64_classify_index): Handle SVE indices, by requiring
14133 a plain register index with a scale that matches the element size.
14134 (aarch64_classify_address): Handle SVE addresses. Assert that
14135 the mode of the address is VOIDmode or an integer mode.
14136 Update call to aarch64_classify_symbol.
14137 (aarch64_classify_symbolic_expression): Update call to
14138 aarch64_classify_symbol.
14139 (aarch64_const_vec_all_in_range_p): New function.
14140 (aarch64_print_vector_float_operand): Likewise.
14141 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
14142 "vN" for FP registers with SVE modes. Handle (const ...) vectors
14143 and the FP immediates 1.0 and 0.5.
14144 (aarch64_print_address_internal): Handle SVE addresses.
14145 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
14146 (aarch64_regno_regclass): Handle predicate registers.
14147 (aarch64_secondary_reload): Handle big-endian reloads of SVE
14149 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
14150 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
14151 (aarch64_convert_sve_vector_bits): New function.
14152 (aarch64_override_options): Use it to handle -msve-vector-bits=.
14153 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
14154 rather than an rtx.
14155 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
14156 Handle SVE vector and predicate modes. Accept VL-based constants
14157 that need only one temporary register, and VL offsets that require
14158 no temporary registers.
14159 (aarch64_conditional_register_usage): Mark the predicate registers
14160 as fixed if SVE isn't available.
14161 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
14162 Return true for SVE vector and predicate modes.
14163 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
14164 rather than an unsigned int. Handle SVE modes.
14165 (aarch64_preferred_simd_mode): Update call accordingly. Handle
14167 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
14169 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
14170 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
14171 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
14172 (aarch64_sve_float_mul_immediate_p): New functions.
14173 (aarch64_sve_valid_immediate): New function.
14174 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
14175 Explicitly reject structure modes. Check for INDEX constants.
14176 Handle PTRUE and PFALSE constants.
14177 (aarch64_check_zero_based_sve_index_immediate): New function.
14178 (aarch64_simd_imm_zero_p): Delete.
14179 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
14180 vector modes. Accept constants in the range of CNT[BHWD].
14181 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
14182 ask for an Advanced SIMD mode.
14183 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
14184 (aarch64_simd_vector_alignment): Handle SVE predicates.
14185 (aarch64_vectorize_preferred_vector_alignment): New function.
14186 (aarch64_simd_vector_alignment_reachable): Use it instead of
14188 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
14189 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
14191 (MAX_VECT_LEN): Delete.
14192 (expand_vec_perm_d): Add a vec_flags field.
14193 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
14194 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14195 (aarch64_evpc_ext): Don't apply a big-endian lane correction
14197 (aarch64_evpc_rev): Rename to...
14198 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
14199 (aarch64_evpc_rev_global): New function.
14200 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
14201 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
14203 (aarch64_evpc_sve_tbl): New function.
14204 (aarch64_expand_vec_perm_const_1): Update after rename of
14205 aarch64_evpc_rev. Handle SVE permutes too, trying
14206 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
14207 than aarch64_evpc_tbl.
14208 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
14209 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
14210 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
14211 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
14212 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
14213 (aarch64_expand_sve_vcond): New functions.
14214 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
14215 of aarch64_vector_mode_p.
14216 (aarch64_dwarf_poly_indeterminate_value): New function.
14217 (aarch64_compute_pressure_classes): Likewise.
14218 (aarch64_can_change_mode_class): Likewise.
14219 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
14220 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
14221 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
14222 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
14223 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
14224 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
14225 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
14226 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
14228 (Dn, Dl, Dr): Accept const as well as const_vector.
14229 (Dz): Likewise. Compare against CONST0_RTX.
14230 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
14231 of "vector" where appropriate.
14232 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
14233 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
14234 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
14235 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
14236 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
14237 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
14238 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
14239 (v_int_equiv): Extend to SVE modes.
14240 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
14242 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
14243 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
14244 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
14245 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
14246 (SVE_COND_FP_CMP): New int iterators.
14247 (perm_hilo): Handle the new unpack unspecs.
14248 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
14250 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
14251 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
14252 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
14253 (aarch64_equality_operator, aarch64_constant_vector_operand)
14254 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
14255 (aarch64_sve_nonimmediate_operand): Likewise.
14256 (aarch64_sve_general_operand): Likewise.
14257 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
14258 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
14259 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
14260 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
14261 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
14262 (aarch64_sve_float_arith_immediate): Likewise.
14263 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
14264 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
14265 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
14266 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
14267 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
14268 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
14269 (aarch64_sve_float_arith_operand): Likewise.
14270 (aarch64_sve_float_arith_with_sub_operand): Likewise.
14271 (aarch64_sve_float_mul_operand): Likewise.
14272 (aarch64_sve_vec_perm_operand): Likewise.
14273 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
14274 (aarch64_mov_operand): Accept const_poly_int and const_vector.
14275 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
14276 as well as const_vector.
14277 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
14278 in file. Use CONST0_RTX and CONSTM1_RTX.
14279 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
14280 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
14281 Use aarch64_simd_imm_zero.
14282 * config/aarch64/aarch64-sve.md: New file.
14283 * config/aarch64/aarch64.md: Include it.
14284 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
14285 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
14286 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
14287 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
14288 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
14289 (sve): New attribute.
14290 (enabled): Disable instructions with the sve attribute unless
14292 (movqi, movhi): Pass CONST_POLY_INT operaneds through
14293 aarch64_expand_mov_immediate.
14294 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
14295 CNT[BHSD] immediates.
14296 (movti): Split CONST_POLY_INT moves into two halves.
14297 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
14298 Split additions that need a temporary here if the destination
14299 is the stack pointer.
14300 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
14301 (*add<mode>3_poly_1): New instruction.
14302 (set_clobber_cc): New expander.
14304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14306 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
14307 parameter and use it instead of GET_MODE_SIZE (innermode). Use
14308 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
14309 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
14310 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
14311 Change innermode from fixed_mode_size to machine_mode.
14312 (simplify_subreg): Update call accordingly. Handle a constant-sized
14313 subreg of a variable-length CONST_VECTOR.
14315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
14316 Alan Hayward <alan.hayward@arm.com>
14317 David Sherwood <david.sherwood@arm.com>
14319 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
14320 (add_offset_to_base): New function, split out from...
14321 (create_mem_ref): ...here. When handling a scale other than 1,
14322 check first whether the address is valid without the offset.
14323 Add it into the base if so, leaving the index and scale as-is.
14325 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14328 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
14329 fold_for_warn before checking if arg2 is INTEGER_CST.
14331 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
14333 * config/rs6000/predicates.md (load_multiple_operation): Delete.
14334 (store_multiple_operation): Delete.
14335 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
14336 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
14337 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
14338 guarded by TARGET_STRING.
14339 (rs6000_output_load_multiple): Delete.
14340 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
14341 OPTION_MASK_STRING / TARGET_STRING handling.
14342 (print_operand) <'N', 'O'>: Add comment that these are unused now.
14343 (const rs6000_opt_masks) <"string">: Change mask to 0.
14344 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
14345 (MASK_STRING): Delete.
14346 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
14348 (load_multiple): Delete.
14355 (store_multiple): Delete.
14362 (movmemsi_8reg): Delete.
14363 (corresponding unnamed define_insn): Delete.
14364 (movmemsi_6reg): Delete.
14365 (corresponding unnamed define_insn): Delete.
14366 (movmemsi_4reg): Delete.
14367 (corresponding unnamed define_insn): Delete.
14368 (movmemsi_2reg): Delete.
14369 (corresponding unnamed define_insn): Delete.
14370 (movmemsi_1reg): Delete.
14371 (corresponding unnamed define_insn): Delete.
14372 * config/rs6000/rs6000.opt (mno-string): New.
14373 (mstring): Replace by deprecation warning stub.
14374 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
14376 2018-01-12 Jakub Jelinek <jakub@redhat.com>
14378 * regrename.c (regrename_do_replace): If replacing the same
14379 reg multiple times, try to reuse last created gen_raw_REG.
14382 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
14383 main to workaround a bug in GDB.
14385 2018-01-12 Tom de Vries <tom@codesourcery.com>
14388 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
14390 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
14392 PR rtl-optimization/80481
14393 * ira-color.c (get_cap_member): New function.
14394 (allocnos_conflict_by_live_ranges_p): Use it.
14395 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
14396 (setup_slot_coalesced_allocno_live_ranges): Ditto.
14398 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
14401 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
14402 (*saddl_se_1): Ditto.
14403 (*ssubsi_1): Ditto.
14404 (*ssubl_se_1): Ditto.
14406 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14408 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
14409 rather than wi::to_widest for DR_INITs.
14410 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
14411 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
14412 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
14414 (vect_analyze_group_access_1): Note that here.
14416 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14418 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
14419 polynomial type sizes.
14421 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
14423 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
14424 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
14425 (gimple_add_tmp_var): Likewise.
14427 2018-01-12 Martin Liska <mliska@suse.cz>
14429 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
14430 (gimple_alloc_sizes): Likewise.
14431 (dump_gimple_statistics): Use PRIu64 in printf format.
14432 * gimple.h: Change uint64_t to int.
14434 2018-01-12 Martin Liska <mliska@suse.cz>
14436 * tree-core.h: Use uint64_t instead of int.
14437 * tree.c (tree_node_counts): Likewise.
14438 (tree_node_sizes): Likewise.
14439 (dump_tree_statistics): Use PRIu64 in printf format.
14441 2018-01-12 Martin Liska <mliska@suse.cz>
14443 * Makefile.in: As qsort_chk is implemented in vec.c, add
14444 vec.o to linkage of gencfn-macros.
14445 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
14446 passing the info to record_node_allocation_statistics.
14447 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
14449 * ggc-common.c (struct ggc_usage): Add operator== and use
14450 it in operator< and compare function.
14451 * mem-stats.h (struct mem_usage): Likewise.
14452 * vec.c (struct vec_usage): Remove operator< and compare
14453 function. Can be simply inherited.
14455 2018-01-12 Martin Jambor <mjambor@suse.cz>
14458 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
14459 * tree-ssa-math-opts.c: Include domwalk.h.
14460 (convert_mult_to_fma_1): New function.
14461 (fma_transformation_info): New type.
14462 (fma_deferring_state): Likewise.
14463 (cancel_fma_deferring): New function.
14464 (result_of_phi): Likewise.
14465 (last_fma_candidate_feeds_initial_phi): Likewise.
14466 (convert_mult_to_fma): Added deferring logic, split actual
14467 transformation to convert_mult_to_fma_1.
14468 (math_opts_dom_walker): New type.
14469 (math_opts_dom_walker::after_dom_children): New method, body moved
14470 here from pass_optimize_widening_mul::execute, added deferring logic
14472 (pass_optimize_widening_mul::execute): Moved most of code to
14473 math_opts_dom_walker::after_dom_children.
14474 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
14475 * config/i386/i386.c (ix86_option_override_internal): Added
14476 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
14478 2018-01-12 Richard Biener <rguenther@suse.de>
14481 * dwarf2out.c (gen_variable_die): Do not reset old_die for
14482 inline instance vars.
14484 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
14487 * config/rx/rx.c (rx_is_restricted_memory_address):
14488 Handle SUBREG case.
14490 2018-01-12 Richard Biener <rguenther@suse.de>
14492 PR tree-optimization/80846
14493 * target.def (split_reduction): New target hook.
14494 * targhooks.c (default_split_reduction): New function.
14495 * targhooks.h (default_split_reduction): Declare.
14496 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
14497 target requests first reduce vectors by combining low and high
14499 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
14500 (get_vectype_for_scalar_type_and_size): Export.
14501 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
14502 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
14503 * doc/tm.texi: Regenerate.
14504 * config/i386/i386.c (ix86_split_reduction): Implement
14505 TARGET_VECTORIZE_SPLIT_REDUCTION.
14507 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14510 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
14511 in PIC mode except for TARGET_VXWORKS_RTP.
14512 * config/sparc/sparc.c: Include cfgrtl.h.
14513 (TARGET_INIT_PIC_REG): Define.
14514 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
14515 (sparc_pic_register_p): New predicate.
14516 (sparc_legitimate_address_p): Use it.
14517 (sparc_legitimize_pic_address): Likewise.
14518 (sparc_delegitimize_address): Likewise.
14519 (sparc_mode_dependent_address_p): Likewise.
14520 (gen_load_pcrel_sym): Remove 4th parameter.
14521 (load_got_register): Adjust call to above. Remove obsolete stuff.
14522 (sparc_expand_prologue): Do not call load_got_register here.
14523 (sparc_flat_expand_prologue): Likewise.
14524 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14525 (sparc_use_pseudo_pic_reg): New function.
14526 (sparc_init_pic_reg): Likewise.
14527 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14528 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14530 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14532 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14533 Add item for branch_cost.
14535 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14537 PR rtl-optimization/83565
14538 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14539 not extend the result to a larger mode for rotate operations.
14540 (num_sign_bit_copies1): Likewise.
14542 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14545 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14547 Use values-Xc.o for -pedantic.
14548 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14550 2018-01-12 Martin Liska <mliska@suse.cz>
14553 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14555 (possible_polymorphic_call_targets): Use it.
14556 (ipa_devirt): Likewise.
14558 2018-01-12 Martin Liska <mliska@suse.cz>
14560 * profile-count.h (enum profile_quality): Use 0 as invalid
14561 enum value of profile_quality.
14563 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14565 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14566 -mext-string options.
14568 2018-01-12 Richard Biener <rguenther@suse.de>
14570 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14571 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14572 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14574 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14576 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14578 * configure.ac (--with-long-double-format): Add support for the
14579 configuration option to change the default long double format on
14581 * config.gcc (powerpc*-linux*-*): Likewise.
14582 * configure: Regenerate.
14583 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14584 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14585 used without modification.
14587 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14589 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14590 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14591 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14592 MISC_BUILTIN_SPEC_BARRIER.
14593 (rs6000_init_builtins): Likewise.
14594 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14596 (speculation_barrier): New define_insn.
14597 * doc/extend.texi: Document __builtin_speculation_barrier.
14599 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14602 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14603 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14604 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14606 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14607 integral modes instead of "ss" and "sd".
14608 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14609 vectors with 32-bit and 64-bit elements.
14610 (vecdupssescalarmodesuffix): New mode attribute.
14611 (vec_dup<mode>): Use it.
14613 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14616 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14617 frame if argument is passed on stack.
14619 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14622 * ree.c (combine_reaching_defs): Optimize also
14623 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14624 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14626 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14628 PR middle-end/83189
14629 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14631 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14633 PR middle-end/83718
14634 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14635 after they are computed.
14637 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14639 PR tree-optimization/83695
14640 * gimple-loop-linterchange.cc
14641 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14642 reset cached scev information after interchange.
14643 (pass_linterchange::execute): Remove call to scev_reset_htab.
14645 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14647 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14648 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14649 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14650 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14651 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14652 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14653 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14654 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14655 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14656 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14657 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14658 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14659 (V_lane_reg): Likewise.
14660 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14662 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14663 (vfmal_lane_low<mode>_intrinsic,
14664 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14665 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14666 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14667 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14668 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14669 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14671 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14673 * config/arm/arm-cpus.in (fp16fml): New feature.
14674 (ALL_SIMD): Add fp16fml.
14675 (armv8.2-a): Add fp16fml as an option.
14676 (armv8.3-a): Likewise.
14677 (armv8.4-a): Add fp16fml as part of fp16.
14678 * config/arm/arm.h (TARGET_FP16FML): Define.
14679 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14681 * config/arm/arm-modes.def (V2HF): Define.
14682 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14683 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14684 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14685 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14686 vfmsl_low, vfmsl_high): New set of builtins.
14687 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14688 (vfml_op): New code attribute.
14689 (VFMLHALVES): New int iterator.
14690 (VFML, VFMLSEL): New mode attributes.
14691 (V_reg): Define mapping for V2HF.
14692 (V_hi, V_lo): New mode attributes.
14693 (VF_constraint): Likewise.
14694 (vfml_half, vfml_half_selector): New int attributes.
14695 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14697 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14698 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14700 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14701 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14702 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14703 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14705 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14706 Document new effective target and option set.
14708 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14710 * config/arm/arm-cpus.in (armv8_4): New feature.
14711 (ARMv8_4a): New fgroup.
14712 (armv8.4-a): New arch.
14713 * config/arm/arm-tables.opt: Regenerate.
14714 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14715 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14716 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14717 Add matching rules for -march=armv8.4-a and extensions.
14718 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14720 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14723 * config/rx/rx.md (BW): New mode attribute.
14724 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14726 2018-01-11 Richard Biener <rguenther@suse.de>
14728 PR tree-optimization/83435
14729 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14730 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14731 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14733 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14734 Alan Hayward <alan.hayward@arm.com>
14735 David Sherwood <david.sherwood@arm.com>
14737 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14739 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14740 (aarch64_print_address_internal): Use it to check for a zero offset.
14742 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14743 Alan Hayward <alan.hayward@arm.com>
14744 David Sherwood <david.sherwood@arm.com>
14746 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14747 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14748 Return a poly_int64 rather than a HOST_WIDE_INT.
14749 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14750 rather than a HOST_WIDE_INT.
14751 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14752 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14753 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14754 final_offset from HOST_WIDE_INT to poly_int64.
14755 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14756 to_constant when getting the number of units in an Advanced SIMD
14758 (aarch64_builtin_vectorized_function): Check for a constant number
14760 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14762 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14763 attribute instead of GET_MODE_NUNITS.
14764 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14765 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14766 GET_MODE_SIZE for fixed-size registers.
14767 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14768 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14769 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14770 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14771 (aarch64_print_operand, aarch64_print_address_internal)
14772 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14773 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14774 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14775 Handle polynomial GET_MODE_SIZE.
14776 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14777 wider than SImode without modification.
14778 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14779 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14780 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14781 passing and returning SVE modes.
14782 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14783 rather than GEN_INT.
14784 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14785 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14786 (aarch64_allocate_and_probe_stack_space): Likewise.
14787 (aarch64_layout_frame): Cope with polynomial offsets.
14788 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14789 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14790 polynomial offsets.
14791 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14792 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14793 poly_int64 rather than a HOST_WIDE_INT.
14794 (aarch64_get_separate_components, aarch64_process_components)
14795 (aarch64_expand_prologue, aarch64_expand_epilogue)
14796 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14797 (aarch64_anchor_offset): New function, split out from...
14798 (aarch64_legitimize_address): ...here.
14799 (aarch64_builtin_vectorization_cost): Handle polynomial
14800 TYPE_VECTOR_SUBPARTS.
14801 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14803 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14804 number of elements from the PARALLEL rather than the mode.
14805 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14806 rather than GET_MODE_BITSIZE.
14807 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14808 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14809 (aarch64_expand_vec_perm_const_1): Handle polynomial
14810 d->perm.length () and d->perm elements.
14811 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14812 Apply to_constant to d->perm elements.
14813 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14814 polynomial CONST_VECTOR_NUNITS.
14815 (aarch64_move_pointer): Take amount as a poly_int64 rather
14817 (aarch64_progress_pointer): Avoid temporary variable.
14818 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14819 the mode attribute instead of GET_MODE.
14821 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14822 Alan Hayward <alan.hayward@arm.com>
14823 David Sherwood <david.sherwood@arm.com>
14825 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14826 x exists before using it.
14827 (aarch64_add_constant_internal): Rename to...
14828 (aarch64_add_offset_1): ...this. Replace regnum with separate
14829 src and dest rtxes. Handle the case in which they're different,
14830 including when the offset is zero. Replace scratchreg with an rtx.
14831 Use 2 additions if there is no spare register into which we can
14832 move a 16-bit constant.
14833 (aarch64_add_constant): Delete.
14834 (aarch64_add_offset): Replace reg with separate src and dest
14835 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14836 Use aarch64_add_offset_1.
14837 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14838 an rtx rather than an int. Take the delta as a poly_int64
14839 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14840 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14841 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14842 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14843 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14844 and aarch64_add_sp.
14845 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14846 aarch64_add_constant.
14848 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14850 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14851 Use scalar_float_mode.
14853 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14855 * config/aarch64/aarch64-simd.md
14856 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14857 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14858 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14859 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14860 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14861 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14862 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14863 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14864 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14865 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14867 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14870 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14871 targ_options->x_arm_arch_string is non NULL.
14873 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14875 * config/aarch64/aarch64.h
14876 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14878 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14881 * expmed.c (emit_store_flag_force): Swap if const op0
14882 and change VOIDmode to mode of op0.
14884 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14886 PR rtl-optimization/83761
14887 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14888 than bytes to mode_for_size.
14890 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14892 PR middle-end/83189
14893 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14894 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14897 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14899 PR middle-end/83575
14900 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14901 when in layout mode.
14902 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14903 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14906 2018-01-10 Michael Collison <michael.collison@arm.com>
14908 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14909 * config/aarch64/aarch64-option-extension.def: Add
14910 AARCH64_OPT_EXTENSION of 'fp16fml'.
14911 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14912 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14913 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14914 * config/aarch64/constraints.md (Ui7): New constraint.
14915 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14916 (VFMLA_SEL_W): Ditto.
14919 (VFMLA16_LOW): New int iterator.
14920 (VFMLA16_HIGH): Ditto.
14921 (UNSPEC_FMLAL): New unspec.
14922 (UNSPEC_FMLSL): Ditto.
14923 (UNSPEC_FMLAL2): Ditto.
14924 (UNSPEC_FMLSL2): Ditto.
14925 (f16mac): New code attribute.
14926 * config/aarch64/aarch64-simd-builtins.def
14927 (aarch64_fmlal_lowv2sf): Ditto.
14928 (aarch64_fmlsl_lowv2sf): Ditto.
14929 (aarch64_fmlalq_lowv4sf): Ditto.
14930 (aarch64_fmlslq_lowv4sf): Ditto.
14931 (aarch64_fmlal_highv2sf): Ditto.
14932 (aarch64_fmlsl_highv2sf): Ditto.
14933 (aarch64_fmlalq_highv4sf): Ditto.
14934 (aarch64_fmlslq_highv4sf): Ditto.
14935 (aarch64_fmlal_lane_lowv2sf): Ditto.
14936 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14937 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14938 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14939 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14940 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14941 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14942 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14943 (aarch64_fmlal_lane_highv2sf): Ditto.
14944 (aarch64_fmlsl_lane_highv2sf): Ditto.
14945 (aarch64_fmlal_laneq_highv2sf): Ditto.
14946 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14947 (aarch64_fmlalq_lane_highv4sf): Ditto.
14948 (aarch64_fmlsl_lane_highv4sf): Ditto.
14949 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14950 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14951 * config/aarch64/aarch64-simd.md:
14952 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14953 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14954 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14955 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14956 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14957 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14958 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14959 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14960 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14961 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14962 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14963 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14964 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14965 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14966 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14967 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14968 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14969 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14970 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14971 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14972 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14973 (vfmlsl_low_u32): Ditto.
14974 (vfmlalq_low_u32): Ditto.
14975 (vfmlslq_low_u32): Ditto.
14976 (vfmlal_high_u32): Ditto.
14977 (vfmlsl_high_u32): Ditto.
14978 (vfmlalq_high_u32): Ditto.
14979 (vfmlslq_high_u32): Ditto.
14980 (vfmlal_lane_low_u32): Ditto.
14981 (vfmlsl_lane_low_u32): Ditto.
14982 (vfmlal_laneq_low_u32): Ditto.
14983 (vfmlsl_laneq_low_u32): Ditto.
14984 (vfmlalq_lane_low_u32): Ditto.
14985 (vfmlslq_lane_low_u32): Ditto.
14986 (vfmlalq_laneq_low_u32): Ditto.
14987 (vfmlslq_laneq_low_u32): Ditto.
14988 (vfmlal_lane_high_u32): Ditto.
14989 (vfmlsl_lane_high_u32): Ditto.
14990 (vfmlal_laneq_high_u32): Ditto.
14991 (vfmlsl_laneq_high_u32): Ditto.
14992 (vfmlalq_lane_high_u32): Ditto.
14993 (vfmlslq_lane_high_u32): Ditto.
14994 (vfmlalq_laneq_high_u32): Ditto.
14995 (vfmlslq_laneq_high_u32): Ditto.
14996 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14997 (AARCH64_FL_FOR_ARCH8_4): New.
14998 (AARCH64_ISA_F16FML): New ISA flag.
14999 (TARGET_F16FML): New feature flag for fp16fml.
15000 (doc/invoke.texi): Document new fp16fml option.
15002 2018-01-10 Michael Collison <michael.collison@arm.com>
15004 * config/aarch64/aarch64-builtins.c:
15005 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
15006 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15007 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
15008 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
15009 (AARCH64_ISA_SHA3): New ISA flag.
15010 (TARGET_SHA3): New feature flag for sha3.
15011 * config/aarch64/iterators.md (sha512_op): New int attribute.
15012 (CRYPTO_SHA512): New int iterator.
15013 (UNSPEC_SHA512H): New unspec.
15014 (UNSPEC_SHA512H2): Ditto.
15015 (UNSPEC_SHA512SU0): Ditto.
15016 (UNSPEC_SHA512SU1): Ditto.
15017 * config/aarch64/aarch64-simd-builtins.def
15018 (aarch64_crypto_sha512hqv2di): New builtin.
15019 (aarch64_crypto_sha512h2qv2di): Ditto.
15020 (aarch64_crypto_sha512su0qv2di): Ditto.
15021 (aarch64_crypto_sha512su1qv2di): Ditto.
15022 (aarch64_eor3qv8hi): Ditto.
15023 (aarch64_rax1qv2di): Ditto.
15024 (aarch64_xarqv2di): Ditto.
15025 (aarch64_bcaxqv8hi): Ditto.
15026 * config/aarch64/aarch64-simd.md:
15027 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
15028 (aarch64_crypto_sha512su0qv2di): Ditto.
15029 (aarch64_crypto_sha512su1qv2di): Ditto.
15030 (aarch64_eor3qv8hi): Ditto.
15031 (aarch64_rax1qv2di): Ditto.
15032 (aarch64_xarqv2di): Ditto.
15033 (aarch64_bcaxqv8hi): Ditto.
15034 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
15035 (vsha512h2q_u64): Ditto.
15036 (vsha512su0q_u64): Ditto.
15037 (vsha512su1q_u64): Ditto.
15038 (veor3q_u16): Ditto.
15039 (vrax1q_u64): Ditto.
15040 (vxarq_u64): Ditto.
15041 (vbcaxq_u16): Ditto.
15042 * config/arm/types.md (crypto_sha512): New type attribute.
15043 (crypto_sha3): Ditto.
15044 (doc/invoke.texi): Document new sha3 option.
15046 2018-01-10 Michael Collison <michael.collison@arm.com>
15048 * config/aarch64/aarch64-builtins.c:
15049 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
15050 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15051 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
15052 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
15053 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
15054 (AARCH64_ISA_SM4): New ISA flag.
15055 (TARGET_SM4): New feature flag for sm4.
15056 * config/aarch64/aarch64-simd-builtins.def
15057 (aarch64_sm3ss1qv4si): Ditto.
15058 (aarch64_sm3tt1aq4si): Ditto.
15059 (aarch64_sm3tt1bq4si): Ditto.
15060 (aarch64_sm3tt2aq4si): Ditto.
15061 (aarch64_sm3tt2bq4si): Ditto.
15062 (aarch64_sm3partw1qv4si): Ditto.
15063 (aarch64_sm3partw2qv4si): Ditto.
15064 (aarch64_sm4eqv4si): Ditto.
15065 (aarch64_sm4ekeyqv4si): Ditto.
15066 * config/aarch64/aarch64-simd.md:
15067 (aarch64_sm3ss1qv4si): Ditto.
15068 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
15069 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
15070 (aarch64_sm4eqv4si): Ditto.
15071 (aarch64_sm4ekeyqv4si): Ditto.
15072 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
15073 (sm3part_op): Ditto.
15074 (CRYPTO_SM3TT): Ditto.
15075 (CRYPTO_SM3PART): Ditto.
15076 (UNSPEC_SM3SS1): New unspec.
15077 (UNSPEC_SM3TT1A): Ditto.
15078 (UNSPEC_SM3TT1B): Ditto.
15079 (UNSPEC_SM3TT2A): Ditto.
15080 (UNSPEC_SM3TT2B): Ditto.
15081 (UNSPEC_SM3PARTW1): Ditto.
15082 (UNSPEC_SM3PARTW2): Ditto.
15083 (UNSPEC_SM4E): Ditto.
15084 (UNSPEC_SM4EKEY): Ditto.
15085 * config/aarch64/constraints.md (Ui2): New constraint.
15086 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
15087 * config/arm/types.md (crypto_sm3): New type attribute.
15088 (crypto_sm4): Ditto.
15089 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
15090 (vsm3tt1aq_u32): Ditto.
15091 (vsm3tt1bq_u32): Ditto.
15092 (vsm3tt2aq_u32): Ditto.
15093 (vsm3tt2bq_u32): Ditto.
15094 (vsm3partw1q_u32): Ditto.
15095 (vsm3partw2q_u32): Ditto.
15096 (vsm4eq_u32): Ditto.
15097 (vsm4ekeyq_u32): Ditto.
15098 (doc/invoke.texi): Document new sm4 option.
15100 2018-01-10 Michael Collison <michael.collison@arm.com>
15102 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
15103 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
15104 (AARCH64_FL_FOR_ARCH8_4): New.
15105 (AARCH64_FL_V8_4): New flag.
15106 (doc/invoke.texi): Document new armv8.4-a option.
15108 2018-01-10 Michael Collison <michael.collison@arm.com>
15110 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
15111 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
15112 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
15113 * config/aarch64/aarch64-option-extension.def: Add
15114 AARCH64_OPT_EXTENSION of 'sha2'.
15115 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
15116 (crypto): Disable sha2 and aes if crypto disabled.
15117 (crypto): Enable aes and sha2 if enabled.
15118 (simd): Disable sha2 and aes if simd disabled.
15119 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
15121 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
15122 (TARGET_SHA2): New feature flag for sha2.
15123 (TARGET_AES): New feature flag for aes.
15124 * config/aarch64/aarch64-simd.md:
15125 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
15126 conditional on TARGET_AES.
15127 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
15128 (aarch64_crypto_sha1hsi): Make pattern conditional
15130 (aarch64_crypto_sha1hv4si): Ditto.
15131 (aarch64_be_crypto_sha1hv4si): Ditto.
15132 (aarch64_crypto_sha1su1v4si): Ditto.
15133 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
15134 (aarch64_crypto_sha1su0v4si): Ditto.
15135 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
15136 (aarch64_crypto_sha256su0v4si): Ditto.
15137 (aarch64_crypto_sha256su1v4si): Ditto.
15138 (doc/invoke.texi): Document new aes and sha2 options.
15140 2018-01-10 Martin Sebor <msebor@redhat.com>
15142 PR tree-optimization/83781
15143 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
15146 2018-01-11 Martin Sebor <msebor@gmail.com>
15147 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15149 PR tree-optimization/83501
15150 PR tree-optimization/81703
15152 * tree-ssa-strlen.c (get_string_cst): Rename...
15153 (get_string_len): ...to this. Handle global constants.
15154 (handle_char_store): Adjust.
15156 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
15157 Jim Wilson <jimw@sifive.com>
15159 * config/riscv/riscv-protos.h (riscv_output_return): New.
15160 * config/riscv/riscv.c (struct machine_function): New naked_p field.
15161 (riscv_attribute_table, riscv_output_return),
15162 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
15163 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
15164 (riscv_compute_frame_info): Only compute frame->mask if not a naked
15166 (riscv_expand_prologue): Add early return for naked function.
15167 (riscv_expand_epilogue): Likewise.
15168 (riscv_function_ok_for_sibcall): Return false for naked function.
15169 (riscv_set_current_function): New.
15170 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
15171 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
15172 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
15173 * doc/extend.texi (RISC-V Function Attributes): New.
15175 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15177 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
15178 check for 128-bit long double before checking TCmode.
15179 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
15180 128-bit long doubles before checking TFmode or TCmode.
15181 (FLOAT128_IBM_P): Likewise.
15183 2018-01-10 Martin Sebor <msebor@redhat.com>
15185 PR tree-optimization/83671
15186 * builtins.c (c_strlen): Unconditionally return zero for the empty
15188 Use -Warray-bounds for warnings.
15189 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
15190 for non-constant array indices with COMPONENT_REF, arrays of
15191 arrays, and pointers to arrays.
15192 (gimple_fold_builtin_strlen): Determine and set length range for
15193 non-constant character arrays.
15195 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
15197 PR middle-end/81897
15198 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
15201 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
15203 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
15205 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15208 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
15209 VECTOR_MEM_ALTIVEC_OR_VSX_P.
15210 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
15211 indexed_or_indirect_operand predicate.
15212 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
15213 (*vsx_le_perm_load_v8hi): Likewise.
15214 (*vsx_le_perm_load_v16qi): Likewise.
15215 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
15216 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
15217 (*vsx_le_perm_store_v8hi): Likewise.
15218 (*vsx_le_perm_store_v16qi): Likewise.
15219 (eight unnamed splitters): Likewise.
15221 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
15223 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
15224 * config/rs6000/emmintrin.h: Likewise.
15225 * config/rs6000/mmintrin.h: Likewise.
15226 * config/rs6000/xmmintrin.h: Likewise.
15228 2018-01-10 David Malcolm <dmalcolm@redhat.com>
15231 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
15233 * tree.c (tree_nop_conversion): Return true for location wrapper
15235 (maybe_wrap_with_location): New function.
15236 (selftest::check_strip_nops): New function.
15237 (selftest::test_location_wrappers): New function.
15238 (selftest::tree_c_tests): Call it.
15239 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
15240 (maybe_wrap_with_location): New decl.
15241 (EXPR_LOCATION_WRAPPER_P): New macro.
15242 (location_wrapper_p): New inline function.
15243 (tree_strip_any_location_wrapper): New inline function.
15245 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
15248 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
15249 stack_realign_offset for the largest alignment of stack slot
15251 (ix86_find_max_used_stack_alignment): New function.
15252 (ix86_finalize_stack_frame_flags): Use it. Set
15253 max_used_stack_alignment if we don't realign stack.
15254 * config/i386/i386.h (machine_function): Add
15255 max_used_stack_alignment.
15257 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
15259 * config/arm/arm.opt (-mbranch-cost): New option.
15260 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
15263 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
15266 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
15267 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
15269 2018-01-10 Richard Biener <rguenther@suse.de>
15272 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
15273 early out so it also covers the case where we have a non-NULL
15276 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15278 PR tree-optimization/83753
15279 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
15280 for non-strided grouped accesses if the number of elements is 1.
15282 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
15285 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
15286 * i386.h (TARGET_USE_GATHER): Define.
15287 * x86-tune.def (X86_TUNE_USE_GATHER): New.
15289 2018-01-10 Martin Liska <mliska@suse.cz>
15292 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
15293 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
15295 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
15296 CLEANUP_NO_PARTITIONING is not set.
15298 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
15300 * doc/rtl.texi: Remove documentation of (const ...) wrappers
15301 for vectors, as a partial revert of r254296.
15302 * rtl.h (const_vec_p): Delete.
15303 (const_vec_duplicate_p): Don't test for vector CONSTs.
15304 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
15305 * expmed.c (make_tree): Likewise.
15308 * common.md (E, F): Use CONSTANT_P instead of checking for
15310 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
15311 checking for CONST_VECTOR.
15313 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15315 PR middle-end/83575
15316 * predict.c (force_edge_cold): Handle in more sane way edges
15317 with no prediction.
15319 2018-01-09 Carl Love <cel@us.ibm.com>
15321 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
15323 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
15324 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
15325 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
15326 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
15327 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
15328 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
15329 * config/rs6000/rs6000-protos.h: Add extern defition for
15330 rs6000_generate_float2_double_code.
15331 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
15333 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
15334 (float2_v2df): Add define_expand.
15336 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
15339 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
15340 op_mode in the force_to_mode call.
15342 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15344 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
15345 instead of checking each element individually.
15346 (aarch64_evpc_uzp): Likewise.
15347 (aarch64_evpc_zip): Likewise.
15348 (aarch64_evpc_ext): Likewise.
15349 (aarch64_evpc_rev): Likewise.
15350 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
15351 instead of checking each element individually. Return true without
15353 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
15354 whether all selected elements come from the same input, instead of
15355 checking each element individually. Remove calls to gen_rtx_REG,
15356 start_sequence and end_sequence and instead assert that no rtl is
15359 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15361 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
15362 order of HIGH and CONST checks.
15364 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
15366 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
15367 if the destination isn't an SSA_NAME.
15369 2018-01-09 Richard Biener <rguenther@suse.de>
15371 PR tree-optimization/83668
15372 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
15374 (canonicalize_loop_form): ... here, renamed from ...
15375 (canonicalize_loop_closed_ssa_form): ... this and amended to
15376 swap successor edges for loop exit blocks to make us use
15377 the RPO order we need for initial schedule generation.
15379 2018-01-09 Joseph Myers <joseph@codesourcery.com>
15381 PR tree-optimization/64811
15382 * match.pd: When optimizing comparisons with Inf, avoid
15383 introducing or losing exceptions from comparisons with NaN.
15385 2018-01-09 Martin Liska <mliska@suse.cz>
15388 * asan.c (shadow_mem_size): Add gcc_assert.
15390 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
15392 Don't save registers in main().
15395 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
15396 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
15397 * config/avr/avr.c (avr_set_current_function): Don't error if
15398 naked, OS_task or OS_main are specified at the same time.
15399 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
15401 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
15403 * common/config/avr/avr-common.c (avr_option_optimization_table):
15404 Switch on -mmain-is-OS_task for optimizing compilations.
15406 2018-01-09 Richard Biener <rguenther@suse.de>
15408 PR tree-optimization/83572
15409 * graphite.c: Include cfganal.h.
15410 (graphite_transform_loops): Connect infinite loops to exit
15411 and remove fake edges at the end.
15413 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15415 * ipa-inline.c (edge_badness): Revert accidental checkin.
15417 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
15420 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
15421 symbols; not inline clones.
15423 2018-01-09 Jakub Jelinek <jakub@redhat.com>
15426 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
15427 hard registers. Formatting fixes.
15429 PR preprocessor/83722
15430 * gcc.c (try_generate_repro): Pass
15431 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
15432 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
15435 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
15436 Kito Cheng <kito.cheng@gmail.com>
15438 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
15439 (riscv_leaf_function_p): Delete.
15440 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
15442 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15444 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
15446 (do_ifelse): New function.
15447 (do_isel): New function.
15448 (do_sub3): New function.
15449 (do_add3): New function.
15450 (do_load_mask_compare): New function.
15451 (do_overlap_load_compare): New function.
15452 (expand_compare_loop): New function.
15453 (expand_block_compare): Call expand_compare_loop() when appropriate.
15454 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
15455 option description.
15456 (-mblock-compare-inline-loop-limit): New option.
15458 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15461 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
15462 Reverse order of second and third operands in first alternative.
15463 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
15464 of first and second elements in UNSPEC_VPERMR vector.
15465 (altivec_expand_vec_perm_le): Likewise.
15467 2018-01-08 Jeff Law <law@redhat.com>
15469 PR rtl-optimizatin/81308
15470 * tree-switch-conversion.c (cfg_altered): New file scoped static.
15471 (process_switch): If group_case_labels makes a change, then set
15473 (pass_convert_switch::execute): If a switch is converted, then
15474 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
15476 PR rtl-optimization/81308
15477 * recog.c (split_all_insns): Conditionally cleanup the CFG after
15480 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
15482 PR target/83663 - Revert r255946
15483 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
15484 generation for cases where splatting a value is not useful.
15485 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
15486 across a vec_duplicate and a paradoxical subreg forming a vector
15487 mode to a vec_concat.
15489 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15491 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
15492 -march=armv8.3-a variants.
15493 * config/arm/t-multilib: Likewise.
15494 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
15496 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15498 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
15500 (cceq_ior_compare_complement): Give it a name so I can use it, and
15501 change boolean_or_operator predicate to boolean_operator so it can
15502 be used to generate a crand.
15503 (eqne): New code iterator.
15504 (bd/bd_neg): New code_attrs.
15505 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
15506 a single define_insn.
15507 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
15508 decrement (bdnzt/bdnzf/bdzt/bdzf).
15509 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
15510 with the new names of the branch decrement patterns, and added the
15511 names of the branch decrement conditional patterns.
15513 2018-01-08 Richard Biener <rguenther@suse.de>
15515 PR tree-optimization/83563
15516 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
15519 2018-01-08 Richard Biener <rguenther@suse.de>
15521 PR middle-end/83713
15522 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15524 2018-01-08 Richard Biener <rguenther@suse.de>
15526 PR tree-optimization/83685
15527 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15528 references to abnormals.
15530 2018-01-08 Richard Biener <rguenther@suse.de>
15533 * dwarf2out.c (output_indirect_strings): Handle empty
15534 skeleton_debug_str_hash.
15535 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15537 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15539 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15540 (emit_store_direct): Likewise.
15541 (arc_trampoline_adjust_address): Likewise.
15542 (arc_asm_trampoline_template): New function.
15543 (arc_initialize_trampoline): Use asm_trampoline_template.
15544 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15545 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15546 * config/arc/arc.md (flush_icache): Delete pattern.
15548 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15550 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15551 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15554 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15557 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15558 by not USED_FOR_TARGET.
15559 (make_pass_resolve_sw_modes): Likewise.
15561 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15563 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15566 2018-01-08 Richard Biener <rguenther@suse.de>
15568 PR middle-end/83580
15569 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15571 2018-01-08 Richard Biener <rguenther@suse.de>
15573 PR middle-end/83517
15574 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15576 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15578 PR middle-end/81897
15579 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15580 basic blocks with a small number of successors.
15581 (convert_control_dep_chain_into_preds): Improve handling of
15583 (dump_predicates): Split apart into...
15584 (dump_pred_chain): ...here...
15585 (dump_pred_info): ...and here.
15586 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15587 (can_chain_union_be_invalidated_p): Improve check for invalidation
15589 (uninit_uses_cannot_happen): Avoid unnecessary if
15590 convert_control_dep_chain_into_preds yielded nothing.
15592 2018-01-06 Martin Sebor <msebor@redhat.com>
15594 PR tree-optimization/83640
15595 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15596 subtracting negative offset from size.
15597 (builtin_access::overlap): Adjust offset bounds of the access to fall
15598 within the size of the object if possible.
15600 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15602 PR rtl-optimization/83699
15603 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15604 extract_bit_field_as_subreg to cases in which the extracted
15605 value is also a vector.
15607 * lra-constraints.c (process_alt_operands): Test for the equivalence
15608 substitutions when detecting a possible reload cycle.
15610 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15613 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15614 by default if flag_selective_schedling{,2}. Formatting fixes.
15616 PR rtl-optimization/83682
15617 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15618 if it has non-VECTOR_MODE element mode.
15619 (vec_duplicate_p): Likewise.
15621 PR middle-end/83694
15622 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15623 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15625 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15628 * config/i386/i386-builtin.def
15629 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15630 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15631 Require also OPTION_MASK_ISA_AVX512F in addition to
15632 OPTION_MASK_ISA_GFNI.
15633 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15634 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15635 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15636 to OPTION_MASK_ISA_GFNI.
15637 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15638 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15639 OPTION_MASK_ISA_AVX512BW.
15640 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15641 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15642 addition to OPTION_MASK_ISA_GFNI.
15643 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15644 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15645 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15646 to OPTION_MASK_ISA_GFNI.
15647 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15648 a requirement for all ISAs rather than any of them with a few
15650 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15652 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15653 bitmasks to be enabled with 3 exceptions, instead of requiring any
15654 enabled ISA with lots of exceptions.
15655 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15656 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15657 Change avx512bw in isa attribute to avx512f.
15658 * config/i386/sgxintrin.h: Add license boilerplate.
15659 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15660 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15661 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15662 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15664 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15665 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15666 temporarily sse2 rather than sse if not enabled already.
15669 * config/i386/sse.md (VI248_VLBW): Rename to ...
15670 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15671 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15672 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15673 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15674 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15675 mode iterator instead of VI248_VLBW.
15677 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15679 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15680 (record_modified): Skip clobbers; add debug output.
15681 (param_change_prob): Use sreal frequencies.
15683 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15685 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15686 punt for user-aligned variables.
15688 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15690 * tree-chrec.c (chrec_contains_symbols): Return true for
15693 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15696 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15697 of (x|y) == x for BICS pattern.
15699 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15701 PR tree-optimization/83605
15702 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15703 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15706 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15708 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15709 * config/epiphany/rtems.h: New file.
15711 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15712 Uros Bizjak <ubizjak@gmail.com>
15715 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15716 QIreg_operand instead of register_operand predicate.
15717 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15718 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15719 comments instead of -fmitigate[-_]rop.
15721 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15724 * cgraphunit.c (symbol_table::compile): Switch to text_section
15725 before calling assembly_start debug hook.
15726 * run-rtl-passes.c (run_rtl_passes): Likewise.
15729 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15731 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15732 range_int_cst_p rather than !symbolic_range_p before calling
15733 extract_range_from_multiplicative_op_1.
15735 2018-01-04 Jeff Law <law@redhat.com>
15737 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15738 redundant test in assertion.
15740 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15742 * doc/rtl.texi: Document machine_mode wrapper classes.
15744 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15746 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15747 using tree_to_uhwi.
15749 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15751 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15752 the VEC_PERM_EXPR fold to fail.
15754 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15757 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15758 to switched_sections.
15760 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15763 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15764 test for d.testing.
15766 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15769 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15770 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15772 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15775 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15776 is BLKmode and bitpos not zero or mode change is needed.
15778 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15781 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15784 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15787 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15788 instead of MULT rtx. Update all corresponding splitters.
15789 (*saddl_se): Ditto.
15790 (*ssub<modesuffix>): Ditto.
15791 (*ssubl_se): Ditto.
15792 (*cmp_sadd_di): Update split patterns.
15793 (*cmp_sadd_si): Ditto.
15794 (*cmp_sadd_sidi): Ditto.
15795 (*cmp_ssub_di): Ditto.
15796 (*cmp_ssub_si): Ditto.
15797 (*cmp_ssub_sidi): Ditto.
15798 * config/alpha/predicates.md (const23_operand): New predicate.
15799 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15800 Look for ASHIFT, not MULT inner operand.
15801 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15803 2018-01-04 Martin Liska <mliska@suse.cz>
15805 PR gcov-profile/83669
15806 * gcov.c (output_intermediate_file): Add version to intermediate
15808 * doc/gcov.texi: Document new field 'version' in intermediate
15809 file format. Fix location of '-k' option of gcov command.
15811 2018-01-04 Martin Liska <mliska@suse.cz>
15814 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15816 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15818 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15820 2018-01-03 Martin Sebor <msebor@redhat.com>
15822 PR tree-optimization/83655
15823 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15824 checking calls with invalid arguments.
15826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15828 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15829 (vectorizable_mask_load_store): Delete.
15830 (vectorizable_call): Return false for masked loads and stores.
15831 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15832 instead of gimple_assign_rhs1.
15833 (vectorizable_load): Handle IFN_MASK_LOAD.
15834 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15838 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15840 (vectorizable_mask_load_store): ...here.
15841 (vectorizable_load): ...and here.
15843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15845 * tree-vect-stmts.c (vect_build_all_ones_mask)
15846 (vect_build_zero_merge_argument): New functions, split out from...
15847 (vectorizable_load): ...here.
15849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15851 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15853 (vectorizable_mask_load_store): ...here.
15854 (vectorizable_store): ...and here.
15856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15858 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15860 (vectorizable_mask_load_store): ...here.
15862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15864 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15865 (vect_model_store_cost): Take a vec_load_store_type instead of a
15867 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15868 (vect_model_store_cost): Take a vec_load_store_type instead of a
15870 (vectorizable_mask_load_store): Update accordingly.
15871 (vectorizable_store): Likewise.
15872 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15876 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15877 IFN_MASK_LOAD calls here rather than...
15878 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15881 Alan Hayward <alan.hayward@arm.com>
15882 David Sherwood <david.sherwood@arm.com>
15884 * expmed.c (extract_bit_field_1): For vector extracts,
15885 fall back to extract_bit_field_as_subreg if vec_extract
15888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15889 Alan Hayward <alan.hayward@arm.com>
15890 David Sherwood <david.sherwood@arm.com>
15892 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15893 they are variable or constant sized.
15894 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15895 slots for constant-sized data.
15897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15898 Alan Hayward <alan.hayward@arm.com>
15899 David Sherwood <david.sherwood@arm.com>
15901 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15902 handling COND_EXPRs with boolean comparisons, try to find a better
15903 basis for the mask type than the boolean itself.
15905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15907 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15908 is calculated and how it can be overridden.
15909 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15910 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15912 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15916 Alan Hayward <alan.hayward@arm.com>
15917 David Sherwood <david.sherwood@arm.com>
15919 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15920 Remove the mode argument.
15921 (aarch64_simd_valid_immediate): Remove the mode and inverse
15923 * config/aarch64/iterators.md (bitsize): New iterator.
15924 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15925 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15926 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15927 aarch64_simd_valid_immediate.
15928 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15929 (aarch64_reg_or_bic_imm): Likewise.
15930 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15931 with an insn_type enum and msl with a modifier_type enum.
15932 Replace element_width with a scalar_mode. Change the shift
15933 to unsigned int. Add constructors for scalar_float_mode and
15934 scalar_int_mode elements.
15935 (aarch64_vect_float_const_representable_p): Delete.
15936 (aarch64_can_const_movi_rtx_p)
15937 (aarch64_simd_scalar_immediate_valid_for_move)
15938 (aarch64_simd_make_constant): Update call to
15939 aarch64_simd_valid_immediate.
15940 (aarch64_advsimd_valid_immediate_hs): New function.
15941 (aarch64_advsimd_valid_immediate): Likewise.
15942 (aarch64_simd_valid_immediate): Remove mode and inverse
15943 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15944 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15945 and aarch64_float_const_representable_p on the result.
15946 (aarch64_output_simd_mov_immediate): Remove mode argument.
15947 Update call to aarch64_simd_valid_immediate and use of
15948 simd_immediate_info.
15949 (aarch64_output_scalar_simd_mov_immediate): Update call
15952 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15953 Alan Hayward <alan.hayward@arm.com>
15954 David Sherwood <david.sherwood@arm.com>
15956 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15957 (mode_nunits): Likewise CONST_MODE_NUNITS.
15958 * machmode.def (ADJUST_NUNITS): Document.
15959 * genmodes.c (mode_data::need_nunits_adj): New field.
15960 (blank_mode): Update accordingly.
15961 (adj_nunits): New variable.
15962 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15964 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15965 listed in adj_nunits.
15966 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15967 listed in adj_nunits. Don't emit case statements for such modes.
15968 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15969 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15970 nothing if adj_nunits is nonnull.
15971 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15972 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15973 (emit_mode_fbit): Update use of print_maybe_const_decl.
15974 (emit_move_size): Likewise. Treat the array as non-const
15976 (emit_mode_adjustments): Handle adj_nunits.
15978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15980 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15981 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15982 (VECTOR_MODES): Use it.
15983 (make_vector_modes): Take the prefix as an argument.
15985 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15986 Alan Hayward <alan.hayward@arm.com>
15987 David Sherwood <david.sherwood@arm.com>
15989 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15990 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15991 for MODE_VECTOR_BOOL.
15992 * machmode.def (VECTOR_BOOL_MODE): Document.
15993 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15994 (make_vector_bool_mode): New function.
15995 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15997 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15998 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
16000 * stor-layout.c (int_mode_for_mode): Likewise.
16001 * tree.c (build_vector_type_for_mode): Likewise.
16002 * varasm.c (output_constant_pool_2): Likewise.
16003 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
16004 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
16005 for MODE_VECTOR_BOOL.
16006 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
16007 of mode class checks.
16008 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
16009 instead of a list of mode class checks.
16010 (expand_vector_scalar_condition): Likewise.
16011 (type_for_widest_vector_mode): Handle BImode as an inner mode.
16013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16014 Alan Hayward <alan.hayward@arm.com>
16015 David Sherwood <david.sherwood@arm.com>
16017 * machmode.h (mode_size): Change from unsigned short to
16019 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
16020 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16021 or if measurement_type is not polynomial.
16022 (fixed_size_mode::includes_p): Check for constant-sized modes.
16023 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
16024 return a poly_uint16 rather than an unsigned short.
16025 (emit_mode_size): Change the type of mode_size from unsigned short
16026 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
16027 (emit_mode_adjustments): Cope with polynomial vector sizes.
16028 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16030 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16032 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
16033 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
16034 * caller-save.c (setup_save_areas): Likewise.
16035 (replace_reg_with_saved_mem): Likewise.
16036 * calls.c (emit_library_call_value_1): Likewise.
16037 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
16038 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
16039 (gen_lowpart_for_combine): Likewise.
16040 * convert.c (convert_to_integer_1): Likewise.
16041 * cse.c (equiv_constant, cse_insn): Likewise.
16042 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
16043 (cselib_subst_to_values): Likewise.
16044 * dce.c (word_dce_process_block): Likewise.
16045 * df-problems.c (df_word_lr_mark_ref): Likewise.
16046 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
16047 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
16048 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
16049 (rtl_for_decl_location): Likewise.
16050 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
16051 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
16052 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
16053 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
16054 (expand_expr_real_1): Likewise.
16055 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
16056 (pad_below): Likewise.
16057 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16058 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
16059 * ira.c (get_subreg_tracking_sizes): Likewise.
16060 * ira-build.c (ira_create_allocno_objects): Likewise.
16061 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
16062 (ira_sort_regnos_for_alter_reg): Likewise.
16063 * ira-costs.c (record_operand_costs): Likewise.
16064 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
16065 (resolve_simple_move): Likewise.
16066 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
16067 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
16068 (lra_constraints): Likewise.
16069 (CONST_POOL_OK_P): Reject variable-sized modes.
16070 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
16071 (add_pseudo_to_slot, lra_spill): Likewise.
16072 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16073 * optabs-query.c (get_best_extraction_insn): Likewise.
16074 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16075 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
16076 (expand_mult_highpart, valid_multiword_target_p): Likewise.
16077 * recog.c (offsettable_address_addr_space_p): Likewise.
16078 * regcprop.c (maybe_mode_change): Likewise.
16079 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
16080 * regrename.c (build_def_use): Likewise.
16081 * regstat.c (dump_reg_info): Likewise.
16082 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
16083 (find_reloads, find_reloads_subreg_address): Likewise.
16084 * reload1.c (eliminate_regs_1): Likewise.
16085 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
16086 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
16087 (simplify_binary_operation_1, simplify_subreg): Likewise.
16088 * targhooks.c (default_function_arg_padding): Likewise.
16089 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
16090 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
16091 (verify_gimple_assign_ternary): Likewise.
16092 * tree-inline.c (estimate_move_cost): Likewise.
16093 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16094 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
16095 (get_address_cost_ainc): Likewise.
16096 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
16097 (vect_supportable_dr_alignment): Likewise.
16098 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16099 (vectorizable_reduction): Likewise.
16100 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
16101 (vectorizable_operation, vectorizable_load): Likewise.
16102 * tree.c (build_same_sized_truth_vector_type): Likewise.
16103 * valtrack.c (cleanup_auto_inc_dec): Likewise.
16104 * var-tracking.c (emit_note_insn_var_location): Likewise.
16105 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
16106 (ADDR_VEC_ALIGN): Likewise.
16108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16109 Alan Hayward <alan.hayward@arm.com>
16110 David Sherwood <david.sherwood@arm.com>
16112 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
16114 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
16115 or if measurement_type is polynomial.
16116 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
16117 * combine.c (make_extraction): Likewise.
16118 * dse.c (find_shift_sequence): Likewise.
16119 * dwarf2out.c (mem_loc_descriptor): Likewise.
16120 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
16121 (extract_bit_field, extract_low_bits): Likewise.
16122 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
16123 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
16124 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
16125 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
16126 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
16127 * reload.c (find_reloads): Likewise.
16128 * reload1.c (alter_reg): Likewise.
16129 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
16130 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
16131 * tree-if-conv.c (predicate_mem_writes): Likewise.
16132 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
16133 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
16134 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
16135 * valtrack.c (dead_debug_insert_temp): Likewise.
16136 * varasm.c (mergeable_constant_section): Likewise.
16137 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
16139 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16140 Alan Hayward <alan.hayward@arm.com>
16141 David Sherwood <david.sherwood@arm.com>
16143 * expr.c (expand_assignment): Cope with polynomial mode sizes
16144 when assigning to a CONCAT.
16146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16147 Alan Hayward <alan.hayward@arm.com>
16148 David Sherwood <david.sherwood@arm.com>
16150 * machmode.h (mode_precision): Change from unsigned short to
16152 (mode_to_precision): Return a poly_uint16 rather than an unsigned
16154 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
16155 or if measurement_type is not polynomial.
16156 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
16157 in which the mode is already known to be a scalar_int_mode.
16158 * genmodes.c (emit_mode_precision): Change the type of mode_precision
16159 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
16161 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16162 for GET_MODE_PRECISION.
16163 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16164 for GET_MODE_PRECISION.
16165 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
16167 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
16168 (expand_field_assignment, make_extraction): Likewise.
16169 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
16170 (get_last_value): Likewise.
16171 * convert.c (convert_to_integer_1): Likewise.
16172 * cse.c (cse_insn): Likewise.
16173 * expr.c (expand_expr_real_1): Likewise.
16174 * lra-constraints.c (simplify_operand_subreg): Likewise.
16175 * optabs-query.c (can_atomic_load_p): Likewise.
16176 * optabs.c (expand_atomic_load): Likewise.
16177 (expand_atomic_store): Likewise.
16178 * ree.c (combine_reaching_defs): Likewise.
16179 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
16180 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
16181 * tree.h (type_has_mode_precision_p): Likewise.
16182 * ubsan.c (instrument_si_overflow): Likewise.
16184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16185 Alan Hayward <alan.hayward@arm.com>
16186 David Sherwood <david.sherwood@arm.com>
16188 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
16189 polynomial numbers of units.
16190 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
16191 (valid_vector_subparts_p): New function.
16192 (build_vector_type): Remove temporary shim and take the number
16193 of units as a poly_uint64 rather than an int.
16194 (build_opaque_vector_type): Take the number of units as a
16195 poly_uint64 rather than an int.
16196 * tree.c (build_vector_from_ctor): Handle polynomial
16197 TYPE_VECTOR_SUBPARTS.
16198 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
16199 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
16200 (build_vector_from_val): If the number of units is variable,
16201 use build_vec_duplicate_cst for constant operands and
16202 VEC_DUPLICATE_EXPR otherwise.
16203 (make_vector_type): Remove temporary is_constant ().
16204 (build_vector_type, build_opaque_vector_type): Take the number of
16205 units as a poly_uint64 rather than an int.
16206 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
16208 * cfgexpand.c (expand_debug_expr): Likewise.
16209 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
16210 (store_constructor, expand_expr_real_1): Likewise.
16211 (const_scalar_mask_from_tree): Likewise.
16212 * fold-const-call.c (fold_const_reduction): Likewise.
16213 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
16214 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
16215 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
16216 (fold_relational_const): Likewise.
16217 (native_interpret_vector): Likewise. Change the size from an
16218 int to an unsigned int.
16219 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
16220 TYPE_VECTOR_SUBPARTS.
16221 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
16222 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
16223 duplicating a non-constant operand into a variable-length vector.
16224 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
16225 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
16226 * ipa-icf.c (sem_variable::equals): Likewise.
16227 * match.pd: Likewise.
16228 * omp-simd-clone.c (simd_clone_subparts): Likewise.
16229 * print-tree.c (print_node): Likewise.
16230 * stor-layout.c (layout_type): Likewise.
16231 * targhooks.c (default_builtin_vectorization_cost): Likewise.
16232 * tree-cfg.c (verify_gimple_comparison): Likewise.
16233 (verify_gimple_assign_binary): Likewise.
16234 (verify_gimple_assign_ternary): Likewise.
16235 (verify_gimple_assign_single): Likewise.
16236 * tree-pretty-print.c (dump_generic_node): Likewise.
16237 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16238 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
16239 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
16240 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
16241 (vect_shift_permute_load_chain): Likewise.
16242 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
16243 (expand_vector_condition, optimize_vector_constructor): Likewise.
16244 (lower_vec_perm, get_compute_type): Likewise.
16245 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16246 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
16247 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
16248 (vect_recog_mask_conversion_pattern): Likewise.
16249 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
16250 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
16251 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16252 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
16253 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
16254 (vectorizable_shift, vectorizable_operation, vectorizable_store)
16255 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
16256 (supportable_widening_operation): Likewise.
16257 (supportable_narrowing_operation): Likewise.
16258 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
16260 * varasm.c (output_constant): Likewise.
16262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16263 Alan Hayward <alan.hayward@arm.com>
16264 David Sherwood <david.sherwood@arm.com>
16266 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
16267 so that both the length == 3 and length != 3 cases set up their
16268 own permute vectors. Add comments explaining why we know the
16269 number of elements is constant.
16270 (vect_permute_load_chain): Likewise.
16272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16273 Alan Hayward <alan.hayward@arm.com>
16274 David Sherwood <david.sherwood@arm.com>
16276 * machmode.h (mode_nunits): Change from unsigned char to
16278 (ONLY_FIXED_SIZE_MODES): New macro.
16279 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
16280 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
16281 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
16283 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
16284 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
16285 or if measurement_type is not polynomial.
16286 * genmodes.c (ZERO_COEFFS): New macro.
16287 (emit_mode_nunits_inline): Make mode_nunits_inline return a
16289 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
16290 Use ZERO_COEFFS when emitting initializers.
16291 * data-streamer.h (bp_pack_poly_value): New function.
16292 (bp_unpack_poly_value): Likewise.
16293 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
16294 for GET_MODE_NUNITS.
16295 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
16296 for GET_MODE_NUNITS.
16297 * tree.c (make_vector_type): Remove temporary shim and make
16298 the real function take the number of units as a poly_uint64
16299 rather than an int.
16300 (build_vector_type_for_mode): Handle polynomial nunits.
16301 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
16302 * emit-rtl.c (const_vec_series_p_1): Likewise.
16303 (gen_rtx_CONST_VECTOR): Likewise.
16304 * fold-const.c (test_vec_duplicate_folding): Likewise.
16305 * genrecog.c (validate_pattern): Likewise.
16306 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
16307 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
16308 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
16309 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
16310 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
16311 * rtlanal.c (subreg_get_info): Likewise.
16312 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16313 (vect_grouped_load_supported): Likewise.
16314 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
16315 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
16316 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
16317 (simplify_const_unary_operation, simplify_binary_operation_1)
16318 (simplify_const_binary_operation, simplify_ternary_operation)
16319 (test_vector_ops_duplicate, test_vector_ops): Likewise.
16320 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
16321 instead of CONST_VECTOR_NUNITS.
16322 * varasm.c (output_constant_pool_2): Likewise.
16323 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
16324 explicit-encoded elements in the XVEC for variable-length vectors.
16326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16328 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
16330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16331 Alan Hayward <alan.hayward@arm.com>
16332 David Sherwood <david.sherwood@arm.com>
16334 * coretypes.h (fixed_size_mode): Declare.
16335 (fixed_size_mode_pod): New typedef.
16336 * builtins.h (target_builtins::x_apply_args_mode)
16337 (target_builtins::x_apply_result_mode): Change type to
16338 fixed_size_mode_pod.
16339 * builtins.c (apply_args_size, apply_result_size, result_vector)
16340 (expand_builtin_apply_args_1, expand_builtin_apply)
16341 (expand_builtin_return): Update accordingly.
16343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16345 * cse.c (hash_rtx_cb): Hash only the encoded elements.
16346 * cselib.c (cselib_hash_rtx): Likewise.
16347 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
16348 CONST_VECTOR encoding.
16350 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16351 Jeff Law <law@redhat.com>
16354 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
16355 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
16356 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
16357 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
16360 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
16361 explicitly probe *sp in a noreturn function if there were any callee
16362 register saves or frame pointer is needed.
16364 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16367 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
16368 BLKmode for ternary, binary or unary expressions.
16371 * var-tracking.c (delete_vta_debug_insn): New inline function.
16372 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
16373 insns from get_insns () to NULL instead of each bb separately.
16374 Use delete_vta_debug_insn. No longer static.
16375 (vt_debug_insns_local, variable_tracking_main_1): Adjust
16376 delete_vta_debug_insns callers.
16377 * rtl.h (delete_vta_debug_insns): Declare.
16378 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
16379 instead of variable_tracking_main.
16381 2018-01-03 Martin Sebor <msebor@redhat.com>
16383 PR tree-optimization/83603
16384 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
16385 arguments past the endof the argument list in functions declared
16386 without a prototype.
16387 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
16388 Avoid checking when arguments are null.
16390 2018-01-03 Martin Sebor <msebor@redhat.com>
16393 * doc/extend.texi (attribute const): Fix a typo.
16394 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
16395 issuing -Wsuggest-attribute for void functions.
16397 2018-01-03 Martin Sebor <msebor@redhat.com>
16399 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
16400 offset_int::from instead of wide_int::to_shwi.
16401 (maybe_diag_overlap): Remove assertion.
16402 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
16403 * gimple-ssa-sprintf.c (format_directive): Same.
16404 (parse_directive): Same.
16405 (sprintf_dom_walker::compute_format_length): Same.
16406 (try_substitute_return_value): Same.
16408 2018-01-03 Jeff Law <law@redhat.com>
16410 PR middle-end/83654
16411 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
16412 non-constant residual for zero at runtime and avoid probing in
16413 that case. Reorganize code for trailing problem to mirror handling
16416 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16418 PR tree-optimization/83501
16419 * tree-ssa-strlen.c (get_string_cst): New.
16420 (handle_char_store): Call get_string_cst.
16422 2018-01-03 Martin Liska <mliska@suse.cz>
16424 PR tree-optimization/83593
16425 * tree-ssa-strlen.c: Include tree-cfg.h.
16426 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
16427 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
16428 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
16430 (strlen_dom_walker::before_dom_children): Call
16431 gimple_purge_dead_eh_edges. Dump tranformation with details
16433 (strlen_dom_walker::before_dom_children): Update call by adding
16434 new argument cleanup_eh.
16435 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
16437 2018-01-03 Martin Liska <mliska@suse.cz>
16440 * cif-code.def (VARIADIC_THUNK): New enum value.
16441 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
16444 2018-01-03 Jan Beulich <jbeulich@suse.com>
16446 * sse.md (mov<mode>_internal): Tighten condition for when to use
16447 vmovdqu<ssescalarsize> for TI and OI modes.
16449 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16451 Update copyright years.
16453 2018-01-03 Martin Liska <mliska@suse.cz>
16456 * ipa-visibility.c (function_and_variable_visibility): Skip
16457 functions with noipa attribure.
16459 2018-01-03 Jakub Jelinek <jakub@redhat.com>
16461 * gcc.c (process_command): Update copyright notice dates.
16462 * gcov-dump.c (print_version): Ditto.
16463 * gcov.c (print_version): Ditto.
16464 * gcov-tool.c (print_version): Ditto.
16465 * gengtype.c (create_file): Ditto.
16466 * doc/cpp.texi: Bump @copying's copyright year.
16467 * doc/cppinternals.texi: Ditto.
16468 * doc/gcc.texi: Ditto.
16469 * doc/gccint.texi: Ditto.
16470 * doc/gcov.texi: Ditto.
16471 * doc/install.texi: Ditto.
16472 * doc/invoke.texi: Ditto.
16474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16476 * vector-builder.h (vector_builder::m_full_nelts): Change from
16477 unsigned int to poly_uint64.
16478 (vector_builder::full_nelts): Update prototype accordingly.
16479 (vector_builder::new_vector): Likewise.
16480 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
16481 (vector_builder::operator ==): Likewise.
16482 (vector_builder::finalize): Likewise.
16483 * int-vector-builder.h (int_vector_builder::int_vector_builder):
16484 Take the number of elements as a poly_uint64 rather than an
16486 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
16487 from unsigned int to poly_uint64.
16488 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
16489 (vec_perm_indices::new_vector): Likewise.
16490 (vec_perm_indices::length): Likewise.
16491 (vec_perm_indices::nelts_per_input): Likewise.
16492 (vec_perm_indices::input_nelts): Likewise.
16493 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
16494 number of elements per input as a poly_uint64 rather than an
16495 unsigned int. Use the original encoding for variable-length
16496 vectors, rather than clamping each individual element.
16497 For the second and subsequent elements in each pattern,
16498 clamp the step and base before clamping their sum.
16499 (vec_perm_indices::series_p): Handle polynomial element counts.
16500 (vec_perm_indices::all_in_range_p): Likewise.
16501 (vec_perm_indices_to_tree): Likewise.
16502 (vec_perm_indices_to_rtx): Likewise.
16503 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
16504 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
16505 (tree_vector_builder::new_binary_operation): Handle polynomial
16506 element counts. Return false if we need to know the number
16507 of elements at compile time.
16508 * fold-const.c (fold_vec_perm): Punt if the number of elements
16509 isn't known at compile time.
16511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16513 * vec-perm-indices.h (vec_perm_builder): Change element type
16514 from HOST_WIDE_INT to poly_int64.
16515 (vec_perm_indices::element_type): Update accordingly.
16516 (vec_perm_indices::clamp): Handle polynomial element_types.
16517 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16518 (vec_perm_indices::all_in_range_p): Likewise.
16519 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
16521 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
16522 polynomial vec_perm_indices element types.
16523 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16524 * fold-const.c (fold_vec_perm): Likewise.
16525 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16526 * tree-vect-generic.c (lower_vec_perm): Likewise.
16527 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16528 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16529 element type to HOST_WIDE_INT.
16531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16532 Alan Hayward <alan.hayward@arm.com>
16533 David Sherwood <david.sherwood@arm.com>
16535 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16536 rather than an int. Use plus_constant.
16537 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16538 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16541 Alan Hayward <alan.hayward@arm.com>
16542 David Sherwood <david.sherwood@arm.com>
16544 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16545 a HOST_WIDE_INT to a poly_int64.
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 * calls.c (load_register_parameters): Cope with polynomial
16552 mode sizes. Require a constant size for BLKmode parameters
16553 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16554 forces a parameter to be padded at the lsb end in order to
16555 fill a complete number of words, require the parameter size
16556 to be ordered wrt UNITS_PER_WORD.
16558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16559 Alan Hayward <alan.hayward@arm.com>
16560 David Sherwood <david.sherwood@arm.com>
16562 * reload1.c (spill_stack_slot_width): Change element type
16563 from unsigned int to poly_uint64_pod.
16564 (alter_reg): Treat mode sizes as polynomial.
16566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16567 Alan Hayward <alan.hayward@arm.com>
16568 David Sherwood <david.sherwood@arm.com>
16570 * reload.c (complex_word_subreg_p): New function.
16571 (reload_inner_reg_of_subreg, push_reload): Use it.
16573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16574 Alan Hayward <alan.hayward@arm.com>
16575 David Sherwood <david.sherwood@arm.com>
16577 * lra-constraints.c (process_alt_operands): Reject matched
16578 operands whose sizes aren't ordered.
16579 (match_reload): Refer to this check here.
16581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16582 Alan Hayward <alan.hayward@arm.com>
16583 David Sherwood <david.sherwood@arm.com>
16585 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16586 that the mode size is in the set {1, 2, 4, 8, 16}.
16588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16589 Alan Hayward <alan.hayward@arm.com>
16590 David Sherwood <david.sherwood@arm.com>
16592 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16593 Use plus_constant instead of gen_rtx_PLUS.
16595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16596 Alan Hayward <alan.hayward@arm.com>
16597 David Sherwood <david.sherwood@arm.com>
16599 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16600 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16601 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16602 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16603 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16604 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16605 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16606 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16607 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16608 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16610 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16611 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16612 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16613 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16614 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16615 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16616 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16617 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16618 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16619 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16621 * expr.c (emit_move_resolve_push): Treat the input and result
16622 of PUSH_ROUNDING as a poly_int64.
16623 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16624 (emit_push_insn): Likewise.
16625 * lra-eliminations.c (mark_not_eliminable): Likewise.
16626 * recog.c (push_operand): Likewise.
16627 * reload1.c (elimination_effects): Likewise.
16628 * rtlanal.c (nonzero_bits1): Likewise.
16629 * calls.c (store_one_arg): Likewise. Require the padding to be
16630 known at compile time.
16632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16633 Alan Hayward <alan.hayward@arm.com>
16634 David Sherwood <david.sherwood@arm.com>
16636 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16637 Use plus_constant instead of gen_rtx_PLUS.
16639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16640 Alan Hayward <alan.hayward@arm.com>
16641 David Sherwood <david.sherwood@arm.com>
16643 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16644 rather than an int.
16646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16647 Alan Hayward <alan.hayward@arm.com>
16648 David Sherwood <david.sherwood@arm.com>
16650 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16651 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16652 via stack temporaries. Treat the mode size as polynomial too.
16654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16655 Alan Hayward <alan.hayward@arm.com>
16656 David Sherwood <david.sherwood@arm.com>
16658 * expr.c (expand_expr_real_2): When handling conversions involving
16659 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16660 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16661 as a poly_uint64 too.
16663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16664 Alan Hayward <alan.hayward@arm.com>
16665 David Sherwood <david.sherwood@arm.com>
16667 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16670 Alan Hayward <alan.hayward@arm.com>
16671 David Sherwood <david.sherwood@arm.com>
16673 * combine.c (can_change_dest_mode): Handle polynomial
16674 REGMODE_NATURAL_SIZE.
16675 * expmed.c (store_bit_field_1): Likewise.
16676 * expr.c (store_constructor): Likewise.
16677 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16678 and polynomial REGMODE_NATURAL_SIZE.
16679 (gen_lowpart_common): Likewise.
16680 * reginfo.c (record_subregs_of_mode): Likewise.
16681 * rtlanal.c (read_modify_subreg_p): Likewise.
16683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16684 Alan Hayward <alan.hayward@arm.com>
16685 David Sherwood <david.sherwood@arm.com>
16687 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16688 numbers of elements.
16690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16691 Alan Hayward <alan.hayward@arm.com>
16692 David Sherwood <david.sherwood@arm.com>
16694 * match.pd: Cope with polynomial numbers of vector elements.
16696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16697 Alan Hayward <alan.hayward@arm.com>
16698 David Sherwood <david.sherwood@arm.com>
16700 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16701 in a POINTER_PLUS_EXPR.
16703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16704 Alan Hayward <alan.hayward@arm.com>
16705 David Sherwood <david.sherwood@arm.com>
16707 * omp-simd-clone.c (simd_clone_subparts): New function.
16708 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16709 (ipa_simd_modify_function_body): Likewise.
16711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16712 Alan Hayward <alan.hayward@arm.com>
16713 David Sherwood <david.sherwood@arm.com>
16715 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16716 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16717 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16718 (expand_vector_condition, vector_element): Likewise.
16719 (subparts_gt): New function.
16720 (get_compute_type): Use subparts_gt.
16721 (count_type_subparts): Delete.
16722 (expand_vector_operations_1): Use subparts_gt instead of
16723 count_type_subparts.
16725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16726 Alan Hayward <alan.hayward@arm.com>
16727 David Sherwood <david.sherwood@arm.com>
16729 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16730 (vect_compile_time_alias): ...this new function. Do the calculation
16731 on poly_ints rather than trees.
16732 (vect_prune_runtime_alias_test_list): Update call accordingly.
16734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16735 Alan Hayward <alan.hayward@arm.com>
16736 David Sherwood <david.sherwood@arm.com>
16738 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16740 (vect_schedule_slp_instance): Likewise.
16742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16743 Alan Hayward <alan.hayward@arm.com>
16744 David Sherwood <david.sherwood@arm.com>
16746 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16747 constant and extern definitions for variable-length vectors.
16748 (vect_get_constant_vectors): Note that the number of units
16749 is known to be constant.
16751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16752 Alan Hayward <alan.hayward@arm.com>
16753 David Sherwood <david.sherwood@arm.com>
16755 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16756 of units as polynomial. Choose between WIDE and NARROW based
16759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16760 Alan Hayward <alan.hayward@arm.com>
16761 David Sherwood <david.sherwood@arm.com>
16763 * tree-vect-stmts.c (simd_clone_subparts): New function.
16764 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16767 Alan Hayward <alan.hayward@arm.com>
16768 David Sherwood <david.sherwood@arm.com>
16770 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16771 vectors as polynomial. Use build_index_vector for
16772 IFN_GOMP_SIMD_LANE.
16774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16775 Alan Hayward <alan.hayward@arm.com>
16776 David Sherwood <david.sherwood@arm.com>
16778 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16779 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16780 for variable-length vectors.
16781 (vectorizable_mask_load_store): Treat the number of units as
16782 polynomial, asserting that it is constant if the condition has
16783 already been enforced.
16784 (vectorizable_store, vectorizable_load): Likewise.
16786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16787 Alan Hayward <alan.hayward@arm.com>
16788 David Sherwood <david.sherwood@arm.com>
16790 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16791 of units as polynomial. Punt if we can't tell at compile time
16792 which vector contains the final result.
16794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16795 Alan Hayward <alan.hayward@arm.com>
16796 David Sherwood <david.sherwood@arm.com>
16798 * tree-vect-loop.c (vectorizable_induction): Treat the number
16799 of units as polynomial. Punt on SLP inductions. Use an integer
16800 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16801 cast of such a series for variable-length floating-point
16804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16805 Alan Hayward <alan.hayward@arm.com>
16806 David Sherwood <david.sherwood@arm.com>
16808 * tree.h (build_index_vector): Declare.
16809 * tree.c (build_index_vector): New function.
16810 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16811 of units as polynomial, forcibly converting it to a constant if
16812 vectorizable_reduction has already enforced the condition.
16813 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16814 to create a {1,2,3,...} vector.
16815 (vectorizable_reduction): Treat the number of units as polynomial.
16816 Choose vectype_in based on the largest scalar element size rather
16817 than the smallest number of units. Enforce the restrictions
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-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16825 number of units as polynomial.
16827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16828 Alan Hayward <alan.hayward@arm.com>
16829 David Sherwood <david.sherwood@arm.com>
16831 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16832 * target.def (autovectorize_vector_sizes): Return the vector sizes
16833 by pointer, using vector_sizes rather than a bitmask.
16834 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16835 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16836 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16838 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16839 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16840 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16841 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16842 * omp-general.c (omp_max_vf): Likewise.
16843 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16844 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16845 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16846 * tree-vect-slp.c (vect_slp_bb): Likewise.
16847 * doc/tm.texi: Regenerate.
16848 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16850 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16851 the vector size as a poly_uint64 rather than an unsigned int.
16852 (current_vector_size): Change from an unsigned int to a poly_uint64.
16853 (get_vectype_for_scalar_type): Update accordingly.
16854 * tree.h (build_truth_vector_type): Take the size and number of
16855 units as a poly_uint64 rather than an unsigned int.
16856 (build_vector_type): Add a temporary overload that takes
16857 the number of units as a poly_uint64 rather than an unsigned int.
16858 * tree.c (make_vector_type): Likewise.
16859 (build_truth_vector_type): Take the number of units as a poly_uint64
16860 rather than an unsigned int.
16862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16863 Alan Hayward <alan.hayward@arm.com>
16864 David Sherwood <david.sherwood@arm.com>
16866 * target.def (get_mask_mode): Take the number of units and length
16867 as poly_uint64s rather than unsigned ints.
16868 * targhooks.h (default_get_mask_mode): Update accordingly.
16869 * targhooks.c (default_get_mask_mode): Likewise.
16870 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16871 * doc/tm.texi: Regenerate.
16873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16874 Alan Hayward <alan.hayward@arm.com>
16875 David Sherwood <david.sherwood@arm.com>
16877 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16878 * omp-general.c (omp_max_vf): Likewise.
16879 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16880 (expand_omp_simd): Handle polynomial safelen.
16881 * omp-low.c (omplow_simd_context): Add a default constructor.
16882 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16883 (lower_rec_simd_input_clauses): Update accordingly.
16884 (lower_rec_input_clauses): Likewise.
16886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16887 Alan Hayward <alan.hayward@arm.com>
16888 David Sherwood <david.sherwood@arm.com>
16890 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16891 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16892 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16893 (vect_analyze_slp_cost): Likewise.
16894 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16895 (vect_model_load_cost): Likewise.
16897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16898 Alan Hayward <alan.hayward@arm.com>
16899 David Sherwood <david.sherwood@arm.com>
16901 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16902 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16903 from an unsigned int * to a poly_uint64_pod *.
16904 (calculate_unrolling_factor): New function.
16905 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16908 Alan Hayward <alan.hayward@arm.com>
16909 David Sherwood <david.sherwood@arm.com>
16911 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16912 from an unsigned int to a poly_uint64.
16913 (_loop_vec_info::slp_unrolling_factor): Likewise.
16914 (_loop_vec_info::vectorization_factor): Change from an int
16916 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16917 (vect_get_num_vectors): New function.
16918 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16919 (vect_get_num_copies): Use vect_get_num_vectors.
16920 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16921 to an unsigned int *.
16922 (vect_analyze_data_refs): Change min_vf from an int * to a
16924 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16925 than an unsigned HOST_WIDE_INT.
16926 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16927 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16928 to an unsigned int *.
16929 (vect_analyze_data_ref_dependences): Likewise.
16930 (vect_compute_data_ref_alignment): Handle polynomial vf.
16931 (vect_enhance_data_refs_alignment): Likewise.
16932 (vect_prune_runtime_alias_test_list): Likewise.
16933 (vect_shift_permute_load_chain): Likewise.
16934 (vect_supportable_dr_alignment): Likewise.
16935 (dependence_distance_ge_vf): Take the vectorization factor as a
16936 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16937 (vect_analyze_data_refs): Change min_vf from an int * to a
16939 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16940 vfm1 as a poly_uint64 rather than an int. Make the same change
16941 for the returned bound_scalar.
16942 (vect_gen_vector_loop_niters): Handle polynomial vf.
16943 (vect_do_peeling): Likewise. Update call to
16944 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16945 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16947 * tree-vect-loop.c (vect_determine_vectorization_factor)
16948 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16949 (vect_get_known_peeling_cost): Likewise.
16950 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16951 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16952 (vect_transform_loop): Likewise. Use the lowest possible VF when
16953 updating the upper bounds of the loop.
16954 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16955 rather than an int.
16956 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16957 polynomial unroll factors.
16958 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16959 (vect_make_slp_decision): Likewise.
16960 (vect_supported_load_permutation_p): Likewise, and polynomial
16962 (vect_analyze_slp_cost): Handle polynomial vf.
16963 (vect_slp_analyze_node_operations): Likewise.
16964 (vect_slp_analyze_bb_1): Likewise.
16965 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16966 than an unsigned HOST_WIDE_INT.
16967 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16968 (vectorizable_load): Handle polynomial vf.
16969 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16971 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16974 Alan Hayward <alan.hayward@arm.com>
16975 David Sherwood <david.sherwood@arm.com>
16977 * match.pd: Handle bit operations involving three constants
16978 and try to fold one pair.
16980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16982 * tree-vect-loop-manip.c: Include gimple-fold.h.
16983 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16984 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16985 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16986 Add a path that uses a step of VF instead of 1, but disable it
16988 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16989 and niters_no_overflow parameters. Update calls to
16990 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16991 Create a new SSA name if the latter choses to use a ste other
16992 than zero, and return it via niters_vector_mult_vf_var.
16993 * tree-vect-loop.c (vect_transform_loop): Update calls to
16994 vect_do_peeling, vect_gen_vector_loop_niters and
16995 slpeel_make_loop_iterate_ntimes.
16996 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16997 (vect_gen_vector_loop_niters): Update declarations after above changes.
16999 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
17001 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
17002 128-bit round to integer instructions.
17003 (ceil<mode>2): Likewise.
17004 (btrunc<mode>2): Likewise.
17005 (round<mode>2): Likewise.
17007 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17009 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
17010 unaligned VSX load/store on P8/P9.
17011 (expand_block_clear): Allow the use of unaligned VSX
17012 load/store on P8/P9.
17014 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17016 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
17018 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
17019 swap associated with both a load and a store.
17021 2018-01-02 Andrew Waterman <andrew@sifive.com>
17023 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
17024 * config/riscv/riscv.md (clear_cache): Use it.
17026 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
17028 * web.c: Remove out-of-date comment.
17030 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17032 * expr.c (fixup_args_size_notes): Check that any existing
17033 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
17034 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
17035 (emit_single_push_insn): ...here.
17037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17039 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
17040 (const_vector_encoded_nelts): New function.
17041 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
17042 (const_vector_int_elt, const_vector_elt): Declare.
17043 * emit-rtl.c (const_vector_int_elt_1): New function.
17044 (const_vector_elt): Likewise.
17045 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
17046 of CONST_VECTOR_ELT.
17048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17050 * expr.c: Include rtx-vector-builder.h.
17051 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
17052 directly on the tree encoding.
17053 (const_vector_from_tree): Likewise.
17054 * optabs.c: Include rtx-vector-builder.h.
17055 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
17056 sequence of "u" values.
17057 * vec-perm-indices.c: Include rtx-vector-builder.h.
17058 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
17059 directly on the vec_perm_indices encoding.
17061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17063 * doc/rtl.texi (const_vector): Describe new encoding scheme.
17064 * Makefile.in (OBJS): Add rtx-vector-builder.o.
17065 * rtx-vector-builder.h: New file.
17066 * rtx-vector-builder.c: Likewise.
17067 * rtl.h (rtx_def::u2): Add a const_vector field.
17068 (CONST_VECTOR_NPATTERNS): New macro.
17069 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
17070 (CONST_VECTOR_DUPLICATE_P): Likewise.
17071 (CONST_VECTOR_STEPPED_P): Likewise.
17072 (CONST_VECTOR_ENCODED_ELT): Likewise.
17073 (const_vec_duplicate_p): Check for a duplicated vector encoding.
17074 (unwrap_const_vec_duplicate): Likewise.
17075 (const_vec_series_p): Check for a non-duplicated vector encoding.
17076 Say that the function only returns true for integer vectors.
17077 * emit-rtl.c: Include rtx-vector-builder.h.
17078 (gen_const_vec_duplicate_1): Delete.
17079 (gen_const_vector): Call gen_const_vec_duplicate instead of
17080 gen_const_vec_duplicate_1.
17081 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
17082 (gen_const_vec_duplicate): Use rtx_vector_builder.
17083 (gen_const_vec_series): Likewise.
17084 (gen_rtx_CONST_VECTOR): Likewise.
17085 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
17086 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17087 Build a new vector rather than modifying a CONST_VECTOR in-place.
17088 (handle_special_swappables): Update call accordingly.
17089 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
17090 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
17091 Build a new vector rather than modifying a CONST_VECTOR in-place.
17092 (handle_special_swappables): Update call accordingly.
17094 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17096 * simplify-rtx.c (simplify_const_binary_operation): Use
17097 CONST_VECTOR_ELT instead of XVECEXP.
17099 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17101 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
17102 the selector elements to be different from the data elements
17103 if the selector is a VECTOR_CST.
17104 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
17105 ssizetype for the selector.
17107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17109 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
17110 before testing each element individually.
17111 * tree-vect-generic.c (lower_vec_perm): Likewise.
17113 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17115 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
17116 * selftest-run-tests.c (selftest::run_tests): Call it.
17117 * vector-builder.h (vector_builder::operator ==): New function.
17118 (vector_builder::operator !=): Likewise.
17119 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
17120 (vec_perm_indices::all_from_input_p): New function.
17121 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
17122 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
17123 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
17124 instead of reading the VECTOR_CST directly. Detect whether both
17125 vector inputs are the same before constructing the vec_perm_indices,
17126 and update the number of inputs argument accordingly. Use the
17127 utility functions added above. Only construct sel2 if we need to.
17129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17131 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
17132 the broadcast of the low byte.
17133 (expand_mult_highpart): Use an explicit encoding for the permutes.
17134 * optabs-query.c (can_mult_highpart_p): Likewise.
17135 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
17136 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17137 (vectorizable_bswap): Likewise.
17138 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
17139 explicit encoding for the power-of-2 permutes.
17140 (vect_permute_store_chain): Likewise.
17141 (vect_grouped_load_supported): Likewise.
17142 (vect_permute_load_chain): Likewise.
17144 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17146 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
17147 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
17148 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
17149 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
17150 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
17151 (vect_gen_perm_mask_any): Likewise.
17153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17155 * int-vector-builder.h: New file.
17156 * vec-perm-indices.h: Include int-vector-builder.h.
17157 (vec_perm_indices): Redefine as an int_vector_builder.
17158 (auto_vec_perm_indices): Delete.
17159 (vec_perm_builder): Redefine as a stand-alone class.
17160 (vec_perm_indices::vec_perm_indices): New function.
17161 (vec_perm_indices::clamp): Likewise.
17162 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
17163 (vec_perm_indices::new_vector): New function.
17164 (vec_perm_indices::new_expanded_vector): Update for new
17165 vec_perm_indices class.
17166 (vec_perm_indices::rotate_inputs): New function.
17167 (vec_perm_indices::all_in_range_p): Operate directly on the
17168 encoded form, without computing elided elements.
17169 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
17170 encoding. Update for new vec_perm_indices class.
17171 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
17172 the given vec_perm_builder.
17173 (expand_vec_perm_var): Update vec_perm_builder constructor.
17174 (expand_mult_highpart): Use vec_perm_builder instead of
17175 auto_vec_perm_indices.
17176 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
17177 vec_perm_indices instead of auto_vec_perm_indices. Use a single
17178 or double series encoding as appropriate.
17179 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
17180 vec_perm_indices instead of auto_vec_perm_indices.
17181 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17182 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17183 (vect_permute_store_chain): Likewise.
17184 (vect_grouped_load_supported): Likewise.
17185 (vect_permute_load_chain): Likewise.
17186 (vect_shift_permute_load_chain): Likewise.
17187 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17188 (vect_transform_slp_perm_load): Likewise.
17189 (vect_schedule_slp_instance): Likewise.
17190 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17191 (vectorizable_mask_load_store): Likewise.
17192 (vectorizable_bswap): Likewise.
17193 (vectorizable_store): Likewise.
17194 (vectorizable_load): Likewise.
17195 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
17196 vec_perm_indices instead of auto_vec_perm_indices. Use
17197 tree_to_vec_perm_builder to read the vector from a tree.
17198 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
17199 vec_perm_builder instead of a vec_perm_indices.
17200 (have_whole_vector_shift): Use vec_perm_builder and
17201 vec_perm_indices instead of auto_vec_perm_indices. Leave the
17202 truncation to calc_vec_perm_mask_for_shift.
17203 (vect_create_epilog_for_reduction): Likewise.
17204 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
17205 from auto_vec_perm_indices to vec_perm_indices.
17206 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17207 instead of changing individual elements.
17208 (aarch64_vectorize_vec_perm_const): Use new_vector to install
17209 the vector in d.perm.
17210 * config/arm/arm.c (expand_vec_perm_d::perm): Change
17211 from auto_vec_perm_indices to vec_perm_indices.
17212 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
17213 instead of changing individual elements.
17214 (arm_vectorize_vec_perm_const): Use new_vector to install
17215 the vector in d.perm.
17216 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
17217 Update vec_perm_builder constructor.
17218 (rs6000_expand_interleave): Likewise.
17219 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
17220 (rs6000_expand_interleave): Likewise.
17222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17224 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
17225 to qimode could truncate the indices.
17226 * optabs.c (expand_vec_perm_var): Likewise.
17228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17230 * Makefile.in (OBJS): Add vec-perm-indices.o.
17231 * vec-perm-indices.h: New file.
17232 * vec-perm-indices.c: Likewise.
17233 * target.h (vec_perm_indices): Replace with a forward class
17235 (auto_vec_perm_indices): Move to vec-perm-indices.h.
17236 * optabs.h: Include vec-perm-indices.h.
17237 (expand_vec_perm): Delete.
17238 (selector_fits_mode_p, expand_vec_perm_var): Declare.
17239 (expand_vec_perm_const): Declare.
17240 * target.def (vec_perm_const_ok): Replace with...
17241 (vec_perm_const): ...this new hook.
17242 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
17243 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
17244 * doc/tm.texi: Regenerate.
17245 * optabs.def (vec_perm_const): Delete.
17246 * doc/md.texi (vec_perm_const): Likewise.
17247 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
17248 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
17249 expand_vec_perm for constant permutation vectors. Assert that
17250 the mode of variable permutation vectors is the integer equivalent
17251 of the mode that is being permuted.
17252 * optabs-query.h (selector_fits_mode_p): Declare.
17253 * optabs-query.c: Include vec-perm-indices.h.
17254 (selector_fits_mode_p): New function.
17255 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
17256 is defined, instead of checking whether the vec_perm_const_optab
17257 exists. Use targetm.vectorize.vec_perm_const instead of
17258 targetm.vectorize.vec_perm_const_ok. Check whether the indices
17259 fit in the vector mode before using a variable permute.
17260 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
17261 vec_perm_indices instead of an rtx.
17262 (expand_vec_perm): Replace with...
17263 (expand_vec_perm_const): ...this new function. Take the selector
17264 as a vec_perm_indices rather than an rtx. Also take the mode of
17265 the selector. Update call to shift_amt_for_vec_perm_mask.
17266 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
17267 Use vec_perm_indices::new_expanded_vector to expand the original
17268 selector into bytes. Check whether the indices fit in the vector
17269 mode before using a variable permute.
17270 (expand_vec_perm_var): Make global.
17271 (expand_mult_highpart): Use expand_vec_perm_const.
17272 * fold-const.c: Includes vec-perm-indices.h.
17273 * tree-ssa-forwprop.c: Likewise.
17274 * tree-vect-data-refs.c: Likewise.
17275 * tree-vect-generic.c: Likewise.
17276 * tree-vect-loop.c: Likewise.
17277 * tree-vect-slp.c: Likewise.
17278 * tree-vect-stmts.c: Likewise.
17279 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
17281 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
17282 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
17283 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
17284 (aarch64_vectorize_vec_perm_const): ...this new function.
17285 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17286 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17287 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
17288 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
17289 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17290 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17291 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
17293 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
17294 check for NEON modes.
17295 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
17296 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
17297 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
17298 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
17300 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
17301 the old VEC_PERM_CONST conditions.
17302 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
17303 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
17304 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
17305 (ia64_vectorize_vec_perm_const_ok): Merge into...
17306 (ia64_vectorize_vec_perm_const): ...this new function.
17307 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
17308 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
17309 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
17310 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
17311 * config/mips/mips.c (mips_expand_vec_perm_const)
17312 (mips_vectorize_vec_perm_const_ok): Merge into...
17313 (mips_vectorize_vec_perm_const): ...this new function.
17314 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
17315 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
17316 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
17317 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
17318 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
17319 (rs6000_expand_vec_perm_const): Delete.
17320 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
17322 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17323 (altivec_expand_vec_perm_const_le): Take each operand individually.
17324 Operate on constant selectors rather than rtxes.
17325 (altivec_expand_vec_perm_const): Likewise. Update call to
17326 altivec_expand_vec_perm_const_le.
17327 (rs6000_expand_vec_perm_const): Delete.
17328 (rs6000_vectorize_vec_perm_const_ok): Delete.
17329 (rs6000_vectorize_vec_perm_const): New function.
17330 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17331 an element count and rtx array.
17332 (rs6000_expand_extract_even): Update call accordingly.
17333 (rs6000_expand_interleave): Likewise.
17334 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
17335 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
17336 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
17337 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
17338 (rs6000_expand_vec_perm_const): Delete.
17339 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
17340 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17341 (altivec_expand_vec_perm_const_le): Take each operand individually.
17342 Operate on constant selectors rather than rtxes.
17343 (altivec_expand_vec_perm_const): Likewise. Update call to
17344 altivec_expand_vec_perm_const_le.
17345 (rs6000_expand_vec_perm_const): Delete.
17346 (rs6000_vectorize_vec_perm_const_ok): Delete.
17347 (rs6000_vectorize_vec_perm_const): New function. Remove stray
17348 reference to the SPE evmerge intructions.
17349 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
17350 an element count and rtx array.
17351 (rs6000_expand_extract_even): Update call accordingly.
17352 (rs6000_expand_interleave): Likewise.
17353 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
17354 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
17356 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
17358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17360 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
17361 vector mode and that that mode matches the mode of the data
17363 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
17364 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
17365 directly using expand_vec_perm_1 when forcing selectors into
17367 (expand_vec_perm_var): New function, split out from expand_vec_perm.
17369 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17371 * optabs-query.h (can_vec_perm_p): Delete.
17372 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
17373 * optabs-query.c (can_vec_perm_p): Split into...
17374 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
17375 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
17376 particular selector is valid.
17377 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
17378 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
17379 (vect_grouped_load_supported): Likewise.
17380 (vect_shift_permute_load_chain): Likewise.
17381 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
17382 (vect_transform_slp_perm_load): Likewise.
17383 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
17384 (vectorizable_bswap): Likewise.
17385 (vect_gen_perm_mask_checked): Likewise.
17386 * fold-const.c (fold_ternary_loc): Likewise. Don't take
17387 implementations of variable permutation vectors into account
17388 when deciding which selector to use.
17389 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
17390 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
17391 with a false third argument.
17392 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
17393 to test whether the constant selector is valid and can_vec_perm_var_p
17394 to test whether a variable selector is valid.
17396 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17398 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
17399 * optabs-query.c (can_vec_perm_p): Likewise.
17400 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
17401 instead of vec_perm_indices.
17402 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
17403 (vect_gen_perm_mask_checked): Likewise,
17404 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
17405 (vect_gen_perm_mask_checked): Likewise,
17407 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
17409 * optabs-query.h (qimode_for_vec_perm): Declare.
17410 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
17411 (qimode_for_vec_perm): ...this new function.
17412 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
17414 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17416 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
17417 does not have a conditional at the top.
17419 2018-01-02 Richard Biener <rguenther@suse.de>
17421 * ipa-inline.c (big_speedup_p): Fix expression.
17423 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17426 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
17429 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
17433 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
17434 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
17435 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
17436 cond_taken_branch_cost 3->4.
17438 2018-01-01 Jakub Jelinek <jakub@redhat.com>
17440 PR tree-optimization/83581
17441 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
17442 TODO_cleanup_cfg if any changes have been made.
17444 PR middle-end/83608
17445 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
17446 convert_modes if target mode has the right side, but different mode
17449 PR middle-end/83609
17450 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
17451 last argument when extracting from CONCAT. If either from_real or
17452 from_imag is NULL, use expansion through memory. If result is not
17453 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
17454 the parts directly to inner mode, if even that fails, use expansion
17457 PR middle-end/83623
17458 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
17459 check for bswap in mode rather than HImode and use that in expand_unop
17462 Copyright (C) 2018 Free Software Foundation, Inc.
17464 Copying and distribution of this file, with or without modification,
17465 are permitted in any medium without royalty provided the copyright
17466 notice and this notice are preserved.