1 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3 * doc/install.texi (Prerequisites): Update link for MPC.
5 2018-09-01 Michael Matz <matz@suse.de>
7 PR tree-optimization/87074
8 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
9 PHIs for outer-loop uses.
11 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
13 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
14 * doc/invoke.texi (C Dialect Options): Ditto.
16 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
18 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
20 2018-08-31 Richard Biener <rguenther@suse.de>
22 PR tree-optimization/87168
23 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
24 (rpo_elim::eliminate_avail): When OP was not visited it must
27 2018-08-31 David Malcolm <dmalcolm@redhat.com>
29 * tree-vrp.c (copy_value_range): Convert param "from" from
30 "value_range *" to "const value_range *".
31 (range_is_null): Likewise for param "vr".
32 (range_int_cst_p): Likewise.
33 (range_int_cst_singleton_p): Likewise.
34 (symbolic_range_p): Likewise.
35 (value_ranges_intersect_p): Likewise for both params.
36 (value_range_nonnegative_p): Likewise for param "vr".
37 (value_range_constant_singleton): Likewise.
38 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
39 (extract_range_into_wide_ints): Likewise for param "vr".
40 (extract_range_from_multiplicative_op): Likewise for params "vr0"
42 (vrp_can_optimize_bit_op): Likewise.
43 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
45 (extract_range_from_unary_expr): Likewise.
46 (debug_value_range): Likewise for param "vr".
47 (value_range::dump): Add "const" qualifier.
48 (vrp_prop::check_array_ref): Convert local "vr" from
49 "value_range *" to "const value_range *".
50 (vrp_prop::check_mem_ref): Likewise.
51 (vrp_prop::visit_stmt): Likewise for local "old_vr".
52 (vrp_intersect_ranges_1): Likewise for param "vr_1".
53 (vrp_intersect_ranges): Likewise.
54 (simplify_stmt_for_jump_threading): Likewise for local "vr".
55 (vrp_prop::vrp_finalize): Likewise.
56 * tree-vrp.h (value_range::dump): Add "const" qualifier.
57 (vrp_intersect_ranges): Add "const" qualifier to params as above.
58 (extract_range_from_unary_expr): Likewise.
59 (value_range_constant_singleton): Likewise.
60 (symbolic_range_p): Likewise.
61 (copy_value_range): Likewise.
62 (extract_range_from_binary_expr_1): Likewise.
63 (range_int_cst_p): Likewise.
64 (vrp_set_zero_nonzero_bits): Likewise.
65 (range_int_cst_singleton_p): Likewise.
67 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
69 * config/aarch64/arm_neon.h (vabsd_s64): New.
70 (vnegd_s64): Likewise.
72 2018-08-31 Martin Jambor <mjambor@suse.cz>
74 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
76 2018-08-31 Martin Liska <mliska@suse.cz>
78 * ipa-icf.c (sem_item::add_type): Use
79 sem_item::m_type_hash_cache.
80 * ipa-icf.h: Move the cache from sem_item_optimizer
83 2018-08-31 Nathan Sidwell <nathan@acm.org>
85 * doc/extend.texi (Backwards Compatibility): Remove implicit
86 extern C leeway of () being (...).
88 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
90 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
92 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
96 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
98 2018-08-31 Jakub Jelinek <jakub@redhat.com>
101 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
102 gen_int_mode. Formatting fixes.
104 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
106 * target.def (custom_function_descriptors): Improve documentation.
107 * doc/tm.texi.in (Trampolines): Expand discussion of function
108 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
109 beginning of the section.
110 * doc/tm.texi: Regenerated.
112 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
114 * cfg.h (class auto_edge_flag): Spell out the template-id of the
115 base class in the initializer list. This is a workaround for
116 building with older GCC.
117 (class auto_bb_flag): Likewise.
119 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
121 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
122 (altivec_vcmpequ<VI_char>_p): Remove star.
123 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
125 (expand_strncmp_vec_sequence): New function.
126 (emit_final_str_compare_vec): New function.
127 (expand_strn_compare): Add support for vector strncmp.
128 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
129 length specification to bytes.
130 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
131 (vcmpnezb_p): New pattern.
132 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
133 for option -mstring-compare-inline-limit.
135 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
137 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
138 (PTA_SKYLAKE): Add PTA_AES.
139 (PTA_GOLDMONT): Likewise.
141 2018-08-29 Jan Hubicka <jh@suse.cz>
144 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
145 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
147 2018-08-29 Jan Hubicka <jh@suse.cz>
149 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
151 (hash_tree): Do not visit TYPE_STUB_DECL.
152 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
153 stream TYPE_STUB_DECL.
154 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
155 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
156 after free_lang_data.
157 (type_in_anonymous_namespace_p): Likewise.
159 2018-08-29 Jan Hubicka <jh@suse.cz>
161 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
162 comment that it has to be even number.
163 (class sreal): Change m_sig type to int32_t.
164 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
165 int64_t for temporary calculations.
166 (sreal_verify_basics): Drop one bit from minimum and maximum.
168 2018-08-30 Richard Biener <rguenther@suse.de>
170 PR tree-optimization/87147
171 * tree-ssa-sccvn.c (SSA_VISITED): New function.
172 (visit_phi): When the degenerate result is from the backedge and
173 we didn't visit its definition yet drop to VARYING.
174 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
176 2018-08-29 Jan Hubicka <jh@suse.cz>
178 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
180 (hash_tree): Likewise.
182 2018-08-29 Jan Hubicka <jh@suse.cz>
184 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
185 and TYPE_NEXT_REF_TO.
187 2018-08-29 Jan Hubicka <jh@suse.cz>
189 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
190 comment that it has to be even number.
191 (class sreal): Change m_sig type to int32_t.
192 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
193 int64_t for temporary calculations.
194 (sreal_verify_basics): Drop one bit from minimum and maximum.
196 2018-08-30 Tamar Christina <tamar.christina@arm.com>
198 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
200 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
203 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
204 if to_add is negative.
206 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
209 * builtins.c (c_strlen): Improve range checks.
211 2018-08-29 Martin Sebor <msebor@redhat.com>
212 Jeff Law <law@redhat.com>
214 PR tree-optimization/86714
215 PR tree-optimization/86711
216 * builtins.c (c_strlen): Add arguments to call to string_constant.
217 * expr.c (string_constant): Add argument. Detect missing nul
218 terminator and outermost declaration it's missing in.
219 * expr.h (string_constant): Add argument.
220 * fold-const.c (read_from_constant_string): Add arguments to call to
222 (c_getstr): Likewise.
223 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
225 * tree-ssa-strlen.c (get_stridx): Likewise.
227 2018-08-29 Jan Hubicka <jh@suse.cz>
229 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
230 Do not stream DECL_VINDEX.
231 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
232 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
233 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
235 2018-08-29 Richard Biener <rguenther@suse.de>
237 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
238 virtual operands that are not default defs to honor region
240 (rpo_vn_valueize): Remove ineffective code here.
242 2018-08-29 Richard Biener <rguenther@suse.de>
244 PR tree-optimization/87132
245 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
246 when skipping defs reachable over backedges.
248 2018-08-29 Richard Biener <rguenther@suse.de>
250 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
251 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
252 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
253 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
254 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
255 (vn_reference_lookup_3): Remove use of const_parms.
256 (free_rpo_vn): Do not free const_parms.
257 (do_rpo_vn): Do not call init_const_parms.
258 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
259 SSA_NAME_POINTS_TO_READONLY_MEMORY.
260 (call_may_clobber_ref_p_1): Likewise.
262 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
265 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
267 (-ftree-scev-cprop): Document.
269 2018-08-29 Jan Hubicka <jh@suse.cz>
271 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
273 (sreal constructor): Update.
274 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
275 sreal:operator/): Update.
277 2018-08-29 Martin Liska <mliska@suse.cz>
279 * tree-switch-conversion.c (switch_conversion::expand):
280 Strenghten assumption about gswitch statements.
282 2018-08-29 Richard Biener <rguenther@suse.de>
284 PR tree-optimization/87117
285 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
286 re-value-number released SSA VDEFs.
288 2018-08-29 Richard Biener <rguenther@suse.de>
290 PR tree-optimization/87126
291 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
293 2018-08-28 Jim Wilson <jimw@sifive.com>
295 * config/riscv/pic.md: Rewrite.
296 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
298 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
299 (SOFTF, default_load, softload, softstore): New.
301 2018-08-28 Jeff Law <law@redhat.com>
303 * fold-const.c (fold_binary_loc): Remove recently added assert.
305 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
307 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
308 to OP parmeter of generated function.
310 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
312 PR tree-optimization/87009
313 * match.pd: Add boolean optimizations.
315 2018-08-28 Martin Sebor <msebor@redhat.com>
318 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
319 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
320 (pass_walloca::gate): Use it.
321 (alloca_call_type): Same.
322 (pass_walloca::execute): Same.
323 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
325 2018-08-28 David Malcolm <dmalcolm@redhat.com>
327 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
328 GCC_VERSION for usage of "__gcc_dump_printf__" format from
331 2018-08-28 Richard Biener <rguenther@suse.de>
333 PR tree-optimization/87124
334 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
335 constants before looking up avail.
337 2018-08-28 Jakub Jelinek <jakub@redhat.com>
340 * calls.c (maybe_warn_nonstring_arg): Punt early if
341 warn_stringop_overflow is zero. Don't call get_range_strlen
342 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
343 Swap comparison operands to have constants on rhs. Only use
344 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
347 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
349 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
350 use of tree_to_shwi. Remove duplicated test for the size being
351 a whole number of bytes.
353 2018-08-28 Richard Biener <rguenther@suse.de>
355 PR tree-optimization/87117
356 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
357 Handle removed stmt without LHS (GIMPLE_NOP).
359 2018-08-28 Richard Biener <rguenther@suse.de>
361 PR tree-optimization/87117
362 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
363 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
365 2018-08-28 Richard Biener <rguenther@suse.de>
367 PR tree-optimization/87117
368 * tree-ssa-pre.c (compute_avail): Do not make expressions
369 with predicated values available.
370 (get_expr_value_id): Assert we do not run into predicated value
373 2018-08-28 Richard Biener <rguenther@suse.de>
375 PR tree-optimization/87117
376 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
377 get virtual operands.
378 (get_expr_operands): Handle STRING_CST like other decls.
380 2018-08-28 Martin Liska <mliska@suse.cz>
382 * tree.h: Update documentation of fndecl_built_in_p
386 2018-08-27 Jeff Law <law@redhat.com>
387 PR tree-optimization/87110
388 * tree-ssa-dse.c (compute_trims): Handle non-constant
391 2018-08-27 Martin Sebor <msebor@redhat.com>
393 PR tree-optimization/86914
394 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
396 2018-08-27 Martin Sebor <msebor@redhat.com>
398 PR tree-optimization/87112
399 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
400 the type of the bound argument.
402 2018-08-27 Jeff Law <law@redhat.com>
404 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
405 type does not have a TYPE_SIZE_UNIT.
407 2018-08-27 Steve Ellcey <sellcey@cavium.com>
409 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
410 with include of backend.h.
412 2018-08-27 Richard Biener <rguenther@suse.de>
414 PR tree-optimization/86927
415 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
416 use const cond reduction code.
418 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
420 PR tree-optimization/85758
421 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
423 2018-08-27 David Malcolm <dmalcolm@redhat.com>
426 * diagnostic-show-locus.c (class layout_range): Update for
427 conversion of show_caret_p to a tri-state.
428 (layout_range::layout_range): Likewise.
429 (make_range): Likewise.
430 (layout::maybe_add_location_range): Likewise.
431 (layout::should_print_annotation_line_p): Don't show annotation
432 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
433 (layout::get_state_at_point): Update for conversion of
434 show_caret_p to a tri-state. Bail out early for
435 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
436 underlining or source colorization.
437 (gcc_rich_location::add_location_if_nearby): Update for conversion
438 of show_caret_p to a tri-state.
439 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
440 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
442 (selftest::test_one_liner_labels): Likewise.
443 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
444 conversion of show_caret_p to a tri-state.
445 * pretty-print.c (text_info::set_location): Likewise.
446 * pretty-print.h (text_info::set_location): Likewise.
447 * substring-locations.c (format_warning_n_va): Likewise.
448 * tree-diagnostic.c (default_tree_printer): Likewise.
449 * tree-pretty-print.c (newline_and_indent): Likewise.
451 2018-08-27 David Malcolm <dmalcolm@redhat.com>
454 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
455 line above for line-insertion fix-it hints.
456 (selftest::test_fixit_insert_containing_newline): Update the
457 expected results, and add a test with line-numbering enabled.
459 2018-08-27 Martin Liska <mliska@suse.cz>
462 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
463 params with DECL_HAS_VALUE_EXPR_P.
465 2018-08-27 Martin Liska <mliska@suse.cz>
467 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
468 selected expansion strategy.
470 2018-08-27 Martin Liska <mliska@suse.cz>
472 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
473 * builtins.c (is_builtin_fn): Likewise.
474 * attribs.c (diag_attr_exclusions): Use new function
475 fndecl_built_in_p and remove check for FUNCTION_DECL if
477 (builtin_mathfn_code): Likewise.
478 (fold_builtin_expect): Likewise.
479 (fold_call_expr): Likewise.
480 (fold_builtin_call_array): Likewise.
481 (fold_call_stmt): Likewise.
482 (set_builtin_user_assembler_name): Likewise.
483 (is_simple_builtin): Likewise.
484 * calls.c (gimple_alloca_call_p): Likewise.
485 (maybe_warn_nonstring_arg): Likewise.
486 * cfgexpand.c (expand_call_stmt): Likewise.
487 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
488 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
489 (cgraph_node::verify_node): Likewise.
490 * cgraphclones.c (build_function_decl_skip_args): Likewise.
491 (cgraph_node::create_clone): Likewise.
492 * config/arm/arm.c (arm_insert_attributes): Likewise.
493 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
494 * dse.c (scan_insn): Likewise.
495 * expr.c (expand_expr_real_1): Likewise.
496 * fold-const.c (operand_equal_p): Likewise.
497 (fold_binary_loc): Likewise.
498 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
499 * gimple-low.c (lower_stmt): Likewise.
500 * gimple-pretty-print.c (dump_gimple_call): Likewise.
501 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
502 * gimple.c (gimple_build_call_from_tree): Likewise.
503 (gimple_call_builtin_p): Likewise.
504 (gimple_call_combined_fn): Likewise.
505 * gimplify.c (gimplify_call_expr): Likewise.
506 (gimple_boolify): Likewise.
507 (gimplify_modify_expr): Likewise.
508 (gimplify_addr_expr): Likewise.
509 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
510 * ipa-cp.c (determine_versionability): Likewise.
511 * ipa-fnsummary.c (compute_fn_summary): Likewise.
512 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
513 * ipa-split.c (visit_bb): Likewise.
514 (split_function): Likewise.
515 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
516 * lto-cgraph.c (input_node): Likewise.
517 * lto-streamer-out.c (write_symbol): Likewise.
518 * omp-low.c (setjmp_or_longjmp_p): Likewise.
519 (lower_omp_1): Likewise.
520 * predict.c (strip_predict_hints): Likewise.
521 * print-tree.c (print_node): Likewise.
522 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
523 * trans-mem.c (is_tm_irrevocable): Likewise.
524 (is_tm_load): Likewise.
525 (is_tm_simple_load): Likewise.
526 (is_tm_store): Likewise.
527 (is_tm_simple_store): Likewise.
528 (is_tm_abort): Likewise.
529 (tm_region_init_1): Likewise.
530 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
531 * tree-cfg.c (verify_gimple_call): Likewise.
532 (move_stmt_r): Likewise.
533 (stmt_can_terminate_bb_p): Likewise.
534 * tree-eh.c (lower_eh_constructs_2): Likewise.
535 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
536 * tree-inline.c (remap_gimple_stmt): Likewise.
538 (estimate_num_insns): Likewise.
539 (fold_marked_statements): Likewise.
540 * tree-sra.c (scan_function): Likewise.
541 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
542 (optimize_stack_restore): Likewise.
543 (pass_fold_builtins::execute): Likewise.
544 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
545 (mark_all_reaching_defs_necessary_1): Likewise.
546 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
547 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
548 (pass_forwprop::execute): Likewise.
549 * tree-ssa-loop-im.c (stmt_cost): Likewise.
550 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
551 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
552 * tree-ssa-strlen.c (get_string_length): Likewise.
553 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
554 (find_func_aliases_for_call): Likewise.
555 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
556 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
557 * tree-tailcall.c (find_tail_calls): Likewise.
558 * tree.c (need_assembler_name_p): Likewise.
559 (free_lang_data_in_decl): Likewise.
560 (get_call_combined_fn): Likewise.
561 * ubsan.c (is_ubsan_builtin_p): Likewise.
562 * varasm.c (incorporeal_function_p): Likewise.
563 * tree.h (DECL_BUILT_IN): Remove and replace with
565 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
566 (fndecl_built_in_p): New.
568 2018-08-27 Martin Liska <mliska@suse.cz>
570 PR tree-optimization/86847
571 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
572 Dump also subtree probability.
573 (switch_decision_tree::do_jump_if_equal): New function.
574 (switch_decision_tree::emit_case_nodes): Handle special
575 situations in balanced tree that can be emitted much simpler.
576 Fix calculation of probabilities that happen in tree expansion.
577 * tree-switch-conversion.h (struct cluster): Add
579 (struct simple_cluster): Likewise.
580 (struct case_tree_node): Add new function has_child.
581 (do_jump_if_equal): New.
583 2018-08-27 Martin Liska <mliska@suse.cz>
585 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
586 Add new argument to bit_test_cluster constructor.
587 (bit_test_cluster::emit): Set bits really number of values
589 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
590 probability argument.
591 * tree-switch-conversion.h (struct bit_test_cluster):
592 Add m_handles_entire_switch.
594 2018-08-27 Martin Liska <mliska@suse.cz>
596 PR tree-optimization/86702
597 * tree-switch-conversion.c (jump_table_cluster::emit):
598 Make probabilities even for values in jump table
599 according to number of cases handled.
600 (switch_decision_tree::compute_cases_per_edge): Pass
601 argument to reset_out_edges_aux function.
602 (switch_decision_tree::analyze_switch_statement): Likewise.
603 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
606 2018-08-27 Martin Liska <mliska@suse.cz>
608 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
609 cfun argument explicitly.
610 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
611 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
612 function gimple_switch_default_bb.
613 (convert_switch_statements):
615 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
616 * stmt.c (label_to_block_fn): Use label_to_block and pass
617 cfun argument explicitly and use gimple_switch_label_bb.
618 (expand_case): Likewise.
619 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
620 cfun argument explicitly. Likewise.
621 (make_edges_bb): Likewise.
622 (make_cond_expr_edges): Likewise.
623 (get_cases_for_edge): Likewise.
624 (make_gimple_switch_edges): Likewise.
625 (label_to_block_fn): Likewise.
626 (label_to_block): Likewise.
627 (make_goto_expr_edges): Likewise.
628 (make_gimple_asm_edges): Likewise.
629 (main_block_label): Likewise.
630 (group_case_labels_stmt): Likewise.
631 (find_taken_edge_computed_goto): Likewise.
632 (find_taken_edge_switch_expr): Likewise.
633 (gimple_verify_flow_info): Likewise.
634 (gimple_redirect_edge_and_branch): Likewise.
635 (gimple_switch_label_bb): New function.
636 (gimple_switch_default_bb): Likewise.
637 (gimple_switch_edge): Likewise.
638 (gimple_switch_default_edge): Likewise.
639 * tree-cfg.h (label_to_block_fn): Remove and replace ...
640 (label_to_block): ... with this.
641 (gimple_switch_label_bb): New.
642 (gimple_switch_default_bb): Likewise.
643 (gimple_switch_edge): Likewise.
644 (gimple_switch_default_edge): Likewise.
645 * tree-cfgcleanup.c (convert_single_case_switch): Use
646 new gimple functions and pass new argument to label_to_block.
647 (cleanup_control_flow_bb):
648 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
649 cfun argument explicitly.
650 (make_eh_edges): Likewise.
651 (redirect_eh_dispatch_edge): Likewise.
652 (lower_resx): Likewise.
653 (lower_eh_dispatch): Likewise.
654 (maybe_remove_unreachable_handlers): Likewise.
655 (unsplit_eh): Likewise.
656 (cleanup_empty_eh): Likewise.
657 (verify_eh_edges): Likewise.
658 (verify_eh_dispatch_edge): Likewise.
659 * tree-ssa-dom.c (record_edge_info): Likewise.
660 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
661 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
662 (thread_through_normal_block): Likewise.
663 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
664 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
665 * tree-switch-conversion.c (switch_conversion::collect): Use new
667 (switch_conversion::check_final_bb): Likewise.
668 (switch_conversion::gather_default_values): Pass new argument
670 (switch_conversion::build_constructors): Likewise.
671 (switch_decision_tree::compute_cases_per_edge): Use new
672 gimple_switch_edge function.
673 (switch_decision_tree::analyze_switch_statement): Pass new argument
675 (switch_decision_tree::try_switch_expansion): Use
676 gimple_switch_default_edge.
677 * tree-vrp.c (find_switch_asserts): Pass new argument
679 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
680 (vr_values::simplify_switch_using_ranges): Likewise.
682 2018-08-27 Richard Biener <rguenther@suse.de>
684 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
685 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
687 * tree-ssa-sccvn.h (struct vn_pval): New structure.
688 (struct vn_nary_op_s): Add unwind_to member. Add
689 predicated_values flag and put result into a union together
690 with a linked list of vn_pval.
691 (struct vn_ssa_aux): Add name member to make maintaining
692 a map of SSA name to vn_ssa_aux possible. Remove no longer
693 needed info, dfsnum, low, visited, on_sccstack, use_processed
694 and range_info_anti_range_p members.
695 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
696 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
698 (vn_valueize): New global.
699 (vn_context_bb): Likewise.
700 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
701 VN_INFO_PTR_INFO): Remove.
702 * tree-ssa-sccvn.c: ... (rewrite)
703 (pass_fre::execute): For -O2+ initialize loops and run
704 RPO VN in optimistic mode (iterating). For -O1 and -Og
705 run RPO VN in non-optimistic mode.
706 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
707 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
708 * doc/invoke.texi (sccvn-max-scc-size): Remove.
709 (rpo-vn-max-loop-depth): Document.
710 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
711 when valuezing the VUSE signals we walked out of the region.
712 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
713 (phi_translate): Set VN context block to use for availability
715 (compute_avail): Likewise.
716 (pre_valueize): New function.
717 (pass_pre::execute): Adjust to the RPO VN API.
719 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
720 (propagate_constants_for_unrolling): Remove.
721 (tree_unroll_loops_completely): Perform value-numbering
722 on the unrolled bodies loop parent.
724 2018-08-27 Richard Biener <rguenther@suse.de>
726 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
727 for partial antic compute.
729 2018-08-27 Jakub Jelinek <jakub@redhat.com>
731 PR rtl-optimization/87065
732 * combine.c (simplify_if_then_else): Formatting fix.
733 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
735 (known_cond): Don't return const_true_rtx for vector modes. Use
736 CONST0_RTX instead of const0_rtx. Formatting fixes.
738 2018-08-27 Martin Liska <mliska@suse.cz>
740 PR gcov-profile/87069
741 * gcov.c (process_file): Record files already processed
742 and warn about a file being processed multiple times.
744 2018-08-27 Martin Liska <mliska@suse.cz>
747 * config/aarch64/aarch64.c (aarch64_override_options_internal):
748 Set default values for x_aarch64_*_string strings.
749 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
750 prefix. For -mabi do not print '=ABI' in help and use
751 <option_value> format for -msve-vector-bits and -moverride
754 2018-08-26 Jeff Law <law@redhat.com>
756 * config/mips/frame-header-opt.c: Include "backend.h" rather than
759 2018-08-26 Marek Polacek <polacek@redhat.com>
761 PR c++/87029, Implement -Wredundant-move.
762 * doc/invoke.texi: Document -Wredundant-move.
764 2018-08-25 Martin Sebor <msebor@redhat.com>
766 PR tree-optimization/87059
767 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
768 to the same type as the other.
769 * fold-const.c (fold_binary_loc): Assert expectation.
771 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
773 * config/darwin.c (machopic_legitimize_pic_address): Clean up
774 extraneous parentheses, dead code section and formatting.
776 2018-08-24 David Malcolm <dmalcolm@redhat.com>
779 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
780 wide enough for jumps in the line-numbering to be visible.
781 (layout::print_gap_in_line_numbering): New member function.
782 (layout::calculate_line_spans): When using line numbering, merge
783 line spans that are only 1 line apart.
784 (diagnostic_show_locus): When printing line numbers, show gaps in
785 line numbering directly, rather than printing headers.
786 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
787 line-numbering with multiple line spans.
788 (selftest::test_fixit_insert_containing_newline_2): Add test of
789 line-numbering, in which the spans are close enough to be merged.
791 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
793 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
794 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
795 * tree-vrp.c (range_is_nonnull): Remove.
796 (range_includes_zero_p): Accept value_range instead of min/max.
797 (extract_range_from_binary_expr_1): Do not early bail on
799 Use range_includes_zero_p instead of range_is_nonnull.
800 (extract_range_from_unary_expr): Use range_includes_zero_p instead
802 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
803 special case VR_ANTI_RANGE.
804 (vrp_finalize): Same.
805 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
807 (range_is_nonnull): Remove.
808 * vr-values.c (vrp_stmt_computes_nonzero): Use
809 range_includes_zero_p instead of range_is_nonnull.
810 (extract_range_basic): Pass value_range to range_includes_zero_p
811 instead of range_is_nonnull.
813 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
815 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
816 * emit-rtl.h (rtl_data): Remove return_bnd.
817 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
818 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
819 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
820 (POINTER_BOUNDS_MODE): Remove definition.
821 (make_pointer_bounds_mode): Remove.
822 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
823 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
824 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
825 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
826 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
827 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
828 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
830 * config/i386/i386-modes.def (BND32, BND64): Remove.
831 * config/i386/i386.c (dbx_register_map): Remove bound registers.
832 (dbx64_register_map): Ditto.
833 (svr4_dbx_register_map): Ditto.
834 (indirect_thunk_bnd_needed): Remove.
835 (indirect_thunks_bnd_used): Ditto.
836 (indirect_return_bnd_needed): Ditto.
837 (indirect_return_via_cx_bnd): Ditto.
838 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
839 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
840 (output_indirect_thunk): Ditto. Remove need_prefix argument.
841 (output_indirect_thunk_function): Remove handling of
842 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
843 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
844 (ix86_save_reg): Remove handling of crtl->return_bnd.
845 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
846 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
847 and UNSPEC_BNDLX_ADDR.
848 (ix86_output_indirect_branch_via_reg): Remove handling of
849 indirect_thunk_prefix_bnd.
850 (ix86_output_indirect_branch_via_push): Ditto.
851 (ix86_output_function_return): Ditto.
852 (ix86_output_indirect_function_return): Ditto.
853 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
854 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
855 (CALL_USED_REGISTERS): Ditto.
856 (REG_ALLOC_ORDER): Update for removal of bound registers.
857 (HI_REGISTER_NAMES): Ditto.
858 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
859 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
860 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
861 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
862 (FIRST_PSEUDO_REG): Update.
863 (BND): Remove mode iterator.
864 * config/i386/predicates.md (bnd_mem_operator): Remove.
866 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
868 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
871 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
873 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
874 the case in which the permute needs only a single element and
875 repeats for every vector of the result. Extend that case to
876 handle variable-length vectors.
877 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
879 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
882 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
883 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
885 2018-08-24 Richard Biener <rguenther@suse.de>
887 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
888 bb_flags_allocated members.
889 (auto_flag): New RAII class for allocating flags.
890 (auto_edge_flag): New RAII class for allocating edge flags.
891 (auto_bb_flag): New RAII class for allocating bb flags.
892 * cfgloop.c (verify_loop_structure): Allocate temporary edge
894 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
895 in favor of temporarily allocated BB flag.
896 * hsa-brig.c: Re-order includes.
897 * hsa-dump.c: Likewise.
898 * hsa-regalloc.c: Likewise.
899 * print-rtl.c: Likewise.
900 * profile-count.c: Likewise.
902 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
905 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
908 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
911 * wide-int-range.cc (wide_int_range_div): Do not ignore result
912 from wide_int_range_multiplicative_op.
914 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
916 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
919 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
921 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
924 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
926 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
927 full GENERIC expression used for replacement.
929 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
931 * tree-vrp.c (abs_extent_range): Remove.
932 (extract_range_into_wide_ints): Pass wide ints by reference.
933 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
934 Pass wide ints by reference in all calls to
935 extract_range_into_wide_ints.
936 * wide-int-range.cc (wide_int_range_div): New.
937 * wide-int-range.h (wide_int_range_div): New.
938 (wide_int_range_includes_zero_p): New.
939 (wide_int_range_zero_p): New.
941 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
943 * config/aarch64/aarch64.md (arches): New enum.
944 (arch): New enum attr.
945 (arch_enabled): New attr.
946 (enabled): Now uses arch_enabled only.
947 (simd, sve, fp16): Removed attribute.
948 (fp): Attr now defined in terms of 'arch'.
949 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
950 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
951 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
952 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
953 attributes into 'arch'.
954 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
955 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
956 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
957 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
958 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
959 'simd' attribute into 'arch'.
960 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
961 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
962 Convert use of 'fp' attribute to 'arch'.
963 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
964 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
966 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
967 (different modes) Merge 'fp' and 'simd' into 'arch'.
968 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
971 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
973 PR rtl-optimization/87026
974 * expmed.c (canonicalize_comparison): If we can no longer create
975 pseudoregisters, don't.
977 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
980 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
982 * config/arm/arm.c (speculation_barrier_libfunc): New static
984 (arm_init_libfuncs): Initialize it.
985 (arm_emit_speculation_barrier): New function.
986 * config/arm/arm.md (speculation_barrier): Call
987 arm_emit_speculation_barrier for architectures that do not have
989 (speculation_barrier_insn): Only match on Armv7 or later.
991 2018-08-23 Richard Biener <rguenther@suse.de>
994 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
997 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
999 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1000 of single-vector TBLs.
1001 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
1004 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1007 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
1008 aarch64_evpc_tbl guard.
1010 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
1012 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
1015 2018-08-22 Martin Sebor <msebor@redhat.com>
1018 * tree-pretty-print.c (pretty_print_string): Add argument.
1019 (dump_generic_node): Call to pretty_print_string with string size.
1021 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
1023 PR rtl-optimization/86771
1024 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
1025 of two SETs into those two SETs, one to be placed at i2, if that SETs
1026 destination is modified between i2 and i3.
1028 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1030 PR tree-optimization/86725
1031 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
1033 (vect_analyze_scalar_cycles_1): Check it.
1035 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1037 PR tree-optimization/86725
1038 * tree-vect-loop.c (vect_is_simple_reduction): When treating
1039 an outer loop phi as a double reduction, make sure that the
1040 single user of the phi result is an inner loop phi.
1042 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1044 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
1045 grouped stores with gaps to a strided group.
1047 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
1049 * tree-vect-stmts.c (get_group_load_store_type)
1050 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
1051 first statement in a group.
1053 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1055 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
1056 the sequence used in gcc/gcc.c.
1058 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1061 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
1064 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1066 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
1067 Darwin10-specific unwinder-shim.
1068 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
1069 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
1070 New to cater for Darwin10 Rosetta.
1072 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1074 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
1075 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
1077 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
1082 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
1083 (dwarf2out_switch_text_section): Generate a local label for the
1084 second function sub-section and apply it as the second FDE start
1086 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
1087 second sub-section start.
1089 2018-08-22 Richard Biener <rguenther@suse.de>
1091 PR tree-optimization/86988
1092 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
1094 2018-08-22 Richard Biener <rguenther@suse.de>
1096 PR tree-optimization/86945
1097 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
1099 2018-08-22 Alexandre Oliva <oliva@adacore.com>
1101 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
1102 a comment about how uses of r2 for .sdata2 come about.
1104 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
1106 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
1108 2018-08-21 Marek Polacek <polacek@redhat.com>
1110 PR c++/86981, Implement -Wpessimizing-move.
1111 * doc/invoke.texi: Document -Wpessimizing-move.
1113 2018-08-21 Jan Hubicka <jh@suse.cz>
1115 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
1116 * tree.h (is_redundant_typedef): Remove.
1117 * dwarf2out.c (is_redundant_typedef): Turn into static function.
1119 2018-08-21 Jan Hubicka <jh@suse.cz>
1121 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
1124 2018-08-21 Tamar Christina <tamar.christina@arm.com>
1126 * expmed.c (extract_low_bits): Reject invalid subregs early.
1128 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1131 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
1134 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1136 * config/vxworks.h: Guard vxworks_asm_out_constructor and
1137 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
1138 * config/vxworks.c: Likewise.
1140 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1142 * config/vxworks.c: Set targetm.have_ctors_dtors
1143 if HAVE_INITFINI_ARRAY_SUPPORT.
1144 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
1145 if HAVE_INITFINI_ARRAY_SUPPORT.
1147 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
1149 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
1150 default search path for VxWorks < 7.
1152 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
1154 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
1155 (get_format_string): Refer to c_getstr.
1157 2018-08-21 Tom de Vries <tdevries@suse.de>
1159 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1160 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1161 (debuginfo_early_stop): Declare.
1162 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
1163 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
1164 (debuginfo_early_stop): New function.
1165 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
1166 and debuginfo_early_stop.
1167 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
1168 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
1169 (general_init): Call debuginfo_early_init.
1170 (finalize): Call debuginfo_fini.
1171 (do_compile): Call debuginfo_init.
1172 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
1174 (@item -fdump-debug, @item -fdump-earlydebug): Add.
1176 2018-08-21 Tom de Vries <tdevries@suse.de>
1178 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
1179 flag_dump_noaddr and flag_dump_unnumbered.
1181 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
1183 * wide-int-range.cc (wide_int_range_abs): New.
1184 (wide_int_range_order_set): Rename from wide_int_range_min_max.
1185 * wide-int-range.h (wide_int_range_abs): New.
1186 (wide_int_range_min_max): New.
1187 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
1188 case to call wide_int_range_abs.
1189 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
1190 (extract_range_from_abs_expr): Delete.
1192 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
1195 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
1196 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
1199 2018-08-20 Nathan Sidwell <nathan@acm.org>
1200 Jeff Law <law@redhat.com>
1202 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
1203 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
1205 2018-08-20 David Malcolm <dmalcolm@redhat.com>
1208 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
1209 (decl_attributes): Likewise.
1210 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
1212 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1213 * diagnostic-core.h (class auto_diagnostic_group): New class.
1214 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
1215 (diagnostic_report_diagnostic): Handle the first diagnostics within
1217 (emit_diagnostic): Add auto_diagnostic_group instance.
1219 (inform_n): Likewise.
1220 (warning): Likewise.
1221 (warning_at): Likewise.
1222 (warning_n): Likewise.
1223 (pedwarn): Likewise.
1224 (permerror): Likewise.
1226 (error_n): Likewise.
1227 (error_at): Likewise.
1229 (fatal_error): Likewise.
1230 (internal_error): Likewise.
1231 (internal_error_no_backtrace): Likewise.
1232 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
1233 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
1234 * diagnostic.h (struct diagnostic_context): Add fields
1235 "diagnostic_group_nesting_depth",
1236 "diagnostic_group_emission_count", "begin_group_cb",
1238 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
1239 Add auto_diagnostic_group instance(s).
1240 (find_explicit_erroneous_behavior): Likewise.
1241 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
1242 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
1243 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
1244 (gimplify_va_arg_expr): Likewise.
1245 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
1246 (HSA_SORRY_AT): Likewise.
1247 * ipa-devirt.c (compare_virtual_tables): Likewise.
1248 (warn_odr): Likewise.
1249 * multiple_target.c (expand_target_clones): Likewise.
1250 * opts-common.c (cmdline_handle_error): Likewise.
1251 * reginfo.c (globalize_reg): Likewise.
1252 * substring-locations.c (format_warning_n_va): Likewise.
1253 * tree-inline.c (expand_call_inline): Likewise.
1254 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
1255 * tree-ssa-loop-niter.c
1256 (do_warn_aggressive_loop_optimizations): Likewise.
1257 * tree-ssa-uninit.c (warn_uninit): Likewise.
1258 * tree.c (warn_deprecated_use): Likewise.
1260 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
1263 * config/i386/i386.md (eh_return): Always update EH return
1264 address in word_mode.
1266 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
1268 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
1269 TARGET_SPLIT_COMPLEX_ARG is defined.
1271 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1273 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
1275 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
1278 * expr.c (expand_assignment): Assert that bitpos is positive.
1279 (store_field): Likewise
1280 (expand_expr_real_1): Make sure that bitpos is positive.
1281 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
1284 2018-08-20 Nathan Sidwell <nathan@acm.org>
1286 * Makefile.in (CPP_ID_DATA_H): Delete.
1287 (CPP_INTERNAL_H): Don't add it.
1288 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
1289 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
1291 2018-08-20 Richard Biener <rguenther@suse.de>
1293 PR tree-optimization/78655
1294 * tree-vrp.c (extract_range_from_binary_expr_1): Make
1295 pointer + offset nonnull if either operand is nonnull work.
1297 2018-08-20 Tom de Vries <tdevries@suse.de>
1299 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
1300 unless the referenced die describes the added property using
1301 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
1302 Otherwise, add a DW_AT_location to the referenced die.
1304 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
1307 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
1308 register_operand when calling ix86_set_reg_reg_cost.
1309 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
1310 Set *total to 0 for operands that satisfy x86_64_immediate_operand
1311 predicate and to 1 otherwise.
1313 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
1315 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
1316 emit a diagnostic that it is not supported and reset the option.
1317 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
1318 supported and consume the option. (ASM_FINAL_SPEC): New.
1320 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1322 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
1325 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
1327 C-SKY port: Documentation
1329 * doc/extend.texi (C-SKY Function Attributes): New section.
1330 * doc/invoke.texi (Option Summary): Add C-SKY options.
1331 (C-SKY Options): New section.
1332 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
1334 2018-08-17 Jojo <jijie_rong@c-sky.com>
1335 Huibin Wang <huibin_wang@c-sky.com>
1336 Sandra Loosemore <sandra@codesourcery.com>
1337 Chung-Lin Tang <cltang@codesourcery.com>
1339 C-SKY port: Backend implementation
1341 * config/csky/*: New.
1342 * common/config/csky/*: New.
1344 2018-08-17 Jojo <jijie_rong@c-sky.com>
1345 Huibin Wang <huibin_wang@c-sky.com>
1346 Sandra Loosemore <sandra@codesourcery.com>
1347 Chung-Lin Tang <cltang@codesourcery.com>
1348 Andrew Jenner <andrew@codesourcery.com>
1350 C-SKY port: Configury
1352 * config.gcc (csky-*-*): New.
1353 * configure.ac: Add csky to targets for dwarf2 debug_line support.
1354 * configure: Regenerated.
1356 2018-08-17 David Malcolm <dmalcolm@redhat.com>
1358 * dump-context.h: Include "dumpfile.h".
1359 (dump_context::dump_printf_va): Convert final param from va_list
1360 to va_list *. Convert from ATTRIBUTE_PRINTF to
1361 ATTRIBUTE_GCC_DUMP_PRINTF.
1362 (dump_context::dump_printf_loc_va): Likewise.
1363 * dumpfile.c: Include "stringpool.h".
1364 (make_item_for_dump_printf_va): Delete.
1365 (make_item_for_dump_printf): Delete.
1366 (class dump_pretty_printer): New class.
1367 (dump_pretty_printer::dump_pretty_printer): New ctor.
1368 (dump_pretty_printer::emit_items): New member function.
1369 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
1371 (dump_pretty_printer::emit_item): New member function.
1372 (dump_pretty_printer::stash_item): New member function.
1373 (dump_pretty_printer::format_decoder_cb): New member function.
1374 (dump_pretty_printer::decode_format): New member function.
1375 (dump_context::dump_printf_va): Reimplement in terms of
1376 dump_pretty_printer.
1377 (dump_context::dump_printf_loc_va): Convert final param from va_list
1379 (dump_context::begin_scope): Reimplement call to
1380 make_item_for_dump_printf.
1381 (dump_printf): Update for change to dump_printf_va.
1382 (dump_printf_loc): Likewise.
1383 (selftest::test_capture_of_dump_calls): Convert "stmt" from
1384 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
1385 with %T, %E, and %G.
1386 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
1387 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
1388 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
1389 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
1390 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
1391 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
1392 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
1393 within a dump_printf_loc call to "%wu".
1394 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
1395 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
1396 missing space after "=".
1397 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
1398 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
1399 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
1400 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
1401 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
1402 duplicate "vectorized" from message.
1404 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
1406 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
1407 polyNxK_t element's TYPE_STRING_FLAG.
1409 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
1411 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
1412 (they were unnamed before). Fix comments.
1414 2018-08-17 Nathan Sidwell <nathan@acm.org>
1416 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
1418 2018-08-17 Richard Biener <rguenther@suse.de>
1420 PR tree-optimization/86841
1421 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
1423 2018-08-17 Martin Liska <mliska@suse.cz>
1425 * common.opt: Remove Warn, Init and Report for options with
1426 Ignore/Deprecated flag. Warning is done automatically for
1428 * config/i386/i386.opt: Likewise.
1429 * config/ia64/ia64.opt: Likewise.
1430 * config/rs6000/rs6000.opt: Likewise.
1431 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1432 Remove usage of flag_check_pointer_bounds.
1433 * lto-wrapper.c (merge_and_complain): Do not handle
1434 OPT_fcheck_pointer_bounds.
1435 (append_compiler_options): Likewise.
1436 * opt-functions.awk: Do not handle Deprecated.
1437 * optc-gen.awk: Check that Var, Report and Init are not
1438 used for an option with Ignore/Deprecated flag.
1439 * opts-common.c (decode_cmdline_option): Do not report
1441 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
1443 * opts.h (struct cl_option): Remove cl_deprecated flag.
1444 (CL_ERR_DEPRECATED): Remove error enum value.
1446 2018-08-17 Richard Biener <rguenther@suse.de>
1449 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
1450 across a va-arg-pack using call adjust its return value accordingly.
1452 2018-08-16 Martin Sebor <msebor@redhat.com>
1454 PR tree-optimization/86853
1455 * gimple-ssa-sprintf.c (struct format_result): Rename member.
1456 (struct fmtresult): Add member and initialize it in ctors.
1457 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
1458 (format_string): Handle %S the same as %ls. Set MAYFAIL.
1459 (format_directive): Set POSUNDER4K when MAYFAIL is set.
1460 (parse_directive): Handle %C same as %c.
1461 (sprintf_dom_walker::compute_format_length): Adjust.
1462 (is_call_safe): Adjust.
1464 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1466 * builtins.c (c_strlen): Add new parameter eltsize. Use it
1467 for determining how to count the elements.
1468 * builtins.h (c_strlen): Adjust prototype.
1469 * expr.c (string_constant): Add new parameter mem_size.
1470 Set *mem_size appropriately.
1471 * expr.h (string_constant): Adjust protoype.
1472 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
1473 * gimple-fold.h (get_range_strlen): Adjust prototype.
1474 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
1475 (format_string): Call get_string_length with eltsize.
1477 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1479 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
1480 to emit the span, rather than setting it as the prefix.
1482 2018-08-16 David Malcolm <dmalcolm@redhat.com>
1484 * diagnostic-show-locus.c (layout::start_annotation_line): Add
1485 "margin_char" parameter, defaulting to space. Use it in place
1486 of pp_space for the initial part of the margin.
1487 (layout::print_leading_fixits): Use '+' when filling the margin
1488 of line-insertion fix-it hints.
1490 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1492 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
1495 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1497 * config/rs6000/altivec.md: Don't set length attribute to the default
1499 * config/rs6000/darwin.md: Ditto.
1500 * config/rs6000/dfp.md: Ditto.
1501 * config/rs6000/htm.md: Ditto.
1502 * config/rs6000/rs6000.md: Ditto.
1503 * config/rs6000/sync.md: Ditto.
1504 * config/rs6000/vsx.md: Ditto.
1506 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1508 * config/rs6000/altivec.md: Don't set length attribute to the default
1509 value, for branch instructions.
1510 * config/rs6000/darwin.md: Ditto.
1511 * config/rs6000/rs6000.md: Ditto.
1513 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
1515 * config/rs6000/rs6000.md (length): Always define as const_int 4.
1516 (unnamed conditional branch define_insn): Set length to 4 or 8
1517 depending on offset.
1518 (<bd>_<mode>): Similar, for alternative 0.
1519 (<bd>tf_<mode>): Ditto.
1521 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1523 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
1525 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
1527 * doc/rtl.texi: Replace old RTX class names with new names.
1530 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
1532 * expmed.h (canonicalize_comparison): New declaration.
1533 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
1534 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
1535 * optabs.c (prepare_cmp_insn): Likewise.
1536 * rtl.h (unsigned_condition_p): New function which checks if a
1537 comparison operator is unsigned.
1539 2018-08-16 Nathan Sidwell <nathan@acm.org>
1541 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
1542 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
1544 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1547 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
1548 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
1549 (mov<mov>): ..this and enable unconditionally.
1551 2018-08-16 Tamar Christina <tamar.christina@arm.com>
1553 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
1555 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
1557 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
1558 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
1559 "Common" with "Target".
1561 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1563 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
1564 * doc/invoke.texi (mmitigate-rop): Remove.
1565 * config/i386/i386.c: Do not include "regrename.h".
1566 (ix86_rop_should_change_byte_p, reg_encoded_number)
1567 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
1569 (ix86_reorg): Remove call to ix86_mitigate_rop.
1570 * config/i386/i386.md (attr "modrm_class"): Remove.
1571 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
1572 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
1573 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
1575 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
1577 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
1578 allow folding of mergeh() and mergel() for the float and double types.
1579 (fold_mergehl_helper): Rework to handle building a permute tree
1582 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
1584 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
1587 2018-08-15 David Malcolm <dmalcolm@redhat.com>
1589 * common.opt (fdiagnostics-show-labels): New option.
1590 * diagnostic-show-locus.c (class layout_range): Add field
1592 (class layout): Add field "m_show_labels_p".
1593 (layout_range::layout_range): Add param "label" and use it to
1595 (make_range): Pass in NULL for new "label" param of layout_range's
1597 (layout::layout): Initialize m_show_labels_p.
1598 (layout::maybe_add_location_range): Pass in loc_range->m_label
1599 when constructing layout_range instances.
1600 (struct line_label): New struct.
1601 (layout::print_any_labels): New member function.
1602 (layout::print_line): Call it if label-printing is enabled.
1603 (selftest::test_one_liner_labels): New test.
1604 (selftest::test_diagnostic_show_locus_one_liner): Call it.
1605 * diagnostic.c (diagnostic_initialize): Initialize
1606 context->show_labels_p.
1607 * diagnostic.h (struct diagnostic_context): Add field
1609 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1610 -fno-diagnostics-show-labels.
1611 * dwarf2out.c (gen_producer_string): Add
1612 OPT_fdiagnostics_show_labels to the ignored options.
1613 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
1615 (gcc_rich_location::maybe_add_expr): Likewise.
1616 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
1617 label" param, defaulting to NULL.
1618 (gcc_rich_location::add_expr): Add "label" param.
1619 (gcc_rich_location::maybe_add_expr): Likewise.
1620 (class text_range_label): New class.
1621 (class range_label_for_type_mismatch): New class.
1622 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
1623 of format_warning_va.
1624 (fmtwarn_n): Likewise for new params of format_warning_n_va.
1625 * lto-wrapper.c (merge_and_complain): Add
1626 OPT_fdiagnostics_show_labels to the "pick one setting" options.
1627 (append_compiler_options): Likewise to the dropped options.
1628 (append_diag_options): Likewise to the passed-on options.
1629 * opts.c (common_handle_option): Handle the new option.
1630 * selftest-diagnostic.c
1631 (test_diagnostic_context::test_diagnostic_context): Enable
1633 * substring-locations.c: Include "gcc-rich-location.h".
1634 (format_warning_n_va): Add "fmt_label" and "param_label" params
1635 and use them as appropriate.
1636 (format_warning_va): Add "fmt_label" and "param_label" params,
1637 passing them on to format_warning_n_va.
1638 (format_warning_at_substring): Likewise.
1639 (format_warning_at_substring_n): Likewise.
1640 * substring-locations.h (format_warning_va): Add "fmt_label" and
1641 "param_label" params.
1642 (format_warning_n_va): Likewise.
1643 (format_warning_at_substring): Likewise.
1644 (format_warning_at_substring_n): Likewise.
1645 * toplev.c (general_init): Initialize global_dc->show_labels_p.
1647 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
1650 * builtins.c (expand_builtin_memcmp): Do not expand the call
1651 when overflow is detected.
1653 2018-08-15 Martin Sebor <msebor@redhat.com>
1655 PR tree-optimization/71625
1656 * config/aarch64/aarch64-builtins.c
1657 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
1659 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
1661 * config/s390/s390.c (s390_reorg): Remove loop.
1663 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1666 (darwin_function_switched_text_sections): Delete.
1667 * gcc/config/darwin.h
1668 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
1670 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
1673 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
1674 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
1675 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
1677 2018-08-15 Martin Liska <mliska@suse.cz>
1679 PR tree-optimization/86925
1680 * predict.c (expr_expected_value_1): When taking
1681 later predictor, assign also probability.
1682 Use fold_build2_initializer_loc in order to fold
1683 the expression in -frounding-math.
1685 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
1687 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
1689 (expand_vec_perm_1): Try the new method.
1691 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
1694 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
1695 Check whether lra_live_max_point is 0 before dividing.
1697 2018-08-14 Martin Sebor <msebor@redhat.com>
1699 PR tree-optimization/86650
1700 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
1701 (vrp_prop::check_mem_ref): Same.
1703 2018-08-13 Liu Hao <lh_mouse@126.com>
1705 * pretty-print.c (eat_esc_sequence): Swap the foreground and
1706 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
1707 and clear it thereafter, as it only works for DBCS.
1709 2018-08-13 Liu Hao <lh_mouse@126.com>
1711 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
1712 handle returned by _get_osf_handle().
1714 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1716 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
1717 for folding vec_perm.
1719 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1721 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
1722 Add support for gimple-folding of vec_pack() and vec_unpack()
1725 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
1727 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
1728 vec_xst variants to the list.
1729 (rs6000_gimple_fold_builtin): Add support for folding unaligned
1730 vector loads and stores.
1732 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
1734 * config.gcc (rs6000-ibm-aix4.x): Delete.
1735 (rs6000-ibm-aix5.1): Delete.
1736 (rs6000-ibm-aix5.2): Delete.
1737 (rs6000-ibm-aix5.3): Delete.
1738 * config/rs6000/aix43.h: Delete.
1739 * config/rs6000/aix51.h: Delete.
1740 * config/rs6000/aix52.h: Delete.
1741 * config/rs6000/t-aix43: Delete.
1743 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
1745 * config/s390/s390.c (s390_decompose_constant_pool_ref):
1747 (s390_decompose_address): Factor out constant pool ref
1750 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1752 * config/nds32/nds32-predicates.c
1753 (nds32_can_use_bclr_p): Change return type as bool.
1754 (nds32_can_use_bset_p): Ditto.
1755 (nds32_can_use_btgl_p): Ditto.
1756 (nds32_can_use_bitci_p): Ditto.
1757 * config/nds32/nds32-protos.h
1758 (nds32_can_use_bclr_p): Change declaration.
1759 (nds32_can_use_bset_p): Ditto.
1760 (nds32_can_use_btgl_p): Ditto.
1761 (nds32_can_use_bitci_p): Ditto.
1763 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1765 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
1766 Support -msched-prolog-epilog option.
1767 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
1769 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1771 * common/config/nds32/nds32-common.c
1772 (nds32_option_optimization_table): Enalbe -malways-align.
1774 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
1776 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
1778 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
1779 OPT_misr_secure_ case.
1780 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
1781 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
1782 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
1784 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
1785 (nds32_isr_info): New field security_level.
1786 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
1787 * config/nds32/nds32.md (return_internal): Consider critical attribute.
1788 * config/nds32/nds32.opt (misr-secure): New option.
1789 * config/nds32/nds32_init.inc: New file.
1790 * config/nds32/nds32_isr.h: New file.
1792 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
1794 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
1795 Update comment for atomic instructions.
1796 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
1797 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
1799 (atomic_loaddi): Revise fence expansion to only emit fence prior to
1800 load for __ATOMIC_SEQ_CST model.
1801 (atomic_loaddi_1): Remove float register target.
1802 (atomic_storedi): Handle CONST_INT values.
1803 (atomic_storedi_1): Remove float register source. Add special case
1805 (memory_barrier): New expander and insn.
1807 2018-08-11 Jakub Jelinek <jakub@redhat.com>
1809 PR tree-optimization/86835
1810 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
1811 new_stmt after def_gsi, make sure to insert new_square_stmt after
1812 that stmt, not 2 stmts before it.
1814 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
1817 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
1820 2018-08-10 Martin Liska <mliska@suse.cz>
1823 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
1825 * builtins.c (expand_builtin_expect_with_probability):
1827 (expand_builtin_expect_with_probability): New function.
1828 (build_builtin_expect_predicate): Add new argumnet probability
1829 for BUILT_IN_EXPECT_WITH_PROBABILITY.
1830 (fold_builtin_expect):
1833 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
1834 * builtins.h (fold_builtin_expect): Set new argument.
1835 * doc/extend.texi: Document __builtin_expect_with_probability.
1836 * doc/invoke.texi: Likewise.
1837 * gimple-fold.c (gimple_fold_call): Pass new argument.
1838 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
1839 also BUILT_IN_EXPECT_WITH_PROBABILITY.
1840 * predict.c (get_predictor_value): New function.
1841 (expr_expected_value): Add new argument probability. Assume
1842 that predictor and probability are always non-null.
1843 (expr_expected_value_1): Likewise. For __builtin_expect and
1844 __builtin_expect_with_probability set probability. Handle
1845 combination in binary expressions.
1846 (tree_predict_by_opcode): Simplify code by simply calling
1847 get_predictor_value.
1848 (pass_strip_predict_hints::execute): Add handling of
1849 BUILT_IN_EXPECT_WITH_PROBABILITY.
1850 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
1852 * tree.h (DECL_BUILT_IN_P): New function.
1854 2018-08-10 Martin Liska <mliska@suse.cz>
1856 PR tree-optimization/85799
1857 * passes.def: Add argument for pass_strip_predict_hints.
1858 * predict.c (class pass_strip_predict_hints): Add new argument
1860 (strip_predictor_early): New function.
1861 (pass_strip_predict_hints::execute): Call the function to
1863 (strip_predict_hints): New function.
1864 * predict.def: Fix comment.
1866 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
1868 * Makefile.in: Clarify which tm.texi to copy over to assert the
1869 right to grant a GFDL license for all.
1871 2018-08-09 Jeff Law <law@redhat.com>
1873 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
1876 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
1878 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
1881 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1883 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
1884 reductions for variable-length vectors.
1886 2018-08-09 David Malcolm <dmalcolm@redhat.com>
1889 * common.opt (fdiagnostics-show-line-numbers): New option.
1890 * diagnostic-show-locus.c (class layout): Add fields
1891 "m_show_line_numbers_p" and "m_linenum_width";
1892 (num_digits): New function.
1893 (test_num_digits): New function.
1894 (layout::layout): Initialize new fields. Update m_x_offset
1895 logic to handle any left margin.
1896 (layout::print_source_line): Print line number when requested.
1897 (layout::start_annotation_line): New member function.
1898 (layout::print_annotation_line): Call it.
1899 (layout::print_leading_fixits): Likewise.
1900 (layout::print_trailing_fixits): Likewise. Update calls to
1901 move_to_column for new parameter.
1902 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
1903 it to potentially call start_annotation_line.
1904 (layout::show_ruler): Call start_annotation_line.
1905 (selftest::test_line_numbers_multiline_range): New selftest.
1906 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
1907 and selftest::test_line_numbers_multiline_range.
1908 * diagnostic.c (diagnostic_initialize): Initialize
1909 show_line_numbers_p.
1910 * diagnostic.h (struct diagnostic_context): Add field
1911 "show_line_numbers_p".
1912 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
1913 -fno-diagnostics-show-line-numbers.
1914 * dwarf2out.c (gen_producer_string): Add
1915 OPT_fdiagnostics_show_line_numbers to the ignored options.
1916 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
1917 one setting" options.
1918 (append_compiler_options): Likewise to the dropped options.
1919 (append_diag_options): Likewise to the passed-on options.
1920 * opts.c (common_handle_option): Handle the new option.
1921 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
1923 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
1925 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
1926 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
1927 third argument of type "const signed char" to descriptions of
1928 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
1929 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
1930 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
1931 __builtin_bcdsub_ov functions.
1933 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1935 PR tree-optimization/86858
1936 * tree-vect-loop.c (vect_is_simple_reduction): Restore
1937 flow_bb_inside_loop_p calls.
1939 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
1941 PR tree-optimization/86871
1942 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
1943 instead of gimple_assign_lhs.
1945 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
1948 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
1949 register constraint to operand 0.
1950 (add<mode>3_carryinC): Likewise.
1951 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
1953 2018-08-09 Martin Liska <mliska@suse.cz>
1956 * common.opt: Remove extra line.
1958 2018-08-09 Martin Liska <mliska@suse.cz>
1960 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
1961 at the end of a line, make first letter capital and end up
1962 a sentence with a dot.
1963 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
1964 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
1965 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
1966 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
1967 (PARAM_MAX_ISL_OPERATIONS): Likewise.
1968 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
1969 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
1970 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
1971 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
1972 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
1973 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
1974 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
1975 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
1976 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
1977 (PARAM_TREE_REASSOC_WIDTH): Likewise.
1978 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
1979 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
1980 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
1982 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
1985 * config/s390/s390.c (s390_option_override_internal): Reduce the
1986 stack-clash-protection-probe-interval param if it would be too big
1989 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
1992 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
1993 * config/m68k/m68k.c (handle_move_double): Don't call
1994 m68k_final_prescan_insn.
1995 (m68k_adjust_decorated_operand): Renamed from
1996 m68k_final_prescan_insn, remove first and third operand and
1998 (print_operand): Call it.
1999 (print_operand_address): Call it.
2001 2018-08-08 Nathan Sidwell <nathan@acm.org>
2003 * diagnostic.c (diagnostic_report_current_module): Use
2004 linemap_included_from & linemap_included_from_linemap.
2006 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
2008 * config/aarch64/aarch64-cores.def: Add phecda core.
2009 * config/aarch64/aarch64-tune.md: Regenerate.
2010 * doc/invoke.texi: Add phecda core.
2012 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2015 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
2017 * config/s390/s390.md ("movti"): Add more alternatives for
2018 constant to GPR copies.
2020 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
2022 * config/s390/s390.c: Fix whitespace damage throughout the file.
2023 * config/s390/s390.h: Likewise.
2024 * config/s390/tpf.h: Likewise.
2026 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2028 * config/s390/s390.c (s390_loadrelative_operand_p):
2029 Remove TARGET_CPU_ZARCH usages.
2030 (s390_rtx_costs): Likewise.
2031 (s390_legitimate_constant_p): Likewise.
2032 (s390_cannot_force_const_mem): Likewise.
2033 (legitimate_reload_constant_p): Likewise.
2034 (s390_preferred_reload_class): Likewise.
2035 (legitimize_pic_address): Likewise.
2036 (legitimize_tls_address): Likewise.
2037 (s390_split_branches): Removed.
2038 (s390_add_execute): Removed.
2039 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
2040 (s390_mainpool_start): Likewise.
2041 (s390_mainpool_finish): Likewise.
2042 (s390_mainpool_cancel): Removed.
2043 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
2044 (s390_chunkify_cancel): Likewise.
2045 (s390_return_addr_rtx): Likewise.
2046 (s390_register_info): Remove split_branches_pending_p uages.
2047 (s390_optimize_register_info): Likewise.
2048 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
2049 split_branches_pending_p usages.
2050 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
2051 (s390_load_got): Likewise.
2052 (s390_expand_split_stack_prologue): Likewise.
2053 (output_asm_nops): Likewise.
2054 (s390_function_profiler): Likewise.
2055 (s390_emit_call): Likewise.
2056 (s390_conditional_register_usage): Likewise.
2057 (s390_optimize_prologue): Likewise.
2058 (s390_reorg): Remove TARGET_CPU_ZARCH and
2059 split_branches_pending_p usages.
2060 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
2062 (s390_output_indirect_thunk_function): Likewise.
2063 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
2064 (TARGET_CPU_ZARCH_P): Removed.
2065 (struct machine_function): Remove split_branches_pending_p.
2066 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
2068 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
2070 * common/config/s390/s390-common.c (processor_flags_table):
2072 * config.gcc: Remove with_arch/with_tune support.
2073 * config/s390/2064.md: Remove cpu attribute comparisons.
2074 * config/s390/driver-native.c (s390_host_detect_local_cpu):
2076 * config/s390/linux.h (ASM_SPEC):
2077 Remove -march support.
2078 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
2079 Use a table to get an arch level.
2080 * config/s390/s390-opts.h (enum processor_type):
2082 * config/s390/s390.c
2083 (processor_table): Remove entries, add arch_level values.
2084 (s390_issue_rate): Remove cases.
2085 (s390_option_override): Adjust
2086 s390_option_override_internal() call.
2087 (s390_option_override_internal): Remove deprecation warning.
2088 (s390_valid_target_attribute_tree): Adjust
2089 s390_option_override_internal() call.
2090 * config/s390/s390.h (struct s390_processor):
2091 Share with s390-c.c, add arch_level field.
2092 * config/s390/s390.md:
2093 Remove occurrences in cpu attribute.
2094 * config/s390/s390.opt: Remove -march/-mtune support.
2095 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
2096 * doc/invoke.texi: Remove deprecation warning.
2098 2018-08-08 Luis Machado <luis.machado@linaro.org>
2100 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
2102 (qdf24xx_tunings): Set vector cost structure to
2103 qdf24xx_vector_cost.
2105 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
2106 <register_sextend>: Set to 3.
2108 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
2111 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
2112 * config/aarch64/aarch64-simd.md
2113 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
2114 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
2115 (aarch64_frecpx<mode>): ...this new pattern.
2116 * config/aarch64/aarch64-simd-builtins.def: Remove comment
2117 about aarch64_frecp<FRECP:frecp_suffix><mode>.
2119 2018-08-07 Martin Liska <mliska@suse.cz>
2122 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
2123 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
2124 * predict.def (PRED_MALLOC_NONNULL): New predictor.
2125 * doc/extend.texi: Document that malloc attribute adds
2128 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
2131 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2132 Define to speculation_safe_value_not_needed.
2134 2018-08-06 Jeff Law <law@redhat.com>
2136 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
2137 the vr_values instance to cprop_into_stmt.
2138 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
2139 (cprop_operand): Also query EVRP to determine if OP is a constant.
2141 2018-08-06 Nathan Sidwell <nathan@acm.org>
2143 * diagnostic.c (diagnostic_report_current_module): Reroll
2144 included-at loop. Translate text.
2146 2018-08-06 David Malcolm <dmalcolm@redhat.com>
2148 * function-tests.c (selftest::test_expansion_to_rtl): Call
2149 free_after_compilation.
2151 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2153 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
2155 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
2157 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
2158 loops with memory block operations from getting unrolled.
2160 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
2163 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2164 Define to speculation_safe_value_not_needed.
2166 2018-08-06 Jeff Law <law@redhat.com>
2168 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
2171 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2174 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
2175 with all enabled __intN types.
2177 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
2179 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2181 * alias.c (record_set): Check for clobber high.
2182 * cfgexpand.c (expand_gimple_stmt): Likewise.
2183 * combine-stack-adj.c (single_set_for_csa): Likewise.
2184 * combine.c (find_single_use_1): Likewise.
2185 (set_nonzero_bits_and_sign_copies): Likewise.
2186 (get_combine_src_dest): Likewise.
2187 (is_parallel_of_n_reg_sets): Likewise.
2188 (try_combine): Likewise.
2189 (record_dead_and_set_regs_1): Likewise.
2190 (reg_dead_at_p_1): Likewise.
2191 (reg_dead_at_p): Likewise.
2192 * dce.c (deletable_insn_p): Likewise.
2193 (mark_nonreg_stores_1): Likewise.
2194 (mark_nonreg_stores_2): Likewise.
2195 * df-scan.c (df_find_hard_reg_defs): Likewise.
2196 (df_uses_record): Likewise.
2197 (df_get_call_refs): Likewise.
2198 * dwarf2out.c (mem_loc_descriptor): Likewise.
2199 * haifa-sched.c (haifa_classify_rtx): Likewise.
2200 * ira-build.c (create_insn_allocnos): Likewise.
2201 * ira-costs.c (scan_one_insn): Likewise.
2202 * ira.c (equiv_init_movable_p): Likewise.
2203 (rtx_moveable_p): Likewise.
2204 (interesting_dest_for_shprep): Likewise.
2205 * jump.c (mark_jump_label_1): Likewise.
2206 * postreload-gcse.c (record_opr_changes): Likewise.
2207 * postreload.c (reload_cse_simplify): Likewise.
2208 (struct reg_use): Add source expr.
2209 (reload_combine): Check for clobber high.
2210 (reload_combine_note_use): Likewise.
2211 (reload_cse_move2add): Likewise.
2212 (move2add_note_store): Likewise.
2213 * print-rtl.c (print_pattern): Likewise.
2214 * recog.c (decode_asm_operands): Likewise.
2215 (store_data_bypass_p): Likewise.
2216 (if_test_bypass_p): Likewise.
2217 * regcprop.c (kill_clobbered_value): Likewise.
2218 (kill_set_value): Likewise.
2219 * reginfo.c (reg_scan_mark_refs): Likewise.
2220 * reload1.c (maybe_fix_stack_asms): Likewise.
2221 (eliminate_regs_1): Likewise.
2222 (elimination_effects): Likewise.
2223 (mark_not_eliminable): Likewise.
2224 (scan_paradoxical_subregs): Likewise.
2225 (forget_old_reloads_1): Likewise.
2226 * reorg.c (find_end_label): Likewise.
2227 (try_merge_delay_insns): Likewise.
2228 (redundant_insn): Likewise.
2229 (own_thread_p): Likewise.
2230 (fill_simple_delay_slots): Likewise.
2231 (fill_slots_from_thread): Likewise.
2232 (dbr_schedule): Likewise.
2233 * resource.c (update_live_status): Likewise.
2234 (mark_referenced_resources): Likewise.
2235 (mark_set_resources): Likewise.
2236 * rtl.c (copy_rtx): Likewise.
2237 * rtlanal.c (reg_referenced_p): Likewise.
2238 (single_set_2): Likewise.
2239 (noop_move_p): Likewise.
2240 (note_stores): Likewise.
2241 * sched-deps.c (sched_analyze_reg): Likewise.
2242 (sched_analyze_insn): Likewise.
2244 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2246 * cse.c (invalidate_reg): New function extracted from...
2247 (invalidate): ...here.
2248 (canonicalize_insn): Check for clobber high.
2249 (invalidate_from_clobbers): invalidate clobber highs.
2250 (invalidate_from_sets_and_clobbers): Likewise.
2251 (count_reg_usage): Check for clobber high.
2252 (insn_live_p): Likewise.
2253 * cselib.c (cselib_expand_value_rtx_1):Likewise.
2254 (cselib_invalidate_regno): Check for clobber in setter.
2255 (cselib_invalidate_rtx): Pass through setter.
2256 (cselib_invalidate_rtx_note_stores):
2257 (cselib_process_insn): Check for clobber high.
2258 * cselib.h (cselib_invalidate_rtx): Add operand.
2260 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2262 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
2263 (mark_not_eliminable): Likewise.
2264 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
2265 * lra-lives.c (process_bb_lives): Check for clobber high.
2266 * lra.c (new_insn_reg): Remember clobber highs.
2267 (collect_non_operand_hard_regs): Check for clobber high.
2268 (lra_set_insn_recog_data): Likewise.
2269 (add_regs_to_insn_regno_info): Likewise.
2270 (lra_update_insn_regno_info): Likewise.
2272 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2274 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
2275 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
2277 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2279 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
2280 (copy_insn_1): Likewise.
2281 (gen_hard_reg_clobber_high): New gen function.
2282 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
2283 * genemit.c (gen_exp): Likewise.
2284 (gen_emit_seq): Pass through info.
2285 (gen_insn): Check for CLOBBER_HIGH.
2286 (gen_expand): Pass through info.
2287 (gen_split): Likewise.
2288 (output_add_clobbers): Likewise.
2289 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
2290 (remove_clobbers): Likewise.
2291 * rtl.h (gen_hard_reg_clobber_high): New declaration.
2293 2018-08-06 Alan Hayward <alan.hayward@arm.com>
2295 * doc/rtl.texi (clobber_high): Add.
2296 (parallel): Add in clobber high
2297 * rtl.c (rtl_check_failed_code3): Add function.
2298 * rtl.def (CLOBBER_HIGH): Add expression.
2299 * rtl.h (RTL_CHECKC3): Add macro.
2300 (rtl_check_failed_code3): Add declaration.
2301 (XC3EXP): Add macro.
2303 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
2306 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
2307 cfun->machine->max_used_stack_alignment if needed.
2309 2018-08-04 Martin Sebor <msebor@redhat.com>
2311 PR tree-optimization/86571
2312 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
2315 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
2318 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2321 2018-08-03 Jeff Law <law@redhat.com>
2324 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2325 Define to speculation_safe_value_not_needed.
2327 2018-08-03 David Malcolm <dmalcolm@redhat.com>
2329 * doc/gcov.texi (-x): Remove duplicate "to".
2330 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
2331 (-Wif-not-aligned): Remove duplicate "is".
2332 (-flto): Remove duplicate "the".
2333 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
2334 duplicate "v5.00.b".
2335 (MSP430 Options): Remove duplicate "and" from the description
2336 of "-mgprel-sec=regexp".
2337 (x86 Options): Remove duplicate copies of "vmldLog102" and
2338 vmlsLog104 from description of "-mveclibabi=type".
2340 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
2342 * internal-fn.h (first_commutative_argument): Declare.
2343 * internal-fn.c (first_commutative_argument): New function.
2344 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
2345 restrictions for pattern statements. Use first_commutative_argument
2346 to look for commutative operands in calls to internal functions.
2348 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
2350 * Makefile.in (wide-int-range.o): New.
2351 * tree-vrp.c: Move all the wide_int_* functions to...
2352 * wide-int-range.cc: ...here.
2353 * tree-vrp.h: Move all the wide_int_* prototypes to...
2354 * wide-int-range.h: ...here.
2356 2018-08-03 Tom de Vries <tdevries@suse.de>
2358 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
2360 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
2361 * except.c (output_function_exception_table): Do early exit if
2362 targetm_common.except_unwind_info (&global_options) == UI_NONE.
2364 2018-08-03 Martin Liska <mliska@suse.cz>
2366 * predict.c (dump_prediction): Change to 2 digits
2369 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
2371 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
2373 (falkor_am_1_gtov_gtov): ... a new insn reservation.
2375 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
2377 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
2378 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
2379 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
2380 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
2381 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
2383 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2385 * diagnostic-show-locus.c (diagnostic_show_locus): Use
2386 pp_take_prefix when saving the existing prefix.
2387 * diagnostic.c (diagnostic_append_note): Likewise.
2388 * langhooks.c (lhd_print_error_function): Likewise.
2389 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
2390 param's type. Free the existing prefix.
2391 (pp_take_prefix): New function.
2392 (pretty_printer::pretty_printer): Drop the prefix parameter.
2393 Rename the length parameter to match the comment.
2394 (pretty_printer::~pretty_printer): Free the prefix.
2395 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
2397 (struct pretty_printer): Drop the "const" from "prefix" field's
2398 type and clarify memory management.
2399 (pp_set_prefix): Drop the "const" from the 2nd param.
2400 (pp_take_prefix): New decl.
2402 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2404 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
2405 for word_mode_ok here instead of passing as argument.
2406 (expand_block_compare): Change select_block_compare_mode() call.
2407 (expand_strncmp_gpr_sequence): New function.
2408 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
2410 2018-08-02 Jeff Law <law@redhat.com>
2413 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2414 Define to speculation_safe_value_not_needed.
2417 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2418 Define to speculation_safe_value_not_needed.
2420 2018-08-02 Tom de Vries <tdevries@suse.de>
2423 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
2424 function. Return UI_TARGET unconditionally.
2425 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
2426 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
2428 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2430 * genemit.c (print_overload_test): Fix typo.
2432 2018-08-02 Richard Biener <rguenther@suse.de>
2434 PR tree-optimization/86816
2435 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
2436 which checks for value availability before querying it.
2437 (gvn_uses_equal): Use it.
2438 (same_succ_hash): Likewise.
2439 (gimple_equal_p): Likewise.
2441 2018-08-02 Nick Clifton <nickc@redhat.com>
2444 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2445 Define to speculation_safe_value_not_needed.
2448 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2449 Define to speculation_safe_value_not_needed.
2452 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2453 Define to speculation_safe_value_not_needed.
2456 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2457 Define to speculation_safe_value_not_needed.
2460 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2461 Define to speculation_safe_value_not_needed.
2464 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2465 Define to speculation_safe_value_not_needed.
2468 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2469 Define to speculation_safe_value_not_needed.
2472 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
2473 Define to speculation_safe_value_not_needed.
2476 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2477 speculation_safe_value_not_needed.
2480 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
2481 to speculation_safe_value_not_needed.
2483 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2485 * doc/md.texi: Expand the documentation of instruction names
2486 to mention port-local uses. Document '@' in pattern names.
2487 * read-md.h (overloaded_instance, overloaded_name): New structs.
2489 (md_reader::handle_overloaded_name): New member function.
2490 (md_reader::get_overloads): Likewise.
2491 (md_reader::m_first_overload): New member variable.
2492 (md_reader::m_next_overload_ptr): Likewise.
2493 (md_reader::m_overloads_htab): Likewise.
2494 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
2495 m_next_overload_ptr and m_overloads_htab.
2496 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
2497 (get_mode_token, get_code_token, get_int_token): New functions.
2498 (map_attr_string): Add an optional argument that passes back
2499 the associated iterator.
2500 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
2501 (md_reader::handle_overloaded_name, add_overload_instance): New
2503 (apply_iterators): Handle '@' names. Report an error if '@'
2504 is used without iterators.
2505 (initialize_iterators): Initialize the new iterator_group fields.
2506 * genopinit.c (handle_overloaded_code_for)
2507 (handle_overloaded_gen): New functions.
2508 (main): Use them to print declarations of maybe_code_for_* and
2509 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
2510 * genemit.c (print_overload_arguments, print_overload_test)
2511 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
2512 (main): Use it to print definitions of maybe_code_for_* and
2513 maybe_gen_* functions.
2514 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
2515 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
2516 instead of explicit mode checks.
2517 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
2518 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
2519 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
2520 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
2521 (aarch64_expand_compare_and_swap): Likewise
2522 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
2523 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
2524 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
2525 (aarch64_constant_pool_reload_icode): Delete.
2526 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
2527 instead of aarch64_constant_pool_reload_icode. Use
2528 code_for_aarch64_reload_mov instead of explicit mode checks.
2529 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
2530 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
2531 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
2532 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
2533 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
2534 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
2535 (aarch64_atomic_load_op_code): Delete.
2536 (aarch64_emit_atomic_load_op): Likewise.
2537 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
2538 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
2539 instead of aarch64_emit_atomic_load_op.
2540 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
2541 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
2542 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
2543 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
2544 character before the pattern name.
2545 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
2546 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
2547 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
2548 (aarch64_frecps<mode>): Likewise.
2549 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
2550 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
2551 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
2552 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
2553 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
2555 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
2557 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
2558 Allow HFmode constants if TARGET_FP_F16INST.
2560 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
2563 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
2564 No longer check last store for clobber of address register.
2566 2018-08-02 Martin Liska <mliska@suse.cz>
2568 PR gcov-profile/86817
2569 * gcov.c (process_all_functions): New function.
2571 (process_file): Move functions processing to
2572 process_all_functions.
2574 2018-08-02 David Malcolm <dmalcolm@redhat.com>
2576 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
2577 "const" to the "gimple *" and "rtx_insn *" parameters.
2578 * dumpfile.h (dump_user_location_t::dump_user_location_t):
2580 (dump_location_t::dump_location_t): Likewise.
2582 2018-08-01 Martin Sebor <msebor@redhat.com>
2584 PR tree-optimization/86650
2585 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
2586 rather than a "gcall *". Directly pass the data of interest
2587 to percent_K_format, rather than building a temporary CALL_EXPR
2589 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
2590 (gimple_fold_builtin_strncat): Adjust.
2591 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
2592 gcall* argument with gimple*.
2593 * gimple-ssa-warn-restrict.c (check_call): Same.
2594 (wrestrict_dom_walker::before_dom_children): Same.
2595 (builtin_access::builtin_access): Same.
2596 (check_bounds_or_overlap): Same
2597 (maybe_diag_overlap): Same.
2598 (maybe_diag_offset_bounds): Same.
2599 * tree-diagnostic.c (default_tree_printer): Move usage of
2600 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
2602 * tree-pretty-print.c (percent_K_format): Add argument.
2603 * tree-pretty-print.h: Add argument.
2604 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
2605 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
2606 (maybe_diag_stxncpy_trunc): Same.
2607 (handle_builtin_stxncpy): Same.
2608 (handle_builtin_strcat): Same.
2610 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2612 * match.pd: Optimise pointer range checks.
2614 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2616 PR tree-optimization/86758
2617 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
2618 to remove pattern statements.
2620 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2622 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
2623 result of dfs_enumerate_from when constructing stmt_vec_infos,
2624 instead of additionally calling get_loop_body.
2626 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2628 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
2630 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
2631 When creating an iv, assert that the step is not known to be zero.
2632 (vect_setup_realignment): Update call accordingly.
2633 * tree-vect-stmts.c (vectorizable_store): Likewise.
2634 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
2636 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2638 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
2639 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
2640 (vectorizable_reduction): Likewise.
2641 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2642 (vect_detect_hybrid_slp_stmts): Likewise.
2643 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2645 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
2647 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
2648 (wide_int_set_zero_nonzero_bits): ...this.
2649 (zero_nonzero_bits_from_vr): Rename to...
2650 (vrp_set_zero_nonzero_bits): ...this.
2651 (extract_range_from_multiplicative_op_1): Abstract wide int
2653 (wide_int_range_multiplicative_op): ...here.
2654 (extract_range_from_binary_expr_1): Extract wide int binary
2655 operations into their own functions.
2656 (wide_int_range_lshift): New.
2657 (wide_int_range_can_optimize_bit_op): New.
2658 (wide_int_range_shift_undefined_p): New.
2659 (wide_int_range_bit_xor): New.
2660 (wide_int_range_bit_ior): New.
2661 (wide_int_range_bit_and): New.
2662 (wide_int_range_trunc_mod): New.
2663 (extract_range_into_wide_ints): New.
2664 (vrp_shift_undefined_p): New.
2665 (extract_range_from_multiplicative_op): New.
2666 (vrp_can_optimize_bit_op): New.
2667 * tree-vrp.h (value_range::dump): New.
2668 (wide_int_range_multiplicative_op): New.
2669 (wide_int_range_lshift):New.
2670 (wide_int_range_shift_undefined_p): New.
2671 (wide_int_range_bit_xor): New.
2672 (wide_int_range_bit_ior): New.
2673 (wide_int_range_bit_and): New.
2674 (wide_int_range_trunc_mod): New.
2675 (zero_nonzero_bits_from_bounds): Rename to...
2676 (wide_int_set_zero_nonzero_bits): ...this.
2677 (zero_nonzero_bits_from_vr): Rename to...
2678 (vrp_set_zero_nonzero_bits): ...this.
2679 (range_easy_mask_min_max): Rename to...
2680 (wide_int_range_can_optimize_bit_op): this.
2681 * vr-values.c (simplify_bit_ops_using_ranges): Rename
2682 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
2684 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2686 * tree-vectorizer.h (vect_orig_stmt): New function.
2687 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
2688 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
2689 (vect_create_epilog_for_reduction): Likewise.
2690 (vectorizable_live_operation): Likewise.
2691 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2692 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
2693 * tree-vect-stmts.c (vectorizable_call): Likewise.
2694 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
2696 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2698 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
2700 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
2701 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
2702 (vect_transform_loop): Likewise.
2703 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2705 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2707 * tree-vectorizer.h (vect_schedule_slp): Return void.
2708 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
2709 (vect_schedule_slp): Likewise.
2711 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2713 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
2715 (vect_transform_loop): Update calls accordingly. Schedule SLP
2716 instances before the main loop, if any exist.
2718 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
2720 PR tree-optimization/86749
2721 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
2722 If the lhs is used in a COND_EXPR, check that it is being used
2723 as the "then" or "else" value.
2725 2018-08-01 Tom de Vries <tdevries@suse.de>
2728 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
2729 speculation_safe_value_not_needed.
2731 2018-08-01 Richard Biener <rguenther@suse.de>
2733 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
2736 2018-08-01 Martin Liska <mliska@suse.cz>
2738 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
2739 format how successful transformation is dumped.
2740 (gimple_mod_pow2_value_transform): Likewise.
2741 (gimple_mod_subtract_transform): Likewise.
2742 (gimple_stringops_transform): Likewise.
2744 2018-08-01 Martin Liska <mliska@suse.cz>
2747 * value-prof.c (interesting_stringop_to_profile_p):
2748 Simplify the code and add BUILT_IN_MEMMOVE.
2749 (gimple_stringops_transform): Likewise.
2751 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
2753 * config/aarch64/aarch64-simd.md
2754 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
2755 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
2756 use GPI iterator instead of SI mode.
2758 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2760 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
2761 rs6000_speculation_barrier.
2762 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
2763 new barrier pattern name.
2765 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
2767 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
2768 (speculation_barrier): New insn.
2770 2018-08-01 Richard Biener <rguenther@suse.de>
2773 * graphite.h: Include isl/id.h and isl/space.h to allow build
2776 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
2779 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2780 mode for COFF targets.
2781 * defaults.h (TARGET_COFF): Define.
2782 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
2783 TARGET_COFF): Define.
2784 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
2785 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
2786 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
2788 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2790 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
2791 * gimple-streamer-out.c (output_bb): Save it.
2792 * lto-streamer-in.c (input_struct_function_base): Restore
2793 instance discriminator if available. Create map on demand.
2794 * lto-streamer-out.c (output_struct_function_base): Save it if
2796 * final.c (decl_to_instance_map): Document LTO strategy.
2798 2018-07-31 Alexandre Oliva <oliva@adacore.com>
2799 Olivier Hainque <hainque@adacore.com>
2801 * debug.h (decl_to_instance_map_t): New type.
2802 (decl_to_instance_map): Declare.
2803 (maybe_create_decl_to_instance_map): New inline function.
2804 * final.c (bb_discriminator, last_bb_discriminator): New statics,
2805 to track basic block discriminators.
2806 (final_start_function_1): Initialize them.
2807 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
2809 (decl_to_instance_map): New variable.
2810 (map_decl_to_instance, maybe_set_discriminator): New functions.
2811 (notice_source_line): Set discriminator.
2813 2018-07-31 Ian Lance Taylor <iant@golang.org>
2815 * targhooks.c (default_have_speculation_safe_value): Add
2818 2018-07-31 David Malcolm <dmalcolm@redhat.com>
2820 * dump-context.h: Include "pretty-print.h".
2821 (dump_context::refresh_dumps_are_enabled): New decl.
2822 (dump_context::emit_item): New decl.
2823 (class dump_context): Add fields "m_test_pp" and
2825 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
2826 (temp_dump_context::get_dumped_text): New decl.
2827 (class temp_dump_context): Add field "m_pp".
2828 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
2829 (dump_context::refresh_dumps_are_enabled): ...and add a test for
2831 (set_dump_file): Update for above change.
2832 (set_alt_dump_file): Likewise.
2833 (dump_loc): New overload, taking a pretty_printer *.
2834 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
2835 to any test pretty-printer.
2836 (make_item_for_dump_gimple_stmt): New function, adapted from
2837 optinfo::add_gimple_stmt.
2838 (dump_context::dump_gimple_stmt): Call it, and use the result,
2839 eliminating the direct usage of dump_file and alt_dump_file in
2840 favor of indirectly using them via emit_item.
2841 (make_item_for_dump_gimple_expr): New function, adapted from
2842 optinfo::add_gimple_expr.
2843 (dump_context::dump_gimple_expr): Call it, and use the result,
2844 eliminating the direct usage of dump_file and alt_dump_file in
2845 favor of indirectly using them via emit_item.
2846 (make_item_for_dump_generic_expr): New function, adapted from
2848 (dump_context::dump_generic_expr): Call it, and use the result,
2849 eliminating the direct usage of dump_file and alt_dump_file in
2850 favor of indirectly using them via emit_item.
2851 (make_item_for_dump_printf_va): New function, adapted from
2852 optinfo::add_printf_va.
2853 (make_item_for_dump_printf): New function.
2854 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
2855 and use the result, eliminating the direct usage of dump_file and
2856 alt_dump_file in favor of indirectly using them via emit_item.
2857 (make_item_for_dump_dec): New function.
2858 (dump_context::dump_dec): Call it, and use the result,
2859 eliminating the direct usage of dump_file and alt_dump_file in
2860 favor of indirectly using them via emit_item.
2861 (make_item_for_dump_symtab_node): New function, adapted from
2862 optinfo::add_symtab_node.
2863 (dump_context::dump_symtab_node): Call it, and use the result,
2864 eliminating the direct usage of dump_file and alt_dump_file in
2865 favor of indirectly using them via emit_item.
2866 (dump_context::begin_scope): Reimplement, avoiding direct usage
2867 of dump_file and alt_dump_file in favor of indirectly using them
2869 (dump_context::emit_item): New member function.
2870 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
2871 Set up test pretty-printer on the underlying context. Call
2872 refresh_dumps_are_enabled.
2873 (temp_dump_context::~temp_dump_context): Call
2874 refresh_dumps_are_enabled.
2875 (temp_dump_context::get_dumped_text): New member function.
2876 (selftest::verify_dumped_text): New function.
2877 (ASSERT_DUMPED_TEXT_EQ): New macro.
2878 (selftest::test_capture_of_dump_calls): Run all tests twice, with
2879 and then without optinfo enabled. Add uses of
2880 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
2881 * dumpfile.h: Update comment for the dump_* API.
2882 * optinfo-emit-json.cc
2883 (selftest::test_building_json_from_dump_calls): Update for new
2884 param for temp_dump_context ctor.
2885 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
2886 and "m_owned" field.
2887 (optinfo_item::~optinfo_item): Likewise.
2888 (optinfo::add_item): New member function.
2889 (optinfo::emit): Update comment.
2890 (optinfo::add_string): Delete.
2891 (optinfo::add_printf): Delete.
2892 (optinfo::add_printf_va): Delete.
2893 (optinfo::add_gimple_stmt): Delete.
2894 (optinfo::add_gimple_expr): Delete.
2895 (optinfo::add_tree): Delete.
2896 (optinfo::add_symtab_node): Delete.
2897 (optinfo::add_dec): Delete.
2898 * optinfo.h (class dump_context): New forward decl.
2899 (optinfo::add_item): New decl.
2900 (optinfo::add_string): Delete.
2901 (optinfo::add_printf): Delete.
2902 (optinfo::add_printf_va): Delete.
2903 (optinfo::add_gimple_stmt): Delete.
2904 (optinfo::add_gimple_expr): Delete.
2905 (optinfo::add_tree): Delete.
2906 (optinfo::add_symtab_node): Delete.
2907 (optinfo::add_dec): Delete.
2908 (optinfo::add_poly_int): Delete.
2909 (optinfo_item::optinfo_item): Remove "owned" param.
2910 (class optinfo_item): Remove field "m_owned".
2912 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2915 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
2916 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
2917 requested variable alignment.
2918 (expand_one_ssa_partition): Likewise.
2919 (expand_one_var): Likewise.
2921 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2923 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2924 to speculation_safe_value_not_needed.
2926 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2928 * targhooks.h (speculation_safe_value_not_needed): New prototype.
2929 * targhooks.c (speculation_safe_value_not_needed): New function.
2930 * target.def (have_speculation_safe_value): Update documentation.
2931 * doc/tm.texi: Regenerated.
2933 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2935 * config/aarch64/iterators.md (ALLI_TI): New iterator.
2936 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
2938 (despeculate_copy<ALLI:mode>_insn): New insn.
2939 (despeculate_copyti_insn): New insn.
2940 (despeculate_simple<ALLI:mode>): New insn
2941 (despeculate_simpleti): New insn.
2942 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
2944 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
2945 aarch64_speculation_safe_value.
2946 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
2948 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2950 * config/aarch64/aarch64-speculation.cc: New file.
2951 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
2952 before pass_reorder_blocks.
2953 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
2955 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
2956 X14 and X15 when tracking speculation.
2957 * config/aarch64/aarch64.md (register name constants): Add
2958 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
2959 (unspec): Add UNSPEC_SPECULATION_TRACKER.
2960 (speculation_barrier): New insn attribute.
2961 (cmp<mode>): Allow SP in comparisons.
2962 (speculation_tracker): New insn.
2963 (speculation_barrier): Add speculation_barrier attribute.
2964 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
2965 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
2966 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
2968 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2970 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
2971 aarch64_track_speculation is true.
2972 (tb<optab><mode>1): Likewise.
2973 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
2974 generate CB[N]Z when tracking speculation.
2975 (aarch64_split_compare_and_swap): Likewise.
2976 (aarch64_split_atomic_op): Likewise.
2978 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2980 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
2982 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2984 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
2985 (speculation_barrier): New insn.
2987 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2989 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
2990 * config/arm/arm.md (speculation_barrier): New expand.
2991 (speculation_barrier_insn): New pattern.
2993 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
2995 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
2996 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
2997 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
2998 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3000 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
3001 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
3002 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
3003 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
3004 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
3005 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
3006 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
3007 * builtins.c (expand_speculation_safe_value): New function.
3008 (expand_builtin): Call it.
3009 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
3010 * doc/extend.texi: Document __builtin_speculation_safe_value.
3011 * doc/md.texi: Document "speculation_barrier" pattern.
3012 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
3013 TARGET_HAVE_SPECULATION_SAFE_VALUE.
3014 * doc/tm.texi: Regenerated.
3015 * target.def (have_speculation_safe_value, speculation_safe_value): New
3017 * targhooks.c (default_have_speculation_safe_value): New function.
3018 (default_speculation_safe_value): New function.
3019 * targhooks.h (default_have_speculation_safe_value): Add prototype.
3020 (default_speculation_safe_value): Add prototype.
3022 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3024 * dump-context.h (dump_context::dump_loc): New decl.
3025 * dumpfile.c (dump_context::dump_loc): New member function.
3026 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
3027 and dump_gimple_stmt.
3028 (dump_context::dump_gimple_expr_loc): Likewise, using
3030 (dump_context::dump_generic_expr_loc): Likewise, using
3032 (dump_context::dump_printf_loc_va): Likewise, using
3034 (dump_context::begin_scope): Explicitly using the global function
3035 "dump_loc", rather than the member function.
3037 2018-07-31 Martin Sebor <msebor@redhat.com>
3039 PR tree-optimization/86741
3040 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
3042 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
3044 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
3045 depend on whether prefetch instructions will be emitted or not.
3046 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
3047 will be emitted or not.
3048 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
3049 (TARGET_SETMEM_PFD): New macros.
3051 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3053 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
3054 (NULL_STMT_VEC_INFO): Delete.
3055 (stmt_vec_info::operator*): Likewise.
3056 (stmt_vec_info::operator gimple *): Likewise.
3057 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
3058 of NULL_STMT_VEC_INFO.
3059 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3060 (vect_reassociating_reduction_p): Likewise.
3061 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3062 (vectorizable_store): Likewise.
3063 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
3064 (vec_info::free_stmt_vec_infos): Likewise.
3066 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3068 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
3069 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
3071 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3073 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
3074 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3075 (vec_info::free_stmt_vec_info): New private member functions.
3076 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
3077 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
3078 * tree-parloops.c (gather_scalar_reductions): Remove calls to
3079 set_stmt_vec_info_vec and free_stmt_vec_infos.
3080 * tree-vect-loop.c (_loop_vec_info): Remove call to
3081 set_stmt_vec_info_vec.
3082 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
3083 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
3084 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
3085 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
3086 (vec_info::free_stmt_vec_info): ...these new functions. Remove
3087 assignments in {vec_info::,}new_stmt_vec_info that are redundant
3088 with the clearing in the xcalloc.
3089 (stmt_vec_info_vec): Delete.
3090 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
3091 (vectorize_loops): Likewise.
3092 (vec_info::~vec_info): Remove argument from call to
3093 free_stmt_vec_infos.
3094 (vec_info::add_stmt): Remove vinfo argument from call to
3097 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3099 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
3100 rather than a gimple stmt.
3101 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
3102 information for pattern statements when passed the original
3103 statement; instead wait to be passed the pattern statement itself.
3104 Don't call set_vinfo_for_stmt here.
3105 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
3106 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
3107 stmt_vec_infos here.
3108 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
3109 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
3110 stmt_vec_infos entry.
3112 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3114 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
3115 * tree-vectorizer.c (vec_info::replace_stmt): New function.
3116 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
3117 * tree-vect-stmts.c (vectorizable_call): Likewise.
3118 (vectorizable_simd_clone_call): Likewise.
3120 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3122 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
3123 * tree-vectorizer.c (vec_info::remove_stmt): New function.
3124 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
3125 * tree-vect-loop.c (vect_transform_loop): Likewise.
3126 * tree-vect-slp.c (vect_schedule_slp): Likewise.
3127 * tree-vect-stmts.c (vect_remove_stores): Likewise.
3129 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3131 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
3132 (vect_dr_stmt): Delete.
3133 * tree-vectorizer.c (vec_info::lookup_dr): New function.
3134 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
3136 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
3137 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
3138 (vect_verify_datarefs_alignment, vect_peeling_supportable)
3139 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3140 (vect_analyze_data_refs): Likewise.
3141 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
3143 (vect_find_same_alignment_drs): Likewise.
3144 (vect_slp_analyze_node_dependences): Update calls accordingly.
3145 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3146 instead of DR_VECT_AUX.
3147 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
3148 of a vector data references. Use vec_info::lookup_dr instead of
3150 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
3151 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
3152 instead of DR_VECT_AUX.
3154 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3156 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
3158 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3160 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3161 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3162 (vect_gen_prolog_loop_niters): Likewise.
3164 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3166 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
3167 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
3168 (vect_known_alignment_in_bytes, vect_dr_behavior)
3169 (vect_get_scalar_dr_size): Take references as dr_vec_infos
3170 instead of data_references. Update calls to other routines for
3171 which the same change has been made.
3172 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
3173 dr_vec_infos instead of stmt_vec_infos.
3174 (vect_analyze_data_ref_dependence): Update call accordingly.
3175 (vect_slp_analyze_data_ref_dependence)
3176 (vect_record_base_alignments): Use DR_VECT_AUX.
3177 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
3178 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
3179 (vector_alignment_reachable_p, vect_get_data_access_cost)
3180 (vect_peeling_supportable, vect_analyze_group_access_1)
3181 (vect_analyze_group_access, vect_analyze_data_ref_access)
3182 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
3183 (vect_compile_time_alias, vect_small_gap_p)
3184 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
3185 (vect_supportable_dr_alignment): Take references as dr_vec_infos
3186 instead of data_references. Update calls to other routines for
3187 which the same change has been made.
3188 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
3189 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3190 (vect_slp_analyze_and_verify_node_alignment)
3191 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
3192 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3193 (vect_setup_realignment): Use dr_vec_infos. Update calls after
3195 (_vect_peel_info::dr): Replace with...
3196 (_vect_peel_info::dr_info): ...this new field.
3197 (vect_peeling_hash_get_most_frequent)
3198 (vect_peeling_hash_choose_best_peeling): Update accordingly.
3199 (vect_peeling_hash_get_lowest_cost):
3200 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
3201 routines for which the same change has been made.
3202 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
3204 * tree-vect-loop-manip.c (get_misalign_in_elems)
3205 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
3207 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3208 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
3209 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
3210 (get_group_load_store_type, get_negative_load_store_type)
3211 (vect_get_data_ptr_increment, vectorizable_store)
3212 (vectorizable_load): Likewise.
3213 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
3214 Update calls to other routines for which the same change has been made.
3216 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3218 * tree-vectorizer.h (vec_info::move_dr): New member function.
3219 (dataref_aux): Rename to...
3220 (dr_vec_info): ...this and add "dr" and "stmt" fields.
3221 (_stmt_vec_info::dr_aux): Update accordingly.
3222 (_stmt_vec_info::data_ref_info): Delete.
3223 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
3224 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
3225 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
3226 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
3228 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
3229 (STMT_VINFO_DR_INFO): New macro.
3230 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
3231 (set_dr_misalignment): Update after rename of dataref_aux.
3232 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
3233 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
3234 initialization of STMT_VINFO_DATA_REF.
3235 * tree-vectorizer.c (vec_info::move_dr): New function.
3236 * tree-vect-patterns.c (vect_recog_bool_pattern)
3237 (vect_recog_mask_conversion_pattern)
3238 (vect_recog_gather_scatter_pattern): Use it.
3239 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
3240 the "dr" and "stmt" fields of dr_vec_info instead of
3241 STMT_VINFO_DATA_REF.
3243 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3245 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
3246 (is_pattern_stmt_p): Use it.
3247 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
3248 on pattern statements.
3250 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3252 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
3253 original stmt as a stmt_vec_info rather than a gimple stmt.
3254 (vect_pattern_recog_1): Take the statement directly as a
3255 stmt_vec_info, rather than via a gimple_stmt_iterator.
3256 Update call to vect_mark_pattern_stmts.
3257 (vect_pattern_recog): Update calls accordingly.
3259 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3261 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
3262 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
3263 a vect_def_type for the first argument.
3264 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
3265 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
3266 operand if it isn't defined by a vectorized statement.
3267 (vect_build_gather_load_calls): Remove the mask_dt argument and
3268 update calls to vect_get_vec_def_for_stmt_copy.
3269 (vectorizable_bswap): Likewise the dt argument.
3270 (vectorizable_call): Update calls to vectorizable_bswap and
3271 vect_get_vec_def_for_stmt_copy.
3272 (vectorizable_simd_clone_call, vectorizable_assignment)
3273 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
3274 (vectorizable_comparison): Update calls to
3275 vect_get_vec_def_for_stmt_copy.
3276 (vectorizable_store): Likewise. Remove now-unnecessary calls to
3278 (vect_get_loop_based_defs): Remove dt argument and update call
3279 to vect_get_vec_def_for_stmt_copy.
3280 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
3281 and vect_get_vec_def_for_stmt_copy.
3282 (vectorizable_load): Update calls to vect_build_gather_load_calls
3283 and vect_get_vec_def_for_stmt_copy.
3284 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3285 (vectorizable_reduction, vectorizable_live_operation): Update calls
3286 to vect_get_vec_def_for_stmt_copy.
3288 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3290 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
3291 and gimple stmt arguments with a stmt_vec_info.
3292 (vect_record_base_alignments): Update calls accordingly.
3293 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
3294 and gimple stmt arguments with a stmt_vec_info.
3295 (vect_build_slp_tree_1): Remove vinfo argument and update call
3296 to vect_record_max_nunits.
3297 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
3298 and vect_record_max_nunits.
3300 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3302 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
3303 file and take a stmt_vec_info instead of a gimple stmt.
3304 (supportable_widening_operation, vect_finish_replace_stmt)
3305 (vect_finish_stmt_generation, vect_get_store_rhs)
3306 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
3307 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
3308 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
3309 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
3310 (vect_create_data_ref_ptr, bump_vector_ptr)
3311 (vect_permute_store_chain, vect_setup_realignment)
3312 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
3313 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
3314 (vectorizable_reduction, vectorizable_induction)
3315 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
3316 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
3317 than gimple stmts as arguments.
3318 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
3319 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
3320 (can_group_stmts_p, vect_check_gather_scatter)
3321 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
3322 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
3323 (vect_permute_load_chain, vect_shift_permute_load_chain)
3324 (vect_transform_grouped_load)
3325 (vect_record_grouped_load_vectors): Likewise.
3326 * tree-vect-loop.c (vect_fixup_reduc_chain)
3327 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
3328 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
3329 (vectorizable_reduction, vectorizable_induction)
3330 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
3331 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
3332 (vect_get_load_store_mask): Likewise.
3333 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
3334 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
3335 * tree-vect-stmts.c (vect_mark_relevant)
3336 (is_simple_and_all_uses_invariant)
3337 (exist_non_indexing_operands_for_use_p, process_use)
3338 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
3339 (vect_get_vec_def_for_operand, vect_get_vec_defs)
3340 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
3341 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
3342 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
3343 (get_negative_load_store_type, get_load_store_type)
3344 (vect_check_load_store_mask, vect_check_store_rhs)
3345 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
3346 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3347 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3348 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3349 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
3350 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
3351 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
3352 (supportable_widening_operation): Likewise.
3354 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3356 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
3357 a stmt_vec_info instead of a gcall.
3358 (vect_check_gather_scatter): Update call accordingly.
3359 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
3361 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
3362 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
3363 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
3364 instead of a gimple stmt.
3365 (vect_transform_loop): Update calls accordingly.
3366 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
3367 stmt_vec_infos instead of gimple stmts.
3368 (vect_analyze_slp_instance): Update use accordingly.
3369 * tree-vect-stmts.c (read_vector_array, write_vector_array)
3370 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
3371 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
3372 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
3373 (vect_gen_widened_results_half, vect_get_loop_based_defs)
3374 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
3375 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
3376 down to subroutines.
3378 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3380 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
3381 of the worklist from a vector of gimple stmts to a vector of
3383 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3384 (vect_mark_stmts_to_be_vectorized): Likewise
3386 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3388 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
3389 statement before passing it to vect_analyze_stmt.
3390 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
3391 the chain of phi vector definitions. Track the exit phi via its
3393 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
3394 STMT_VINFO_REDUC_DEF.
3395 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
3396 stmt_vec_infos to handle the statement chains.
3397 (vect_get_slp_defs): Record the first statement in the node
3398 using a stmt_vec_info.
3399 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
3400 statements here and pass their stmt_vec_info down to subroutines.
3401 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
3402 down to vect_finish_stmt_generation.
3403 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
3404 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
3405 stmt_vec_infos to subroutines.
3406 (vect_remove_stores): Use stmt_vec_infos to handle the statement
3409 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3411 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
3412 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
3413 (vect_permute_store_chain, vect_setup_realignment)
3414 (vect_permute_load_chain, vect_shift_permute_load_chain)
3415 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
3416 stmts internally, and when passing values to other vectorizer routines.
3417 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3418 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
3419 (vect_analyze_loop_operations, get_initial_def_for_reduction)
3420 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
3421 (vectorizable_reduction, vectorizable_induction)
3422 (vectorizable_live_operation, vect_transform_loop_stmt)
3423 (vect_transform_loop): Likewise.
3424 * tree-vect-patterns.c (vect_reassociating_reduction_p)
3425 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
3426 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
3427 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3428 (vect_slp_analyze_node_operations_1): Likewise.
3429 * tree-vect-stmts.c (vect_mark_relevant, process_use)
3430 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
3431 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
3432 (vect_finish_stmt_generation_1, get_group_load_store_type)
3433 (get_load_store_type, vect_build_gather_load_calls)
3434 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3435 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3436 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3437 (vectorizable_store, vectorizable_load, vectorizable_condition)
3438 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
3439 (supportable_widening_operation): Likewise.
3440 (vect_get_vector_types_for_stmt): Likewise.
3441 * tree-vectorizer.h (vect_dr_behavior): Likewise.
3443 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3445 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3446 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
3447 (vect_permute_store_chain, vect_permute_load_chain)
3448 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
3449 repeated stmt_vec_info lookups.
3450 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
3451 (vect_update_ivs_after_vectorizer): Likewise.
3452 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
3453 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3454 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
3455 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3456 (vect_bb_slp_scalar_cost): Likewise.
3457 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
3459 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3461 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
3462 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
3463 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
3464 to get gassigns and gcalls, rather than operating on generc gimple
3466 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
3467 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
3468 (vectorizable_load, vect_analyze_stmt): Likewise.
3469 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
3471 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3473 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
3474 return stmt_vec_infos rather than gimple stmts. Do not accept
3476 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
3478 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3479 Update use of get_later_stmt.
3480 (vect_get_constant_vectors): Update call accordingly.
3481 (vect_schedule_slp_instance): Likewise
3482 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
3483 (vect_slp_analyze_instance_dependence): Likewise.
3484 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
3486 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3488 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
3489 (stmt_info_for_cost::stmt_info): ...this new field.
3490 (add_stmt_costs): Update accordingly.
3491 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3492 (vect_get_known_peeling_cost): Likewise.
3493 (vect_estimate_min_profitable_iters): Likewise.
3494 * tree-vect-stmts.c (record_stmt_cost): Likewise.
3496 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3498 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
3499 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3500 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
3502 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
3504 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3506 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
3507 a gimple stmt to a stmt_vec_info.
3508 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3510 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3512 * tree-vectorizer.h (vec_info::grouped_stores): Change from
3513 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3514 (_loop_vec_info::reduction_chains): Likewise.
3515 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
3517 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3519 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3521 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
3522 a gimple stmt to a stmt_vec_info.
3523 (_stmt_vec_info::next_element): Likewise.
3524 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
3525 (vect_slp_analyze_and_verify_node_alignment)
3526 (vect_analyze_group_access_1, vect_analyze_group_access)
3527 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
3528 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
3529 (vect_supportable_dr_alignment): Update accordingly.
3530 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
3531 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
3532 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
3533 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
3534 * tree-vect-slp.c (vect_build_slp_tree_1)
3535 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
3536 (vect_split_slp_store_group, vect_analyze_slp_instance)
3537 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
3538 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
3539 (get_group_load_store_type, get_load_store_type)
3540 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
3541 (vect_transform_stmt, vect_remove_stores): Likewise.
3543 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3545 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
3547 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3548 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
3549 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
3550 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
3551 (vector_alignment_reachable_p, vect_get_data_access_cost)
3552 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
3553 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
3554 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
3555 (vect_analyze_group_access_1, vect_analyze_group_access)
3556 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
3557 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
3558 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
3559 result of vect_dr_stmt and use the stmt_vec_info instead of
3560 the associated gimple stmt.
3561 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
3562 (vect_gen_prolog_loop_niters): Likewise.
3563 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
3565 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3567 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
3568 to a vec<stmt_vec_info>.
3569 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
3570 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
3572 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
3573 to a vec<stmt_vec_info>.
3574 (bst_traits::value_type, bst_traits::value_type): Likewise.
3575 (bst_traits::hash): Update accordingly.
3576 (vect_get_and_check_slp_defs): Change the stmts parameter from
3577 a vec<gimple *> to a vec<stmt_vec_info>.
3578 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
3579 (vect_build_slp_tree): Likewise.
3580 (vect_build_slp_tree_2): Likewise. Update uses of
3581 SLP_TREE_SCALAR_STMTS.
3582 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
3583 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
3584 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
3585 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
3586 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
3587 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
3588 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
3589 (vect_get_constant_vectors, vect_get_slp_defs)
3590 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
3591 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
3592 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
3593 instead of gimple stmts.
3594 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
3595 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
3596 (vect_slp_analyze_instance_dependence): Update uses of
3597 SLP_TREE_SCALAR_STMTS.
3598 (vect_slp_analyze_and_verify_node_alignment): Likewise.
3599 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
3600 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
3601 (get_initial_defs_for_reduction): Likewise.
3602 (vect_create_epilog_for_reduction): Likewise.
3603 (vectorize_fold_left_reduction): Likewise.
3604 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
3605 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
3606 (can_vectorize_live_stmts): Likewise.
3608 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3610 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
3611 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
3612 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
3614 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
3616 (gather_scalar_reductions): Update after above interface changes.
3617 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
3618 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
3620 (vect_force_simple_reduction): Likewise.
3621 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
3622 LOOP_VINFO_REDUCTIONS.
3623 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3625 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3627 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
3628 a gimple stmt to a stmt_vec_info.
3629 * tree-vect-loop.c (vect_active_double_reduction_p)
3630 (vect_force_simple_reduction, vectorizable_reduction): Update
3633 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3635 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
3636 vec<gimple *> to a vec<stmt_vec_info>.
3637 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
3638 the reduction_phis argument from a vec<gimple *> to a
3640 (vectorizable_reduction): Likewise the phis local variable that
3641 is passed to vect_create_epilog_for_reduction. Update for new type
3642 of SLP_TREE_VEC_STMTS.
3643 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
3644 (vectorizable_live_operation): Likewise.
3645 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
3646 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
3648 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3650 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
3651 a gimple stmt to a stmt_vec_info.
3652 (vectorizable_condition, vectorizable_live_operation)
3653 (vectorizable_reduction, vectorizable_induction): Pass back the
3654 vectorized statement as a stmt_vec_info.
3655 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
3656 use of STMT_VINFO_VEC_STMT.
3657 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
3658 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
3659 as stmt_vec_infos rather than gimple stmts.
3660 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
3662 (vectorizable_live_operation): Likewise.
3663 (vectorizable_reduction, vectorizable_induction): Likewise,
3664 updating use of STMT_VINFO_VEC_STMT.
3665 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
3666 of STMT_VINFO_VEC_STMT.
3667 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
3668 (vectorizable_simd_clone_call, vectorizable_conversion)
3669 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3670 (vectorizable_store, vectorizable_load, vectorizable_condition)
3671 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
3672 from a gimple stmt to a stmt_vec_info.
3673 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
3674 pointer to a stmt_vec_info to the vectorizable_* routines.
3676 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3678 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
3679 a gimple stmt to a stmt_vec_info.
3680 (is_pattern_stmt_p): Update accordingly.
3681 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
3682 (vect_record_grouped_load_vectors): Likewise.
3683 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
3684 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
3685 (vect_model_reduction_cost): Likewise.
3686 (vect_create_epilog_for_reduction): Likewise.
3687 (vectorizable_reduction, vectorizable_induction): Likewise.
3688 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3689 Return the stmt_vec_info for the pattern statement.
3690 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
3691 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
3692 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3693 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
3694 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
3695 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
3696 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
3697 (free_stmt_vec_info, vect_is_simple_use): Likewise.
3699 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3701 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
3702 (vect_finish_stmt_generation): Likewise.
3703 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3704 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
3705 (vect_build_gather_load_calls): Use the return value of the above
3706 functions instead of a separate call to vinfo_for_stmt. Use narrow
3707 scopes for the input gimple stmt and wider scopes for the associated
3708 stmt_vec_info. Use vec_info::lookup_def when setting these
3709 stmt_vec_infos from an SSA_NAME definition.
3710 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
3711 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
3712 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
3713 (vectorizable_store, vectorizable_load, vectorizable_condition)
3714 (vectorizable_comparison): Likewise.
3715 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
3716 (vectorizable_reduction): Likewise.
3718 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3720 * tree-vectorizer.h (vect_is_simple_use): Add an optional
3721 stmt_vec_info * parameter before the optional gimple **.
3722 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3723 (process_use, vect_get_vec_def_for_operand_1): Update callers.
3724 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
3725 * tree-vect-loop.c (vectorizable_reduction): Likewise.
3726 (vectorizable_live_operation): Likewise.
3727 * tree-vect-patterns.c (type_conversion_p): Likewise.
3728 (vect_look_through_possible_promotion): Likewise.
3729 (vect_recog_rotate_pattern): Likewise.
3730 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3732 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3734 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
3735 a typedef to a wrapper class.
3736 (NULL_STMT_VEC_INFO): New macro.
3737 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
3738 (stmt_vec_info::operator*): New function.
3739 (stmt_vec_info::operator gimple *): Likewise.
3740 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
3741 (add_stmt_costs): Likewise.
3742 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
3743 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
3744 (vect_get_known_peeling_cost): Likewise.
3745 (vect_estimate_min_profitable_iters): Likewise.
3746 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3747 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
3748 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
3749 (vectorizable_store, free_stmt_vec_infos): Likewise.
3750 (new_stmt_vec_info): Change return type of xcalloc to
3753 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3755 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
3756 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
3757 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
3758 a single_imm_use-based sequence.
3759 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
3761 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3763 * tree-vectorizer.h (vec_info::lookup_def): Declare.
3764 * tree-vectorizer.c (vec_info::lookup_def): New function.
3765 * tree-vect-patterns.c (vect_get_internal_def): Use it.
3766 (vect_widened_op_tree): Likewise.
3767 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3768 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
3769 (vectorizable_reduction): Likewise.
3770 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
3772 (vect_is_slp_reduction): Update calls accordingly. Use
3773 vec_info::lookup_def.
3774 (vect_is_simple_reduction): Likewise
3775 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
3777 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3779 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
3780 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
3781 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
3783 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
3784 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
3785 (vect_update_vf_for_slp, vect_analyze_loop_operations)
3786 (vect_is_slp_reduction, vectorizable_induction)
3787 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
3788 * tree-vect-patterns.c (vect_init_pattern_stmt):
3789 (vect_determine_min_output_precision_1, vect_determine_precisions)
3790 (vect_pattern_recog): Likewise.
3791 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
3792 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
3793 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
3794 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
3795 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
3796 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
3797 info field from a loop to a loop_vec_info.
3799 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3801 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
3802 (vec_info::add_stmt): Declare.
3803 * tree-vectorizer.c (vec_info::add_stmt): New function.
3804 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
3805 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
3806 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
3807 (vectorizable_induction): Likewise.
3808 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
3809 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3810 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
3811 (vectorizable_load): Likewise.
3812 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
3813 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
3814 (vect_recog_gather_scatter_pattern): Likewise.
3815 (append_pattern_def_seq): Likewise. Remove a check that is
3816 performed by add_stmt itself.
3818 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3820 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
3821 which make_ssa_name was called with new_stmt before new_stmt
3824 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3826 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
3828 (vect_is_slp_reduction): ...here...
3829 (vect_is_simple_reduction): ...and here. Remove repetition of tests
3830 that are already known to be false.
3832 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3834 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
3835 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
3836 STMT_VINFO_NUM_SLP_USES when it's true.
3837 (vect_free_slp_instance): Add a final_p parameter and pass it to
3839 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
3840 (vect_analyze_slp_instance): Likewise.
3841 (vect_slp_analyze_operations): Likewise.
3842 (vect_slp_analyze_bb_1): Likewise.
3843 * tree-vectorizer.c (vec_info): Likewise.
3844 * tree-vect-loop.c (vect_transform_loop): Likewise.
3846 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3848 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3849 function is not called for second and subsequent members of
3852 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3854 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
3855 cases for nested loops from here to ...
3856 (vect_create_epilog_for_reduction): ...here. Only call
3857 vect_is_simple_use for inner-loop reductions.
3859 2018-07-31 Martin Liska <mliska@suse.cz>
3861 PR gcov-profile/85338
3862 PR gcov-profile/85350
3863 PR gcov-profile/85372
3864 * profile.c (struct location_triplet): New.
3865 (struct location_triplet_hash): Likewise.
3866 (output_location): Do not output a BB that
3867 is already recorded for a line.
3868 (branch_prob): Use streamed_locations.
3870 2018-07-31 Martin Liska <mliska@suse.cz>
3872 PR gcov-profile/85370
3873 * coverage.c (coverage_begin_function): Do not mark target
3874 clones as artificial functions.
3876 2018-07-31 Martin Liska <mliska@suse.cz>
3878 PR gcov-profile/83813
3879 PR gcov-profile/84758
3880 PR gcov-profile/85217
3881 PR gcov-profile/85332
3882 * profile.c (branch_prob): Do not record GOTO expressions
3883 for GIMPLE statements which locations are already streamed.
3885 2018-07-31 Olivier Hainque <hainque@adacore.com>
3887 * gcc.c (handle_spec_function): Accept a soft_matched_part
3888 argument, as do_spec_1. Pass it down to ...
3889 (eval_spec_function): Accept a soft_matched_part argument,
3890 and pass it down to ...
3891 (do_spec_2): Accept a soft_matched_part argument, and pass
3892 it down to do_spec_1.
3893 (do_spec_1): Pass soft_matched_part to handle_spec_function.
3894 (handle_braces): Update call to handle_spec_function.
3895 (driver::set_up_specs): Update calls to do_spec_2.
3896 (compare_debug_dump_opt_spec_function): Likewise.
3897 (compare_debug_self_opt_spec_function): Likewise.
3899 2018-07-31 Olivier Hainque <hainque@adacore.com>
3901 * common.opt (nolibc): New option.
3902 * doc/invoke.texi (Link Options): Document it.
3903 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
3904 * config/alpha/linux.h: Likewise.
3905 * config/arc/elf.h: Likewise.
3906 * config/arm/uclinux-elf.h: Likewise.
3907 * config/arm/unknown-elf.h: Likewise.
3908 * config/avr/avrlibc.h: Likewise.
3909 * config/bfin/bfin.h: Likewise.
3910 * config/bfin/linux.h: Likewise.
3911 * config/bfin/uclinux.h: Likewise.
3912 * config/darwin.h: Likewise.
3913 * config/darwin10.h: Likewise.
3914 * config/darwin12.h: Likewise.
3915 * config/gnu-user.h: Likewise.
3916 * config/lm32/uclinux-elf.h: Likewise.
3917 * config/pa/pa-hpux11.h: Likewise.
3918 * config/pa/pa64-hpux.h: Likewise.
3919 * config/sparc/sparc.h: Likewise.
3921 2018-07-31 Olivier Hainque <hainque@adacore.com>
3923 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
3924 undefined variables.
3926 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3929 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
3932 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3934 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
3935 terminated string literal.
3937 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
3939 PR rtl-optimization/85160
3940 * combine.c (is_just_move): New function.
3941 (try_combine): Allow combining two instructions into two if neither of
3942 the original instructions was a move.
3944 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
3947 * doc/extend.texi (Global Register Variables): Discourage use of type
3949 (Local Register Variables): Likewise.
3951 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
3953 PR tree-optimization/86506
3954 * hwint.h (ceil_log2): Resync with hwint.c implementation.
3956 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
3959 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
3960 hard_regno, make sure no insn between `from` and `to` clobbers it.
3962 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
3963 Tom de Vries <tdevries@suse.de>
3965 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
3966 (PTX_DEFAULT_RUNTIME_DIM): ... this.
3967 (nvptx_goacc_validate_dims): Set default worker and gang dims to
3968 PTX_DEFAULT_RUNTIME_DIM.
3969 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
3971 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
3973 * config/pa/pa.c (pa_output_addr_vec): Align address table.
3974 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
3975 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
3977 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
3979 * config/rs6000/constraints.md (wG constraint): Delete, no longer
3981 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
3982 predicate to reflect toc fusion has been deleted.
3983 (toc_fusion_mem_raw): Delete, no longer used.
3984 (toc_fusion_mem_wrapped): Likewise.
3985 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
3987 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
3988 Delete, no longer used.
3989 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
3990 meant to be used for toc fusion.
3991 (rs6000_debug_print_mode): Delete toc fusion debugging.
3992 (rs6000_debug_reg_global): Likewise.
3993 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
3994 fusion and secondary reload support that were never used.
3995 (rs6000_option_override_internal): Delete TOC fusion, that was only
3996 partially defined, and it did not work unless you also used the
3998 (rs6000_legitimate_address_p): Delete TOC fusion support.
3999 (rs6000_opt_masks): Likewise.
4000 (fusion_wrap_memory_address): Delete function, no longer used.
4001 (fusion_split_address); Delete TOC fusion support.
4002 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
4003 longer used with toc fusion being deleted.
4004 (TARGET_TOC_FUSION_FP): Likewise.
4005 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
4007 (toc fusion spliter): Delete TOC fusion support.
4008 (toc_fusionload_<mode>): Likewise.
4009 (toc_fusionload_di): Likewise.
4010 (fusion_gpr_load_<mode>): Delete generator function, this insn no
4011 longer needs to be named. Rename predicate to delete TOC fusion.
4012 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4013 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4014 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
4015 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
4016 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
4018 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
4020 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
4021 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
4022 __int128 in built-in function prototypes.
4023 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
4024 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
4026 2018-07-27 Martin Sebor <msebor@redhat.com>
4028 PR tree-optimization/86696
4029 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
4030 types, including enums.
4031 (handle_char_store): Be prepared for the above function to fail.
4033 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
4035 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
4036 when optimization level is lower than 2 or optimize for size.
4038 2018-07-26 Martin Sebor <msebor@redhat.com>
4040 PR tree-optimization/86043
4041 PR tree-optimization/86042
4042 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
4043 (get_string_cst_length): Rename...
4044 (get_min_string_length): ...to this. Add argument.
4045 (handle_char_store): Extend to handle multi-character stores by
4047 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
4048 * tree.h (initializer_zerop): Add argument.
4050 2018-07-26 Jakub Jelinek <jakub@redhat.com>
4053 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
4054 declare target to variables if they have always,{to,from,tofrom} map
4057 2018-07-26 Martin Liska <mliska@suse.cz>
4060 * lto-wrapper.c: Add linker_output as prefix
4061 for ltrans_output_file.
4063 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
4065 PR rtl-optimization/85805
4066 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
4067 value for hard registers if that was written in the same mode.
4069 2018-07-26 Martin Liska <mliska@suse.cz>
4071 PR gcov-profile/86536
4072 * gcov.c (format_gcov): Use printf format %.*f directly
4073 and do not handle special values.
4075 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4077 * common/config/arc/arc-common.c (arc_option_optimization_table):
4078 Update default optimizations for size.
4080 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4082 * config/arc/arc.md (movsf_insn): Add short instruction selection.
4083 * config/arc/constraints.md (CfZ): New constraint.
4084 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
4085 (subsf3_fpu): Likewise.
4086 (cmpsf_fpu): Likewise.
4087 (cmpsf_fpu_uneq): Likewise.
4089 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4091 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
4093 (arc_is_uncached_mem_p): uncached applies to both the variable and
4096 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
4098 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
4101 2018-07-25 David Malcolm <dmalcolm@redhat.com>
4103 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
4104 field "m_scopes" from vec to auto_vec.
4106 2018-07-25 Martin Liska <mliska@suse.cz>
4108 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
4111 2018-07-25 Richard Biener <rguenther@suse.de>
4114 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
4115 special wrt context_die late.
4116 (gen_subprogram_die): Re-use DIEs in local scope.
4118 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
4120 PR tree-optimization/86644
4121 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
4123 2018-07-25 Martin Liska <mliska@suse.cz>
4126 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
4127 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
4129 2018-07-25 Martin Liska <mliska@suse.cz>
4132 * params.def: Explain ASan abbreviation and provide
4133 a documentation link.
4135 2018-07-24 Martin Sebor <msebor@redhat.com>
4137 PR tree-optimization/86622
4138 PR tree-optimization/86532
4139 * builtins.h (string_length): Declare.
4140 * builtins.c (c_strlen): Correct handling of non-constant offsets.
4141 (check_access): Be prepared for non-constant length ranges.
4142 (string_length): Make extern.
4143 * expr.c (string_constant): Only handle the minor non-constant
4144 array index. Use string_constant to compute the length of
4145 a generic string constant.
4147 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
4149 PR tree-optimization/86618
4150 * tree-vect-stmts.c (vectorizable_call): Don't take the address
4151 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
4153 2018-07-24 David Malcolm <dmalcolm@redhat.com>
4155 PR tree-optimization/86636
4156 * json.cc (json::object::set): Fix comment. Add assertions.
4157 (json::array::append): Move here from json.h. Add comment and an
4159 (json::string::string): Likewise.
4160 * json.h (json::array::append): Move to json.cc.
4161 (json::string::string): Likewise.
4162 * optinfo-emit-json.cc
4163 (optrecord_json_writer::impl_location_to_json): Assert that we
4164 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
4165 wrapper around it. Expand the location once, rather than three
4167 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
4168 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
4170 (optrecord_json_writer::optinfo_to_json): Likewise, in four
4171 places. Fix some overlong lines.
4173 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
4175 * config/aarch64/aarch64-simd.md
4176 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
4177 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
4178 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
4179 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
4180 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
4181 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
4182 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
4183 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
4184 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
4186 2018-07-24 Jakub Jelinek <jakub@redhat.com>
4189 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
4190 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
4191 and abs_d == d, do the power of two handling if profitable.
4193 2018-07-24 Richard Biener <rguenther@suse.de>
4195 * match.pd: Add BIT_FIELD_REF canonicalizations.
4197 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4200 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
4202 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4204 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
4205 terminated STRING_CST object.
4207 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
4209 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
4211 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4213 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
4214 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
4215 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
4216 the elements into a register.
4217 (rs6000_split_v4si_init_di_reg): Delete.
4218 (rs6000_split_v4si_init): Delete.
4219 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
4220 (vsx_init_v4si): Rewrite as a define_expand.
4222 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
4224 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
4225 zero_extend argument from memory): New.
4227 2018-07-22 Martin Sebor <msebor@redhat.com>
4230 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
4231 diagnosing calls with unknown arguments unless -Walloca-larger-than
4232 is restricted to less than PTRDIFF_MAX bytes.
4234 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
4236 * doc/gcov.texi (Invoking Gcov): Editorial changes.
4238 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4240 * pretty-print.c (text_info::set_location): Remove redundant
4241 "line_table" parameter from call to rich_location::set_range.
4243 2018-07-20 Martin Sebor <msebor@redhat.com>
4246 * builtins.c (expand_builtin_alloca): Adjust.
4247 * calls.c (alloc_max_size): Simplify.
4248 * cgraphunit.c (cgraph_node::expand): Adjust.
4249 * common.opt (larger_than_size, warn_frame_larger_than): Remove
4251 (frame_larger_than_size): Same.
4252 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
4253 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
4255 * doc/invoke.texi (GCC Command Options): Document option arguments.
4256 Explain byte-size arguments and suffixes.
4257 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
4258 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
4259 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
4260 * doc/options.texi (UInteger): Expand.
4261 (Host_Wide_Int, ByteSize): Document new properties.
4262 * final.c (final_start_function_1): Include sizes in an error message.
4263 * function.c (frame_offset_overflow): Same.
4264 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
4265 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
4266 Diagnose unbounded alloca calls only for limits of less than
4268 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
4269 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
4271 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
4272 only for limits of less than PTRDIFF_MAX.
4273 * langhooks-def.h (lhd_handle_option): Change function argument
4275 * langhooks.c (lhd_handle_option): Same.
4276 * langhooks.h (handle_option): Same.
4277 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
4279 (var_type, var_type_struct): Same.
4280 (var_set): Handle ByteSize flag.
4281 * optc-gen.awk: Add comments to output to ease debugging. Make
4282 use of HOST_WIDE_INT where appropriate.
4283 * opts-gen-save.awk: Use %lx to format unsigned long.
4284 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
4285 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
4286 arguments. Parse bytes-size suffixes.
4287 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
4288 (enum_value_to_arg): Same.
4289 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
4290 (handle_option): Adjust.
4291 (generate_option): Change function argument to HOST_WIDE_INT.
4292 (cmdline_handle_error): Adjust.
4293 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
4294 (set_option): Change function argument to HOST_WIDE_INT.
4295 (option_enabled): Handle cl_host_wide_int.
4296 (get_option_state): Handle CLVC_SIZE.
4297 (control_warning_option): Same.
4298 * opts.c (common_handle_option): Change function argument to
4299 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
4300 OPT_Wvla_larger_than_.
4301 * opts.h (enum cl_var_type): Add an enumerator.
4302 * stor-layout.c (layout_decl): Print a more meaningful warning.
4303 * toplev.c (output_stack_usage): Adjust.
4305 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
4307 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
4308 call to inline_expand_builtin_string_cmp.
4309 (expand_builtin_strcmp): Likewise.
4310 (expand_builtin_strncmp): Likewise.
4311 (inline_string_cmp): Delete the last parameter, change char_type_node
4312 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
4314 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
4315 the inlining expansion on target where the type of the call has same or
4316 narrower precision than unsigned char.
4318 2018-07-20 David Malcolm <dmalcolm@redhat.com>
4320 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
4321 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
4322 * common.opt (fsave-optimization-record): New option.
4323 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
4324 * doc/invoke.texi (-fsave-optimization-record): New option.
4325 * dumpfile.c: Include "optinfo-emit-json.h".
4326 (struct kv_pair): Move to coretypes.h.
4327 (optgroup_options): Make non-static.
4328 (dump_context::end_scope): Call
4329 optimization_records_maybe_pop_dump_scope.
4330 * dumpfile.h (optgroup_options): New decl.
4331 * json.cc: New file.
4333 * optinfo-emit-json.cc: New file.
4334 * optinfo-emit-json.h: New file.
4335 * optinfo.cc: Include "optinfo-emit-json.h".
4336 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
4337 (optinfo_enabled_p): Check optimization_records_enabled_p.
4338 (optinfo_wants_inlining_info_p): Likewise.
4339 * optinfo.h: Update comment.
4340 * profile-count.c (profile_quality_as_string): New function.
4341 * profile-count.h (profile_quality_as_string): New decl.
4342 (profile_count::quality): New accessor.
4343 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
4344 and optinfo_emit_json_cc_tests.
4345 * selftest.h (selftest::json_cc_tests): New decl.
4346 (selftest::optinfo_emit_json_cc_tests): New decl.
4347 * toplev.c: Include "optinfo-emit-json.h".
4348 (compile_file): Call optimization_records_finish.
4349 (do_compile): Call optimization_records_start.
4350 * tree-ssa-live.c: Include optinfo.h.
4351 (remove_unused_scope_block_p): Retain inlining information if
4352 optinfo_wants_inlining_info_p returns true.
4354 2018-07-20 Richard Biener <rguenther@suse.de>
4357 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
4358 to cover -flto-partition=none.
4360 2018-07-20 Martin Liska <mliska@suse.cz>
4362 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
4363 (get_decl_source_range): Remove unused function.
4365 2018-07-20 Richard Biener <rguenther@suse.de>
4367 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
4368 (struct vn_phi_s): Likewise.
4369 (struct vn_reference_s): Likewise.
4370 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
4371 for searching the slot of an entry known to be in the hash itself.
4372 (vn_phi_hasher::equal): Likewise.
4373 (vn_reference_hasher::equal): Likewise.
4374 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
4376 (optimistic_info, current_info): Remove, keeping only valid_info.
4377 (vn_reference_lookup_1): Remove fallback lookup.
4378 (vn_reference_lookup_2): Likewise.
4379 (vn_nary_op_lookup_1): Likewise.
4380 (vn_phi_lookup): Likewise.
4381 (vn_nary_build_or_lookup_1): Make sure to not chain the built
4383 (vn_reference_insert): Adjust, chain the inserted hash element
4384 at last_inserted_ref.
4385 (vn_reference_insert_pieces): Likewise.
4386 (visit_reference_op_call): Likewise.
4387 (vn_nary_op_insert_into): Chain the inserted hash element at
4389 (vn_nary_op_insert_pieces): Adjust.
4390 (vn_nary_op_insert): Likewise.
4391 (vn_nary_op_insert_stmt): Likewise.
4392 (vn_phi_insert): Adjust, chain the inserted hash element at
4394 (process_scc): Remove clearing and copying the optimistic
4395 table. Instead remove elements inserted during an optimistic
4396 iteration from the single table we maintain.
4397 (init_scc_vn): Adjust.
4398 (free_scc_vn): Likewise.
4399 (sccvn_dom_walker::record_cond): Likewise.
4400 (sccvn_dom_walker::after_dom_children): Likewise.
4402 2018-07-19 Martin Sebor <msebor@redhat.com>
4404 PR tree-optimization/84047
4405 PR tree-optimization/83776
4406 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
4407 (check_array_bounds): Call it.
4409 2018-07-19 Martin Sebor <msebor@redhat.com>
4411 * align.h (align_flags): Use member initialization.
4413 2018-07-19 David Malcolm <dmalcolm@redhat.com>
4415 * Makefile.in (OBJS): Add optinfo.o.
4416 * coretypes.h (class symtab_node): New forward decl.
4417 (struct cgraph_node): New forward decl.
4418 (class varpool_node): New forward decl.
4419 * dump-context.h: New file.
4420 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
4422 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
4423 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
4424 (set_alt_dump_file): Likewise.
4425 (dump_context::~dump_context): New dtor.
4426 (dump_gimple_stmt): Move implementation to...
4427 (dump_context::dump_gimple_stmt): ...this new member function.
4428 Add the stmt to any pending optinfo, creating one if need be.
4429 (dump_gimple_stmt_loc): Move implementation to...
4430 (dump_context::dump_gimple_stmt_loc): ...this new member function.
4431 Start a new optinfo and add the stmt to it.
4432 (dump_gimple_expr): Move implementation to...
4433 (dump_context::dump_gimple_expr): ...this new member function.
4434 Add the stmt to any pending optinfo, creating one if need be.
4435 (dump_gimple_expr_loc): Move implementation to...
4436 (dump_context::dump_gimple_expr_loc): ...this new member function.
4437 Start a new optinfo and add the stmt to it.
4438 (dump_generic_expr): Move implementation to...
4439 (dump_context::dump_generic_expr): ...this new member function.
4440 Add the tree to any pending optinfo, creating one if need be.
4441 (dump_generic_expr_loc): Move implementation to...
4442 (dump_context::dump_generic_expr_loc): ...this new member
4443 function. Add the tree to any pending optinfo, creating one if
4445 (dump_printf): Move implementation to...
4446 (dump_context::dump_printf_va): ...this new member function. Add
4447 the text to any pending optinfo, creating one if need be.
4448 (dump_printf_loc): Move implementation to...
4449 (dump_context::dump_printf_loc_va): ...this new member function.
4450 Start a new optinfo and add the stmt to it.
4451 (dump_dec): Move implementation to...
4452 (dump_context::dump_dec): ...this new member function. Add the
4453 value to any pending optinfo, creating one if need be.
4454 (dump_context::dump_symtab_node): New member function.
4455 (dump_context::get_scope_depth): New member function.
4456 (dump_context::begin_scope): New member function.
4457 (dump_context::end_scope): New member function.
4458 (dump_context::ensure_pending_optinfo): New member function.
4459 (dump_context::begin_next_optinfo): New member function.
4460 (dump_context::end_any_optinfo): New member function.
4461 (dump_context::s_current): New global.
4462 (dump_context::s_default): New global.
4463 (dump_scope_depth): Delete global.
4464 (dumpfile_ensure_any_optinfo_are_flushed): New function.
4465 (dump_symtab_node): New function.
4466 (get_dump_scope_depth): Reimplement in terms of dump_context.
4467 (dump_begin_scope): Likewise.
4468 (dump_end_scope): Likewise.
4469 (selftest::temp_dump_context::temp_dump_context): New ctor.
4470 (selftest::temp_dump_context::~temp_dump_context): New dtor.
4471 (selftest::verify_item): New function.
4472 (ASSERT_IS_TEXT): New macro.
4473 (ASSERT_IS_TREE): New macro.
4474 (ASSERT_IS_GIMPLE): New macro.
4475 (selftest::test_capture_of_dump_calls): New test.
4476 (selftest::dumpfile_c_tests): Call it.
4477 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
4478 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
4479 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
4480 descriptive comment.
4481 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
4482 (dump_node, dump_bb): Move these unrelated decls.
4483 (class dump_manager): Add leading comment.
4484 * optinfo.cc: New file.
4485 * optinfo.h: New file.
4487 2018-07-19 Michael Collison <michael.collison@arm.com>
4488 Richard Henderson <rth@redhat.com>
4490 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
4491 (subti): Handle op1 zero.
4492 (subvti4, usub4ti4): New.
4493 (*sub<GPI>3_compare1_imm): New.
4494 (sub<GPI>3_carryinCV): New.
4495 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
4496 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
4498 2018-07-19 Michael Collison <michael.collison@arm.com>
4499 Richard Henderson <rth@redhat.com>
4501 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
4502 (addti3): Create simpler code if low part is already known to be 0.
4503 (addvti4, uaddvti4): New.
4504 (*add<GPI>3_compareC_cconly_imm): New.
4505 (*add<GPI>3_compareC_cconly): New.
4506 (*add<GPI>3_compareC_imm): New.
4507 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
4508 handle constants within this pattern..
4509 (*add<GPI>3_compareV_cconly_imm): New.
4510 (*add<GPI>3_compareV_cconly): New.
4511 (*add<GPI>3_compareV_imm): New.
4512 (add<GPI>3_compareV): New.
4513 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
4514 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
4515 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
4516 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
4517 with 'comparison' operator.
4518 (*add<GPI>3_compareV_cconly_imm): Ditto.
4519 (*add<GPI>3_compareV_cconly): Ditto.
4520 (*add<GPI>3_compareV_imm): Ditto.
4521 (add<GPI>3_compareV): Ditto.
4522 (add<mode>3_carryinC): Ditto.
4523 (*add<mode>3_carryinC_zero): Ditto.
4524 (*add<mode>3_carryinC): Ditto.
4525 (add<mode>3_carryinV): Ditto.
4526 (*add<mode>3_carryinV_zero): Ditto.
4527 (*add<mode>3_carryinV): Ditto.
4529 2018-07-19 Michael Collison <michael.collison@arm.com>
4530 Richard Henderson <rth@redhat.com>
4532 * config/aarch64/aarch64-modes.def (CC_V): New.
4533 * config/aarch64/aarch64-protos.h
4534 (aarch64_addti_scratch_regs): Declare
4535 (aarch64_subvti_scratch_regs): Declare.
4536 (aarch64_expand_subvti): Declare.
4537 (aarch64_gen_unlikely_cbranch): Declare
4538 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
4539 for signed overflow using CC_Vmode.
4540 (aarch64_get_condition_code_1): Handle CC_Vmode.
4541 (aarch64_gen_unlikely_cbranch): New function.
4542 (aarch64_addti_scratch_regs): New function.
4543 (aarch64_subvti_scratch_regs): New function.
4544 (aarch64_expand_subvti): New function.
4546 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4548 * config/aarch64/aarch64-option-extensions.def: New entry for profile
4550 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
4551 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
4554 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4557 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
4558 address check not strict.
4560 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
4562 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
4564 (store_pair_lanes<mode>): Likewise.
4565 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
4566 enum value 'ADDR_QUERY_LDP_STP_N'.
4567 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
4568 (aarch64_print_address_internal): Add declaration.
4569 (aarch64_print_ldpstp_address): Remove.
4570 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
4571 (aarch64_print_operand): Change printing of 'y'.
4572 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
4573 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
4574 'true' rather than '1'.
4575 * config/aarch64/constraints.md (Uml): Likewise.
4576 (Uml): Rename to Umn.
4579 2018-07-19 Richard Biener <rguenther@suse.de>
4581 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
4583 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
4584 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
4585 (vn_reference_hasher): Likewise.
4586 (struct vn_tables_s): Remove obstack and alloc-pool members.
4587 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
4588 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
4589 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
4590 (vn_reference_insert_pieces): Likewise.
4591 (alloc_vn_nary_op_noinit): Adjust.
4592 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
4593 (vn_phi_eq): Adjust.
4594 (shared_lookup_phiargs): Remove.
4595 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
4596 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
4597 (visit_reference_op_call): Likewise.
4598 (copy_nary, copy_phi, copy_reference): Remove.
4599 (process_scc): Rewind the obstack when iterating. Do not
4600 copy the elements to valid_info but just move them from one
4601 hashtable to the other.
4602 (allocate_vn_table): Adjust.
4603 (free_vn_table): Likewise.
4604 (init_scc_vn): Likewise.
4605 (free_scc_vn): Likewise.
4607 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
4610 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
4611 indirect_return as function type attribute.
4612 (ix86_attribute_table): Change indirect_return to function
4614 * doc/extend.texi: Update indirect_return attribute.
4616 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
4618 * wide-int.h (widest2_int): New.
4619 * gimple-fold.c (arith_overflowed_p): Use it.
4620 * tree.h (widest2_int_cst): New.
4621 * tree-vrp.c (wide_int_binop_overflow): Rename from
4622 vrp_int_const_binop.
4623 Rewrite to work on trees.
4624 (extract_range_from_multiplicative_op_1): Abstract code to...
4625 (wide_int_range_min_max): ...here.
4626 (wide_int_range_cross_product): ...and here.
4627 (extract_range_from_binary_expr_1): Abstract overflow code to...
4628 (wide_int_range_mult_wrapping): ...here.
4629 * tree-vrp.h (wide_int_range_cross_product): New.
4630 (wide_int_range_mult_wrapping): New.
4632 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
4633 Julia Koval <julia.koval@intel.com>
4635 * config/i386/x86-tune-costs.h (skylake_memcpy,
4636 skylake_memset): Replace rep_prefix with unrolling for size 512.
4638 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
4641 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
4642 comparision with EQ_EXPR in last stmt.
4644 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
4646 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
4647 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
4648 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
4649 previously known as "PowerPC AltiVec Built-in Functions". Move
4650 some material to new subsubsections "PowerPC AltiVec Built-in
4651 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
4653 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
4654 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
4655 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
4656 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
4658 2018-07-18 Richard Biener <rguenther@suse.de>
4660 PR tree-optimization/86557
4661 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
4664 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
4666 * config/s390/s390.c (s390_function_profiler): Generate CFI.
4668 2018-07-17 Jeff Law <law@redhat.com>
4670 * config/arm/arm.c (get_label_padding): Update for recent
4671 changes to label_to_alignment.
4673 PR tree-optimization/86010
4674 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
4676 * config/mips/mips.c (vr4130_align_insns): Update for recent
4677 changes to label_to_alignment.
4679 * config/frv/frv.c (frv_label_align): Update for recent changes
4680 to label_to_alignment.
4682 * config/nios2/nios2.c (nios2_label_align): Update for recent
4683 changes which dropped ALIGN_LABELS_LOG.
4685 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
4687 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
4688 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
4690 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
4692 * config/arc/arc.c (arc_label_align): Use align_labels instead of
4693 deprecated align_labels_log.
4695 2018-07-17 Richard Biener <rguenther@suse.de>
4698 * dwarf2out.c (init_sections_and_labels): Always generate
4699 a debug_line_str_section for early LTO debug.
4700 (dwarf2out_finish): Reset debug_line_str_hash output early.
4701 Bump counter for extra dwarf5 .debug_loc labels to not conflict
4703 (dwarf2out_early_finish): Output debug_line_str.
4705 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4707 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
4708 index register on z196 or later.
4710 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
4712 * config/s390/s390.c (s390_default_align): Set default function
4714 (s390_override_options_after_change): Call s390_default align.
4715 (s390_option_override_internal): Call s390_default_align.
4716 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
4718 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4721 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
4722 to _looptemp_ clauses, other than the first two.
4724 2018-07-17 Martin Liska <mliska@suse.cz>
4726 * opts.c: Do not enable OPT_falign_* for -Os.
4728 2018-07-17 Martin Liska <mliska@suse.cz>
4730 * align.h (MAX_CODE_ALIGN): New.
4731 (MAX_CODE_ALIGN_VALUE): New.
4732 * common/config/i386/i386-common.c (ix86_handle_option):
4733 (MAX_CODE_ALIGN): Moved to align.h.
4734 * final.c (MAX_CODE_ALIGN): Likewise.
4735 * opts.c (parse_and_check_align_values):
4736 (MAX_CODE_ALIGN): Likewise.
4737 (MAX_CODE_ALIGN_VALUE): Likewise.
4739 2018-07-17 Martin Liska <mliska@suse.cz>
4741 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
4742 in order to fulfil coding style.
4743 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
4744 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4745 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4746 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
4747 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
4748 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
4749 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
4750 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
4751 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4753 2018-07-17 Martin Liska <mliska@suse.cz>
4755 * align.h: New file.
4756 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
4758 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
4759 align_flags of label_to_alignment.
4760 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
4762 * config/m68k/m68k.c: Do not use removed align_labels_value and
4764 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
4765 (LOOP_ALIGN): Likewise.
4766 (LABEL_ALIGN): Likewise.
4767 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4768 Remove not used macro.
4769 (rs6000_loop_align): Change return type to align_flags.
4770 (rs6000_loop_align_max_skip): Remove.
4771 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
4772 Change return type to align_flags.
4773 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
4774 Remove not used macro.
4775 (rs6000_loop_align): Change return type to align_flags.
4776 (rs6000_loop_align_max_skip): Remove.
4777 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
4778 * config/rx/rx-protos.h (rx_align_for_label): Make it
4780 * config/rx/rx.c (rx_align_for_label): Change return type
4782 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
4784 into align_flags class.
4785 (LABEL_ALIGN): Likewise.
4786 (LOOP_ALIGN): Likewise.
4787 * config/s390/s390.c (s390_label_align): Use align_flags
4789 (s390_asm_output_function_label): Likewise.
4790 * config/sh/sh.c (sh_override_options_after_change):
4791 Use align_flags class directly without macros.
4792 (find_barrier): Likewise.
4793 (barrier_align): Likewise.
4794 (sh_loop_align): Likewise.
4795 * config/spu/spu.c (spu_option_override):
4796 Use align_flags_tuple::get_value instead of removed macros.
4797 (spu_sched_init): Likewise.
4798 * config/spu/spu.h (GTY): Likewise.
4799 * config/visium/visium.c (visium_option_override):
4800 Set "8" as default secondary alignment.
4801 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
4802 in order to guarantee secondary alignment of 8.
4803 * coretypes.h: Include align.h header file.
4804 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
4805 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
4806 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
4807 * doc/tm.texi.in: Likewise.
4808 * final.c (struct label_alignment): Remove not used structure.
4809 (LABEL_ALIGN): Change type to align_flags.
4810 (LOOP_ALIGN): Likewise.
4811 (JUMP_ALIGN): Likewise.
4812 (default_loop_align_max_skip): Remove.
4813 (default_label_align_max_skip): Likewise.
4814 (default_jump_align_max_skip): Likewise.
4815 (default_label_align_after_barrier_max_skip):
4816 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
4817 (LABEL_TO_MAX_SKIP): Remove.
4818 (label_to_alignment): Return align_flags type instead of integer.
4819 (label_to_max_skip): Remove.
4820 (align_fuzz): Use align_flags type.
4821 (compute_alignments): Use align_flags type and use align_flags::max
4822 to combine multiple alignments.
4823 (grow_label_align): Grow vec instead of C array.
4824 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
4825 (shorten_branches): Use align_flags type and use align_flags::max
4826 to combine multiple alignments.
4827 (final_scan_insn_1): Remove usage of secondary alignment that comes
4828 from label alignment, but instead use proper secondary alignment
4829 which is computed in grow_label_align.
4830 * flags.h (struct align_flags_tuple): Move to align.h.
4831 (struct align_flags): Likewise.
4832 (state_align_loops): Rename to align_loops.
4833 (state_align_jumps): Rename to align_jumps.
4834 (state_align_labels): Rename to align_labels.
4835 (state_align_functions): Rename to align_functions.
4836 (align_loops_log): Remove.
4837 (align_jumps_log): Remove.
4838 (align_labels_log): Remove.
4839 (align_functions_log): Remove.
4840 (align_loops_max_skip): Remove.
4841 (align_jumps_max_skip): Remove.
4842 (align_labels_max_skip): Remove.
4843 (align_functions_max_skip): Remove.
4844 (align_loops_value): Remove.
4845 (align_jumps_value): Remove.
4846 (align_labels_value): Remove.
4847 (align_functions_value): Remove.
4848 * output.h (label_to_alignment): Change return type to align_flags.
4849 (label_to_max_skip): Remove.
4850 * target.def: Remove loop_align_max_skip, label_align_max_skip,
4851 jump_align_max_skip macros.
4852 * targhooks.h (default_loop_align_max_skip): Remove.
4853 (default_label_align_max_skip): Likewise.
4854 (default_jump_align_max_skip): Likewise.
4855 (default_label_align_after_barrier_max_skip): Remove.
4856 * toplev.c (read_log_maxskip): Use ::normalize function.
4857 (parse_N_M): Remove not used argument and also call ::normalize.
4858 (parse_alignment_opts): Do not pass unused arguments.
4859 * varasm.c (assemble_start_function): Use directly align_functions
4860 instead of removed macros.
4861 * system.h: Do not poison removed macros.
4863 2018-07-17 Jakub Jelinek <jakub@redhat.com>
4866 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
4867 and cond temporaries don't have reference type if iterator has
4868 pointer type. For init use &for_pre_body instead of pre_p if
4869 for_pre_body is non-empty.
4871 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4873 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
4874 double-double modes to SFmode directly directly.
4875 (trunc<mode>sf2_fprs): Delete.
4877 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4879 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
4880 for conversions between IFmode and the decimal floating point modes.
4881 (init_float128_ieee): Use the correct names for conversions between
4882 KFmode and the decimal floating point modes.
4884 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
4886 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
4887 for the conversions between TDmode and IFmode.
4888 (init_float128_ieee): Use more correct names for the conversions
4889 between TDmode and KFmode.
4891 2018-07-16 Jakub Jelinek <jakub@redhat.com>
4893 PR tree-optimization/86526
4894 * builtins.c (expand_builtin_memcmp): Formatting fixes.
4895 (inline_expand_builtin_string_cmp): Likewise.
4896 (inline_string_cmp): Likewise. Use c_readstr instead of
4897 builtin_memcpy_read_str. Add unit_mode temporary.
4899 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4902 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
4903 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
4905 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
4907 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
4908 Alphabetize prototypes of built-in functions, separating out
4909 built-in functions that are listed in this section but should be
4910 described elsewhere.
4912 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
4915 * expmed.c (emit_store_flag): Do not emit setcc followed by a
4916 conditional move when trapping comparison was split to a
4917 non-trapping one (and vice versa).
4919 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4921 * config/s390/s390.c (s390_function_profiler): Generate nops
4922 instead of profiler call sequences.
4923 * config/s390/s390.opt: Add the new option.
4925 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4927 * config/s390/s390.c (s390_function_profiler): Generate
4928 __mcount_loc section.
4929 * config/s390/s390.opt: Add the new option.
4931 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
4933 * common.opt: Add the new warning.
4934 * config/s390/s390.c (s390_function_profiler): Emit "brasl
4935 %r0,__fentry__" when -mfentry is specified.
4936 (s390_option_override_internal): Disallow -mfentry for 31-bit
4938 * config/s390/s390.opt: Add the new option.
4940 2018-07-16 Richard Biener <rguenther@suse.de>
4943 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
4944 for function-local FUNCTION_DECL and RESULT_DECL immediately.
4946 2018-07-16 Martin Liska <mliska@suse.cz>
4949 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
4952 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
4954 * config/arc/arcHS.md: Update ARCHS scheduling rules.
4956 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
4958 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
4960 * config/arc/arc-cpus.def (hs4x): New cpu.
4962 * config/arc/arc-tables.opt: Regenerate.
4963 * config/arc/arc.c (arc_sched_issue_rate): New function.
4964 (TARGET_SCHED_ISSUE_RATE): Define.
4965 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
4966 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
4968 (attr tune): Add ARCHS4x tune values.
4969 (attr tune_dspmpy): Define.
4970 (*tst): Correct instruction type.
4971 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
4972 * config/arc/arcHS4x.md: New file.
4973 * config/arc/fpu.md: Update instruction type attributes.
4974 * config/arc/t-multilib: Regenerate.
4976 2018-07-16 Tom de Vries <tdevries@suse.de>
4979 * var-tracking.c (vt_initialize): Fix pre_dec handling.
4981 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4983 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
4986 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
4988 PR tree-optimization/86514
4989 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
4990 conversion to a boolean type from a type with greater precision.
4992 2018-07-16 Tom de Vries <tdevries@suse.de>
4994 * var-tracking.c (vt_initialize): Print adjusted insn slim if
4995 dump_flags request TDF_SLIM.
4997 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
4999 * fold-const.c (int_const_binop_1): Abstract...
5000 (wide_int_binop): ...wide int code here.
5001 (poly_int_binop): ...poly int code here.
5002 Abstract the rest of int_const_binop_1 into int_const_binop.
5003 * fold-const.h (wide_int_binop): New.
5004 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
5005 Remove useless PLUS/MINUS_EXPR case.
5006 (zero_nonzero_bits_from_vr): Move wide int code...
5007 (zero_nonzero_bits_from_bounds): ...here.
5008 (extract_range_from_binary_expr_1): Move mask optimization code...
5009 (range_easy_mask_min_max): ...here.
5010 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
5011 (range_easy_mask_min_max): New.
5013 2018-07-15 Jeff Law <law@redhat.com>
5016 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
5019 2018-07-14 Jim Wilson <jimw@sifive.com>
5021 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
5023 2018-07-14 Paul Koning <ni1d@arrl.net>
5025 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
5027 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
5029 * lto-streamer-out.c (copy_function_or_variable): Dump info about
5032 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
5033 Steve Munroe <munroesj52@gmail.com>
5035 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
5036 (_mm_andnot_si128): Likewise.
5037 (_mm_or_si128): Likewise.
5038 (_mm_xor_si128): Likewise.
5040 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
5043 * builtins.c (expand_builtin_memcmp): Inline the calls first
5044 when result_eq is false.
5045 (expand_builtin_strcmp): Inline the calls first.
5046 (expand_builtin_strncmp): Likewise.
5047 (inline_string_cmp): New routine. Expand a string compare
5048 call by using a sequence of char comparison.
5049 (inline_expand_builtin_string_cmp): New routine. Inline expansion
5050 a call to str(n)cmp/memcmp.
5051 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
5053 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
5055 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
5057 * config/arm/driver-arm.c: Include arm-native.h.
5058 (host_detect_local_cpu): Use auto-generated data tables.
5059 (vendors, arm_cpu_table): Delete. Move part information to ...
5060 * config/arm/arm-cpus.in: ... here.
5061 * config/arm/parsecpu.awk (gen_native): New function.
5062 (vendor, part): New CPU fields.
5063 (END): Add support for building the native CPU detection tables.
5064 * config/arm/t-arm (arm-native.h): Add build rule.
5065 (driver-arm.o): Add dependency on arm-native.h.
5067 2018-07-13 Richard Biener <rguenther@suse.de>
5070 * match.pd (addr1 - addr2): Allow either of the operand to
5073 2018-07-13 Tom de Vries <tdevries@suse.de>
5075 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
5078 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
5080 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
5081 arrays instead of numbered variables.
5083 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
5085 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
5086 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
5088 2018-07-13 Richard Biener <rguenther@suse.de>
5091 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
5092 instead of get_context_die.
5094 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
5095 Richard Biener <rguenther@suse.de>
5098 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
5099 that the loop latch destination where phi is defined.
5101 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
5103 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
5104 (riscv_expand_epilogue): Add assertion to check interrupt mode.
5105 (riscv_set_current_function): Extract getting interrupt type to new
5107 (riscv_get_interrupt_type): New function.
5108 (riscv_merge_decl_attributes): New function, checking interrupt type is
5110 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
5112 2018-07-12 Paul Koning <ni1d@arrl.net>
5114 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
5117 2018-07-12 Paul Koning <ni1d@arrl.net>
5119 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
5120 zero reference, add doloop_end instead.
5121 * doc/md.texi (decrement_and_branch_until_zero): Remove.
5122 (Looping patterns): Remove decrement_and_branch_until_zero. Add
5123 detail for doloop_end.
5125 2018-07-12 Martin Sebor <msebor@redhat.com>
5128 * attribs.c (decl_attributes): Reject conflicting attributes before
5129 calling attribute handlers.
5131 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5133 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
5135 (gcc::dump_manager::get_dump_file_name): likewise.
5136 (dump_begin): Likewise.
5137 * dumpfile.h (dump_begin): Update prototype.
5138 (gcc::dump_manager::get_dump_file_name,
5139 gcc::dump_manager::get_dump_file_name): Update prototype.
5141 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5143 * internal-fn.h (vectorizable_internal_fn_p): New function.
5144 * tree-vect-slp.c (compatible_calls_p): Likewise.
5145 (vect_build_slp_tree_1): Remove nops argument. Handle calls
5146 to internal functions.
5147 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
5149 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5151 * fold-const.h (inverse_conditions_p): Declare.
5152 * fold-const.c (inverse_conditions_p): New function.
5153 * match.pd: Use inverse_conditions_p. Add folds of view_converts
5154 that test the inverse condition of a conditional internal function.
5155 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
5156 * internal-fn.c (internal_fn_mask_index): Handle conditional
5158 (vectorized_internal_fn_supported_p): New function.
5159 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
5160 (any_pred_load_store): Replace with...
5161 (need_to_predicate): ...this new variable.
5162 (redundant_ssa_names): New variable.
5163 (ifcvt_can_use_mask_load_store): Move initial checks to...
5164 (ifcvt_can_predicate): ...this new function. Handle tree codes
5165 for which a conditional internal function exists.
5166 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
5167 instead of ifcvt_can_use_mask_load_store. Update after variable
5169 (predicate_load_or_store): New function, split out from
5170 predicate_mem_writes.
5171 (check_redundant_cond_expr): New function.
5172 (value_available_p): Likewise.
5173 (predicate_rhs_code): Likewise.
5174 (predicate_mem_writes): Rename to...
5175 (predicate_statements): ...this. Use predicate_load_or_store
5176 and predicate_rhs_code.
5177 (combine_blocks, tree_if_conversion): Update after above name changes.
5178 (ifcvt_local_dce): Handle redundant_ssa_names.
5179 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
5180 general conditional functions.
5181 * tree-vect-stmts.c (vectorizable_call): Likewise.
5183 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5184 Alan Hayward <alan.hayward@arm.com>
5185 David Sherwood <david.sherwood@arm.com>
5187 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
5188 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
5189 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
5190 plus and minus and convert them into IFN_COND_FMA-based sequences.
5191 (convert_mult_to_fma): Handle conditional plus and minus.
5193 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5195 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
5196 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
5197 (cond_fnms_optab): New optabs.
5198 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
5200 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
5201 * internal-fn.h (get_conditional_internal_fn): Declare.
5202 (get_unconditional_internal_fn): Likewise.
5203 * internal-fn.c (cond_ternary_direct): New macro.
5204 (expand_cond_ternary_optab_fn): Likewise.
5205 (direct_cond_ternary_optab_supported_p): Likewise.
5206 (FOR_EACH_COND_FN_PAIR): Likewise.
5207 (get_conditional_internal_fn): New function.
5208 (get_unconditional_internal_fn): Likewise.
5209 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
5210 (gimple_match_op::gimple_match_op): Add a new overload for 5
5212 (gimple_match_op::set_op): Likewise.
5213 (gimple_resimplify5): Declare.
5214 * genmatch.c (decision_tree::gen): Generate simplifications for
5216 * gimple-match-head.c (gimple_simplify): Define an overload for
5217 5 operands. Handle calls with 5 arguments in the top-level overload.
5218 (convert_conditional_op): Handle conversions from unconditional
5219 internal functions to conditional ones.
5220 (gimple_resimplify5): New function.
5221 (build_call_internal): Pass a fifth operand.
5222 (maybe_push_res_to_seq): Likewise.
5223 (try_conditional_simplification): Try converting conditional
5224 internal functions to unconditional internal functions.
5225 Handle 3-operand unconditional forms.
5226 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
5227 Define ternary equivalents of the current rules for binary conditional
5229 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
5231 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
5232 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
5233 (optab): Handle them.
5234 (SVE_COND_FP_TERNARY): New int iterator.
5235 (sve_fmla_op, sve_fmad_op): New int attributes.
5236 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
5237 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
5238 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
5240 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
5242 * target.def (preferred_else_value): New target hook.
5243 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
5244 * doc/tm.texi: Regenerate.
5245 * targhooks.h (default_preferred_else_value): Declare.
5246 * targhooks.c (default_preferred_else_value): New function.
5247 * internal-fn.h (conditional_internal_fn_code): Declare.
5248 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
5249 (get_conditional_internal_fn): Use it.
5250 (conditional_internal_fn_code): New function.
5251 * gimple-match.h (gimple_match_cond): New struct.
5252 (gimple_match_op): Add a cond member function.
5253 (gimple_match_op::gimple_match_op): Update all forms to take a
5255 * genmatch.c (expr::gen_transform): Use the same condition as res_op
5256 for the suboperation, but don't specify a particular else_value.
5257 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
5258 (visit_nary_op, visit_reference_op_load): Pass
5259 gimple_match_cond::UNCOND to the gimple_match_op constructor.
5260 * gimple-match-head.c: Include tree-eh.h
5261 (convert_conditional_op): New function.
5262 (maybe_resimplify_conditional_op): Likewise.
5263 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
5264 (gimple_resimplify2): Likewise.
5265 (gimple_resimplify3): Likewise.
5266 (gimple_resimplify4): Likewise.
5267 (maybe_push_res_to_seq): Return null for conditional operations.
5268 (try_conditional_simplification): New function.
5269 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
5271 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
5273 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
5275 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
5277 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
5279 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
5281 (hash_tree): Do not hash DECL_FCONTEXT
5282 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
5283 Do not stream DECL_FCONTEXT.
5284 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
5285 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
5287 2018-07-12 Richard Biener <rguenther@suse.de>
5290 * dwarf2out.c (gen_block_die): Only output blocks when they have
5291 at least one !DECL_IGNORED_P variable.
5293 2018-07-12 Richard Biener <rguenther@suse.de>
5296 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
5297 Remove -mieee-fp handling.
5299 2018-07-12 Richard Biener <rguenther@suse.de>
5301 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
5302 left-over from last patch.
5304 2018-07-12 Jakub Jelinek <jakub@redhat.com>
5306 PR tree-optimization/86492
5307 * gimple-ssa-store-merging.c
5308 (imm_store_chain_info::coalesce_immediate_stores): Call
5309 check_no_overlap even for the merge_overlapping case. Formatting fix.
5311 2018-07-12 Richard Biener <rguenther@suse.de>
5314 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
5315 move possibly trapping operations into the conditional.
5317 2018-07-12 Richard Biener <rguenther@suse.de>
5319 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
5320 (vn_lookup_simplify_result): Remove recursion limit applied
5322 (vn_nary_build_or_lookup_1): Adjust.
5323 (try_to_simplify): Likewise.
5324 * gimple-match-head.c (gimple_resimplify1): Instead apply one
5326 (gimple_resimplify2): Likewise.
5327 (gimple_resimplify3): Likewise.
5328 (gimple_resimplify4): Likewise.
5330 2018-07-11 Jakub Jelinek <jakub@redhat.com>
5332 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
5333 Use __mmask64 type instead of __mmask8 for __M argument.
5334 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
5335 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
5337 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
5338 __mmask16 for __M argument.
5339 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
5340 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
5341 to __mmask16 instead of __mmask8.
5342 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
5343 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
5344 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
5345 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
5346 instead of __mmask16 for __U argument.
5347 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
5348 __mmask16 instead of __mmask8 for __U argument.
5349 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
5351 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
5353 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
5355 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
5357 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
5359 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
5361 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
5363 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
5365 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
5367 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
5369 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
5370 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
5371 return type as well as __M argument type and all casts from __mmask8
5373 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
5374 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
5375 return type as well as __M argument type and all casts from __mmask8
5377 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
5378 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
5379 return type as well as __M argument type and all casts from __mmask8
5381 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
5382 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
5383 return type as well as __M argument type and all casts from __mmask8
5385 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
5386 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
5389 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
5391 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
5392 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
5395 2018-07-11 Paul Koning <ni1d@arrl.net>
5397 * doc/md.texi (define_subst): Document how multiple occurrences of
5398 the same argument in the replacement pattern are handled.
5400 2018-07-11 Paul Koning <ni1d@arrl.net>
5402 * doc/extend.texi (Common Variable Attributes): Move "mode" into
5404 (Common Type Attributes): Add "mode" attribute.
5406 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
5408 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
5409 stream DECL_ORIGINAL_TYPE.
5410 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
5411 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
5412 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
5413 Do not walk original type.
5414 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
5416 (write_ts_decl_non_common_tree_pointers): Do not stream
5418 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
5419 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
5421 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5423 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
5424 thread twice from the same starting edge.
5426 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
5428 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
5429 * gimple.c (gimple_call_nonnull_result_p): ...here...
5430 (gimple_call_nonnull_arg): ...and here.
5431 * gimple.h (gimple_call_nonnull_result_p): New.
5432 (gimple_call_nonnull_arg): New.
5434 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
5436 * config/arm/arm-cpus.in: Move information from fpu field of each
5437 cpu definition to the isa field.
5438 * config/arm/parsecpu.awk (fpu): Delete match rule.
5439 (gen_comm_data): Don't add bits from the CPU's FPU entry.
5441 2018-07-11 Richard Biener <rguenther@suse.de>
5444 * dwarf2out.c (init_sections_and_labels): Use
5445 output_asm_line_debug_info consistently.
5446 (dwarf2out_early_finish): Likewise.
5447 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
5450 2018-07-11 Richard Biener <rguenther@suse.de>
5452 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
5453 Rework father_bb setting in a way to avoid propagating constants
5454 multiple times on a loop body.
5456 2018-07-10 Mark Wielaard <mark@klomp.org>
5459 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
5461 2018-07-10 Richard Biener <rguenther@suse.de>
5463 * hash-map.h (hash_map::iterator::operator*): Return
5464 references to key and value.
5466 2018-07-10 Jakub Jelinek <jakub@redhat.com>
5469 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
5470 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
5471 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
5472 (gimplify_omp_for): For composite loops, move outer
5473 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
5474 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
5475 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
5476 TREE_LIST for both the original class iterator and the "last" helper
5477 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
5478 loop, remember has_decl_expr from outer composite loops for the
5479 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
5481 2018-07-09 Martin Sebor <msebor@redhat.com>
5485 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
5486 accessing implicitly initialized array elements.
5487 * expr.c (string_constant): Handle string initializers of
5488 character arrays within aggregates.
5489 * gimple-fold.c (fold_array_ctor_reference): Add argument.
5490 Store element offset. As a special case, handle zero size.
5491 (fold_nonarray_ctor_reference): Same.
5492 (fold_ctor_reference): Add argument. Store subobject offset.
5493 * gimple-fold.h (fold_ctor_reference): Add argument.
5495 2018-07-09 Paul Koning <ni1d@arrl.net>
5497 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
5498 (pdp11_insn_cost): New function.
5499 (pdp11_md_asm_adjust): New function.
5500 (TARGET_INVALID_WITHIN_DOLOOP): Define.
5501 (pdp11_rtx_costs): Update to match machine better.
5502 (output_addr_const_pdp11): Correct format mismatch warnings.
5503 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
5504 * config/pdp11/pdp11.md: General change to add base_cost and/or
5505 length attributes for use by new pdp11_insn_cost function.
5506 (MIN_BRANCH): Correct definition.
5508 (doloop_end): Use standard pattern name for looping pattern.
5509 (doloop_end_nocc): New.
5510 (movsf): Add another constraint alternative.
5511 (zero_extendqihi2): Add constraint alternatives for not in place
5513 (zero_extendhisi2): Remove.
5514 (shift patterns): Add CC handling variants.
5518 (define_peephole2): New peephole to recognize mov that sets CC for
5521 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5523 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
5524 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
5527 2018-07-09 Jakub Jelinek <jakub@redhat.com>
5530 * real.c (real_nextafter): Return true if result is denormal.
5532 2018-07-09 Martin Liska <mliska@suse.cz>
5534 * common.opt: Add back wrongly removed attribute.
5536 2018-07-09 Richard Biener <rguenther@suse.de>
5539 * dwarf2out.c (gen_block_die): For an early generated DIE
5540 always output high/low PC attributes.
5542 2018-07-09 Tom de Vries <tdevries@suse.de>
5544 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
5545 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
5546 onto VAR_DECL with abstract origin.
5548 2018-07-07 Jim Wilson <jimw@sifive.com>
5550 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
5552 2018-07-07 Tom de Vries <tdevries@suse.de>
5554 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
5556 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
5558 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
5560 (combine_bound): Use wide-int overflow calculation instead of
5562 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
5564 * fold-const.c (int_const_binop_2): Same.
5565 (extract_muldiv_1): Same.
5566 (fold_div_compare): Same.
5567 (fold_abs_const): Same.
5569 * poly-int.h (add): Same.
5573 * predict.c (predict_iv_comparison): Same.
5574 * profile-count.c (slow_safe_scale_64bit): Same.
5575 * simplify-rtx.c (simplify_const_binary_operation): Same.
5576 * tree-chrec.c (tree_fold_binomial): Same.
5577 * tree-data-ref.c (split_constant_offset_1): Same.
5578 * tree-if-conv.c (idx_within_array_bound): Same.
5579 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
5580 * tree-ssa-phiopt.c (minmax_replacement): Same.
5581 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
5582 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
5583 * vr-values.c (vr_values::adjust_range_with_scev): Same.
5584 * wide-int.cc (wi::add_large): Same.
5585 (wi::mul_internal): Same.
5586 (wi::sub_large): Same.
5587 (wi::divmod_internal): Same.
5588 * wide-int.h: Change overflow type to overflow_type for neg, add,
5589 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
5590 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
5591 mul_internal, divmod_internal.
5592 (overflow_type): New enum.
5593 (accumulate_overflow): New.
5595 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
5597 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
5598 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
5600 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5602 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
5603 argument is checked for zero before entering loop, avoid checking again.
5605 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
5607 * gimplify.h (generic_expr_could_trap_p): Set as global function.
5608 * gimplify.h (generic_expr_could_trap_p): Likwise.
5609 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
5611 2018-07-06 Jakub Jelinek <jakub@redhat.com>
5613 PR tree-optimization/86401
5614 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
5615 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
5616 (fold_bit_and_mask): ... here. New helper function for match.pd.
5617 * fold-const.h (fold_bit_and_mask): Declare.
5618 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
5620 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
5623 * target.def (translate_mode_attribute): New hook.
5624 * targhooks.h (default_translate_mode_attribute): Declare.
5625 * targhooks.c (default_translate_mode_attribute): New function.
5626 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
5627 * doc/tm.texi: Regenerate.
5628 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
5629 (rs6000_translate_mode_attribute): New function.
5631 2018-07-06 Paul Koning <ni1d@arrl.net>
5633 * doc/md.texi (define_split): Document DONE and FAIL.
5634 (define_peephole2): Ditto.
5636 2018-07-05 Jeff Law <law@redhat.com>
5638 PR tree-optimization/86010
5639 * tree-ssa-dse.c (compute_trims): More aggressively trim at
5640 both the head and tail of mem* and str* calls.
5642 2018-07-05 Jim Wilson <jimw@sifive.com>
5644 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
5646 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
5648 * config/aarch64/aarch64-simd.md: correct flags text for
5649 MIN_EXPR replacement.
5651 2018-07-05 James Clarke <jrtc27@jrtc27.com>
5653 * configure: Regenerated.
5655 2018-07-05 Carl Love <cel@us.ibm.com>
5657 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
5658 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
5659 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
5660 VSX_BUILTIN_DOUBLEL_V4SF.
5662 2018-07-05 Martin Sebor <msebor@redhat.com>
5665 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
5666 than its domain to compute its the upper bound of a char array.
5668 2018-07-05 Nathan Sidwell <nathan@acm.org>
5670 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
5671 * doc/cpp.texi: Update comment.
5672 * doc/tm.texi: Rebuilt.
5673 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
5674 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
5675 * doc/extend.texi (Backwards Compatibility): Clarify it is system
5676 headers affected by extern "C".
5677 * system.h: Poison NO_IMPLICIT_EXTERN_C.
5678 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
5679 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
5680 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
5681 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
5682 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
5683 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
5684 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
5685 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
5686 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
5687 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
5688 config/sparc/sp64-elf.h, config/spu/spu.h,
5689 config/stormy16/stormy16.h, config/v850/v850.h,
5690 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
5691 define NO_IMPLICIT_EXTERN_C.
5692 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
5694 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5697 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5698 instead of GET_MODE_SIZE when comparing Units.
5700 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5703 * rtlanal.c (set_noop_p): Constrain on mode change,
5704 include hard-reg-set.h
5706 2018-07-05 Tamar Christina <tamar.christina@arm.com>
5708 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
5710 2018-07-05 Jakub Jelinek <jakub@redhat.com>
5713 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5716 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5718 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5720 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
5723 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
5725 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
5727 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
5730 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
5732 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
5734 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
5735 neg_*_op* variables.
5737 2018-07-04 Martin Liska <mliska@suse.cz>
5739 * tree-switch-conversion.c: Define
5740 max_ratio_for_speed and max_ratio_for_size constants.
5742 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5743 Martin Liska <mliska@suse.cz>
5748 * common.opt: Rename align options with 'str_' prefix.
5749 * common/config/i386/i386-common.c (set_malign_value): New
5751 (ix86_handle_option): Use it to set -falign-* options/
5752 * config/aarch64/aarch64-protos.h (struct tune_params): Change
5753 type from int to string.
5754 * config/aarch64/aarch64.c: Update default values from int
5756 * config/alpha/alpha.c (alpha_override_options_after_change):
5758 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
5759 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5760 max skip conditionally.
5761 * config/i386/freebsd.h (SUBALIGN_LOG): New.
5762 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5763 max skip conditionally.
5764 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5765 max skip conditionally.
5766 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
5767 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5768 max skip conditionally.
5769 * config/i386/i386.c (struct ptt): Change type from int to
5771 (ix86_default_align): Set default values.
5772 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
5773 max skip conditionally.
5774 * config/i386/iamcu.h (SUBALIGN_LOG): New.
5775 (ASM_OUTPUT_MAX_SKIP_ALIGN):
5776 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
5777 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5778 max skip conditionally.
5779 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
5780 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
5781 * config/i386/x86-64.h (SUBALIGN_LOG): New.
5782 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
5783 max skip conditionally.
5784 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5785 * config/ia64/ia64.c (ia64_option_override): Set default values
5786 for alignment options.
5787 * config/m68k/m68k.c: Handle new str_align_* options.
5788 * config/mips/mips.c (mips_set_compression_mode): Change
5790 (mips_option_override): Set default values for options.
5791 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5793 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5795 * config/rx/rx.c (rx_option_override): Likewise.
5796 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
5797 (LABEL_ALIGN): Use align_labels_log.
5798 (LOOP_ALIGN): Use align_loops_align.
5799 * config/s390/s390.c (s390_asm_output_function_label): Use new
5801 * config/sh/sh.c (sh_override_options_after_change):
5802 Change type of constants.
5803 * config/spu/spu.c (spu_sched_init): Likewise.
5804 * config/sparc/sparc.c (sparc_option_override): Set default
5806 * config/visium/visium.c (visium_option_override): Likewise.
5807 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
5808 emit p2align format with last argument if it's not needed.
5809 * doc/invoke.texi: Document extended format of -falign-*.
5810 * final.c: Use align_labels alignment.
5811 * flags.h (struct target_flag_state): Change type to use
5813 (struct align_flags_tuple): New.
5814 (struct align_flags): Likewise.
5815 (align_loops_log): Redefine macro to use new types.
5816 (align_loops_max_skip): Redefine macro to use new types.
5817 (align_jumps_log): Redefine macro to use new types.
5818 (align_jumps_max_skip): Redefine macro to use new types.
5819 (align_labels_log): Redefine macro to use new types.
5820 (align_labels_max_skip): Redefine macro to use new types.
5821 (align_functions_log): Redefine macro to use new types.
5822 (align_loops): Redefine macro to use new types.
5823 (align_jumps): Redefine macro to use new types.
5824 (align_labels): Redefine macro to use new types.
5825 (align_functions): Redefine macro to use new types.
5826 (align_functions_max_skip): Redefine macro to use new types.
5827 (align_loops_value): New macro.
5828 (align_jumps_value): New macro.
5829 (align_labels_value): New macro.
5830 (align_functions_value): New macro.
5831 * function.c (invoke_set_current_function_hook): Propagate
5832 alignment values from flags to global variables default in
5834 * ipa-icf.c (sem_function::equals_wpa): Use
5835 cl_optimization_option_eq instead of memcmp.
5836 * lto-streamer.h (cl_optimization_stream_out): Support streaming
5838 (cl_optimization_stream_in): Likewise.
5839 * optc-save-gen.awk: Support strings in cl_optimization.
5840 * opth-gen.awk: Likewise.
5841 * opts.c (finish_options): Remove error checking of invalid
5843 (MAX_CODE_ALIGN): Remove.
5844 (MAX_CODE_ALIGN_VALUE): Likewise.
5845 (parse_and_check_align_values): New function.
5846 (check_alignment_argument): Likewise.
5847 (common_handle_option): Use check_alignment_argument.
5848 * opts.h (parse_and_check_align_values): Declare.
5849 * toplev.c (init_alignments): Remove.
5850 (read_log_maxskip): New.
5851 (parse_N_M): Likewise.
5852 (parse_alignment_opts): Likewise.
5853 (backend_init_target): Remove usage of init_alignments.
5854 * toplev.h (parse_alignment_opts): Declare.
5855 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
5857 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
5858 * tree.c (cl_option_hasher::equal): New.
5859 * varasm.c: Use new global macros.
5861 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
5863 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
5864 Use a simpler align directive also if MAXSKIP = ALIGN-1.
5865 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5866 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5867 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5868 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5869 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
5870 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
5871 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
5872 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5873 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5874 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5875 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5877 2018-07-04 Martin Liska <mliska@suse.cz>
5878 Jonathan Wakely <jwakely@redhat.com>
5880 * coverage.c: Use correct type.
5881 * doc/invoke.texi: Language correction.
5883 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
5886 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
5887 ENDBRANCH for non-tail call which may return via indirect branch.
5888 * doc/extend.texi: Document indirect_return attribute.
5890 2018-07-03 Martin Sebor <msebor@redhat.com>
5892 PR tree-optimization/86274
5893 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
5895 (format_floating): Correct handling of infinities and NaNs.
5897 2018-07-03 Martin Sebor <msebor@redhat.com>
5899 * print-tree.c (print_real_cst): New function.
5900 (print_node_brief): Call it.
5901 (print_node): Ditto.
5903 2018-07-03 Jeff Law <law@redhat.com>
5905 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
5906 into a single pattern.
5908 * config/h8300/h8300.md (ors code_iterator): New.
5909 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
5910 a single pattern and single splitter.
5911 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
5912 (iorqi3_1, xorqi3_1): Likewise.
5913 (iorqi3, xorqi3 expanders): Similarly.
5915 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
5916 (movmd_internal) into a single pattern using the P mode iterator.
5917 (movmd splitters): Similarly.
5918 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
5919 (movsd splitters): Similarly.
5921 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
5922 ADDB, ADDW and ADDL into a single ADD attribute which selects the
5923 right table based on the size of the operand.
5924 * config/h8300/h8300.md (length_table): Corresponding changes. All
5925 references to "addb", "addw" and "addl" changed to "add".
5926 (btst patterns): Merge two variants into a single pattern.
5927 (tstqi, tsthi): Likewise.
5928 (addhi3_incdec, addsi3_incdec): Likewise.
5929 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
5930 (mulhi3, mulsi3): Likewise.
5931 (udivhi3, udivsi3): Likewise.
5932 (divhi3, divsi3): Likewise.
5933 (andorqi3, andorhi3, andorsi3): Likewise.
5935 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
5938 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
5939 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
5941 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5943 PR tree-optimization/85694
5944 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
5945 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
5947 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
5948 (<u>avg<mode>3_ceil): New patterns.
5950 2018-07-03 David Malcolm <dmalcolm@redhat.com>
5952 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
5953 scan-tree-dump directive.
5954 * gcc.dg/vect/slp-perm-2.c: Likewise.
5955 * gcc.dg/vect/slp-perm-3.c: Likewise.
5956 * gcc.dg/vect/slp-perm-5.c: Likewise.
5957 * gcc.dg/vect/slp-perm-6.c: Likewise.
5958 * gcc.dg/vect/slp-perm-7.c: Likewise.
5959 * gcc.dg/vect/slp-perm-8.c: Likewise.
5961 2018-07-03 Marek Polacek <polacek@redhat.com>
5964 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
5966 2018-07-03 Richard Biener <rguenther@suse.de>
5969 * tree-ssa-structalias.c (find_func_clobbers): Properly
5970 handle indirect calls.
5972 2018-07-03 Jeff Law <law@redhat.com>
5974 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
5975 (shifts): New code iterator.
5976 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
5977 expander. Fix HImode handling on H8/SX.
5978 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
5979 (subqi3, subhi3, subsi3 expanders): Likewise.
5980 (andqi3, andhi3, andsi3 expanders): Likewise.
5981 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
5982 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
5983 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
5984 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
5985 (zero_extendqihi2, zero_extendqisi2): Likewise.
5986 (extendqihi2, extendqisi2): Likewise.
5987 (rotlqi3, rotlhi3, rotlsi3): Likewise.
5988 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
5989 (rotlqi3_1, rotlhi3_1): Likewise.
5990 (logicalhi3_sn, logicalsi3_sn): Likewise.
5991 (logicalhi3, logicalsi3): Likewise.
5993 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5995 * tree-vect-patterns.c (vect_recog_rotate_pattern)
5996 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5997 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
5998 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
5999 type to append_pattern_def_seq instead of creating a stmt_vec_info
6001 (build_mask_conversion): Likewise. Remove vinfo argument.
6002 (vect_add_conversion_to_patterm): Likewise, renaming to...
6003 (vect_add_conversion_to_pattern): ...this.
6004 (vect_recog_mask_conversion_pattern): Update call to
6005 build_mask_conversion. Pass the vector type to
6006 append_pattern_def_seq here too.
6007 (vect_recog_gather_scatter_pattern): Update call to
6008 vect_add_conversion_to_pattern.
6010 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6012 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
6013 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6014 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
6015 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
6016 STMT_VINFO_PATTERN_DEF_SEQ to null here.
6017 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
6018 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
6019 append_pattern_def_seq instead of new_pattern_def_seq.
6020 (vect_recog_divmod_pattern): Do both of the above.
6021 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
6024 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6026 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
6027 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
6028 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
6029 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
6030 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
6031 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6032 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
6033 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6034 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6035 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
6036 parameter with a single stmt_vec_info.
6037 (vect_recog_func_ptr): Likewise.
6038 (vect_recog_gather_scatter_pattern): Likewise, folding in...
6039 (vect_try_gather_scatter_pattern): ...this.
6040 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
6041 the stmt_vec_info of the statement to be matched. Don't clear
6042 STMT_VINFO_RELATED_STMT.
6043 (vect_pattern_recog): Update call accordingly.
6045 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6047 PR tree-optimization/85694
6048 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
6049 (uavgM3_ceil): Document new optabs.
6050 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
6051 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
6053 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
6054 (savg_ceil_optab): New optabs.
6055 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
6056 (vect_vect_recog_func_ptrs): Add it.
6057 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
6058 constant directly from the associated lhs.
6060 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6062 * tree-vect-patterns.c (vect_split_statement): New function.
6063 (vect_convert_input): Use it to try to split an existing cast.
6065 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6067 * poly-int.h (print_hex): New function.
6068 * dumpfile.h (dump_dec, dump_hex): Declare.
6069 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
6070 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
6071 min_input_precision, operation_precision and operation_sign.
6072 * tree-vect-patterns.c (vect_get_range_info): New function.
6073 (vect_same_loop_or_bb_p, vect_single_imm_use)
6074 (vect_operation_fits_smaller_type): Delete.
6075 (vect_look_through_possible_promotion): Add an optional
6076 single_use_p parameter.
6077 (vect_recog_over_widening_pattern): Rewrite to use new
6078 stmt_vec_info infomration. Handle one operation at a time.
6079 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
6080 (vect_truncatable_operation_p, vect_set_operation_type)
6081 (vect_set_min_input_precision): New functions.
6082 (vect_determine_min_output_precision_1): Likewise.
6083 (vect_determine_min_output_precision): Likewise.
6084 (vect_determine_precisions_from_range): Likewise.
6085 (vect_determine_precisions_from_users): Likewise.
6086 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
6087 (vect_vect_recog_func_ptrs): Put over_widening first.
6089 (vect_pattern_recog): Call vect_determine_precisions.
6091 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6093 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
6094 statements that have been replaced by further pattern statements.
6095 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
6097 2018-07-03 Richard Biener <rguenther@suse.de>
6099 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
6100 always set *dt. Dump vectype in vectype overload.
6101 * dumpfile.h (dump_gimple_expr): New function.
6102 (dump_gimple_expr_loc): Likewise.
6103 * dumpfile.c (dump_gimple_expr): New function.
6104 (dump_gimple_expr_loc): Likewise.
6106 2018-07-02 Jeff Law <law@redhat.com>
6108 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
6109 the H8/300, H8/300H and H8/S variants into a single pattern.
6110 (movhi_h8300, movqi_h8300hs): Similarly.
6111 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
6112 (QHI mode iterator): New.
6114 * config/h8300/h8300.md: Remove trailing whitespace.
6116 2018-07-02 Jim Wilson <jimw@sifive.com>
6118 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
6119 instead of emit_insn for interrupt returns.
6120 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
6121 (riscv_sret, riscv_uret): Likewise.
6123 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6125 * pretty-print.c (selftest::test_pp_format): Move save and restore
6126 of quotes to class auto_fix_quotes, and add an instance.
6127 * selftest.c: Include "intl.h".
6128 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
6129 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
6130 * selftest.h (selftest::auto_fix_quotes): New class.
6132 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
6134 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
6135 (aarch64_sve_prepare_conditional_op): Remove.
6136 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
6137 Allow aarch64_simd_reg_or_zero as select operand; remove
6138 the aarch64_sve_prepare_conditional_op call.
6139 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
6140 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6141 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
6142 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
6143 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
6144 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
6145 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
6146 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
6147 and a splitters to match all of the *_any patterns.
6148 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
6150 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
6151 (SVE_COND_FP_BINARY_REV): Remove.
6152 (sve_int_op_rev, sve_fp_op_rev): New.
6153 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
6154 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
6155 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
6156 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
6157 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
6158 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
6159 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
6160 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
6161 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
6163 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
6164 Remove match_dup 1 from the inner unspec.
6165 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
6167 * config/aarch64/aarch64.md (movprfx): New attr.
6168 (length): Default movprfx to 8.
6169 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
6170 (*madd<SVE_I>, *msub<SVE_I): Likewise.
6171 (*<su>mul<SVE_I>3_highpart): Likewise.
6172 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
6173 (*v<ASHIFT><SVE_I>3): Likewise.
6174 (*<su><MAXMIN><SVE_I>3): Likewise.
6175 (*<su><MAXMIN><SVE_F>3): Likewise.
6176 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
6177 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
6178 (*div<SVE_F>4): Likewise.
6180 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
6182 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
6185 2018-07-02 Richard Biener <rguenther@suse.de>
6187 PR tree-optimization/86363
6188 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
6189 memset argument refers to a non-variable address.
6191 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6193 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
6194 {PLUS,MINUS}_EXPR code to...
6195 (adjust_symbolic_bound): ...here,
6196 (combine_bound): ...here,
6197 (set_value_range_with_overflow): ...and here.
6199 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
6201 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
6203 (extract_range_from_abs_expr): ...here.
6205 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
6207 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
6208 -fno-omit-frame-pointer when not optimizing.
6210 2018-07-02 Martin Liska <mliska@suse.cz>
6213 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
6214 (propagate_nothrow): Likewise.
6216 2018-07-02 Martin Liska <mliska@suse.cz>
6219 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
6221 2018-07-02 David Malcolm <dmalcolm@redhat.com>
6223 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
6224 function in r262149, changing "loc" param from source_location to
6225 const dump_location_t &.
6226 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
6227 declaration, as above.
6229 2018-07-01 Paul Koning <ni1d@arrl.net>
6231 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6232 -munit-asm, -mgnu-asm, -mdec-asm.
6233 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
6234 (pdp11_output_labelref): New.
6235 (pdp11_output_def): New.
6236 (pdp11_output_addr_vec_elt): New.
6237 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
6238 %# and %@ format codes.
6239 (pdp11_option_override): New.
6240 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
6241 (pdp11_output_ident): New.
6242 (pdp11_asm_named_section): New.
6243 (pdp11_asm_init_sections): New.
6244 (pdp11_file_start): New.
6245 (pdp11_file_end): New.
6246 (output_ascii): Use .ascii/.asciz for -mdec-asm.
6247 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
6248 %o, like %c but octal.
6249 (pdp11_option_override): New.
6250 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
6252 (DATA_SECTION_ASM_OP): Ditto.
6253 (READONLY_DATA_SECTION_ASM_OP): New.
6254 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
6255 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
6256 (ASM_OUTPUT_LABELREF): Ditto.
6257 (ASM_OUTPUT_DEF): Ditto.
6258 (ASM_OUTPUT_EXTERNAL): New.
6259 (ASM_OUTPUT_SOURCE_FILENAME): New.
6260 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
6261 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
6262 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
6263 %# and %@ format codes.
6264 * config/pdp11/pdp11.opt (mgnu-asm): New.
6265 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
6266 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
6267 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
6269 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
6271 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
6272 dereferencing path[] beyond its length.
6274 (debug_all_paths): New.
6275 (rewire_first_differing_edge): New.
6276 (adjust_paths_after_duplication): New.
6277 (duplicate_thread_path): Call adjust_paths_after_duplication.
6279 (thread_through_all_blocks): Add new argument to
6280 duplicate_thread_path.
6282 2018-06-30 Jim Wilson <jimw@sifive.com>
6284 * config/riscv/predicates.md (p2m1_shift_operand): New.
6285 (high_mask_shift_operand): New.
6286 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
6287 pattern using p2m1_shift_operand.
6288 (lshsi3_zero_extend_3+2): New combiner pattern using
6289 high_mask_shift_operand.
6291 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6293 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
6295 (vect_recog_rotate_pattern): ...here.
6296 (vect_convert_input): Try to insert casts of invariants in the
6298 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
6299 preheader to be empty.
6301 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6303 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
6304 vector type. If given, install it in the new statement's
6306 (vect_element_precision): New function.
6307 (vect_unpromoted_value): New struct.
6308 (vect_unpromoted_value::vect_unpromoted_value): New function.
6309 (vect_unpromoted_value::set_op): Likewise.
6310 (vect_look_through_possible_promotion): Likewise.
6311 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
6312 (vect_widened_op_tree, vect_convert_input): Likewise.
6313 (vect_convert_inputs, vect_convert_output): Likewise.
6314 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
6315 to handle the optional cast of the multiplication result and
6316 vect_widened_op_tree to detect the widened multiplication itself.
6317 Do not require the input and output of promotion casts to have
6318 the same sign, but base the signedness of the operation on the
6319 input rather than the result. If the pattern includes two
6320 promotions, check that those promotions have the same sign.
6321 Do not restrict the MULT_EXPR handling to a double-width result;
6322 handle quadruple-width results and wider. Use vect_convert_inputs
6323 to convert the inputs to the common type.
6324 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
6325 to handle the optional cast of the ABS result. Also allow a sign
6326 change or a sign extension between the ABS and MINUS.
6327 Use vect_widened_op_tree to detect the widened subtraction and use
6328 vect_convert_inputs to convert the inputs to the common type.
6329 (vect_handle_widen_op_by_const): Delete.
6330 (vect_recog_widen_op_pattern): New function.
6331 (vect_recog_widen_mult_pattern): Use it.
6332 (vect_recog_widen_shift_pattern): Likewise.
6333 (vect_recog_widen_sum_pattern): Use
6334 vect_look_through_possible_promotion to handle the promoted
6337 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6339 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
6340 the containing gimple_seq *.
6341 * gimple-iterator.h (gsi_for_stmt): Declare it.
6342 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
6343 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
6344 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
6345 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6346 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
6348 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
6350 (vect_mark_pattern_stmts): ...here. Handle cases in which the
6351 statement being replaced is part of an existing pattern
6352 definition sequence, inserting the new pattern statements before
6354 (vect_pattern_recog_1): Don't return a bool. If the statement
6355 is already part of a pattern, instead apply pattern matching
6356 to the pattern definition statements. Don't clear the
6357 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
6358 (vect_pattern_recog): Don't break after the first match;
6359 continue processing the pattern definition statements instead.
6360 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
6362 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6364 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6365 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6366 (vect_recog_widen_sum_pattern): Use it.
6368 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6370 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6371 phi is not a pattern statement and has not been replaced by
6372 a pattern statement.
6373 * tree-vect-patterns.c (type_conversion_p): Don't check
6374 STMT_VINFO_IN_PATTERN_P.
6375 (vect_recog_vector_vector_shift_pattern): Likewise.
6376 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
6377 the pattern statement rather than the original statement; check
6378 directly for a WIDEN_MULT_EXPR here.
6379 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
6380 vect_is_simple_use to return the pattern statement rather
6381 than the original statement; use is_pattern_stmt_p to check
6382 for such a pattern statement.
6383 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
6384 to return the pattern statement rather than the original statement;
6385 don't do the same transformation here.
6386 (vect_is_simple_use): If the defining statement has been replaced
6387 by a pattern statement, return the pattern statement instead.
6388 Remove the corresponding (local) transformation from the vectype
6391 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
6393 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
6394 end and default to null.
6395 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6396 (vectorizable_reduction): Update calls accordingly, dropping the
6397 gimple ** argument if the passed-back statement isn't needed.
6398 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
6399 (vect_recog_rotate_pattern): Likewise.
6400 (vect_recog_mask_conversion_pattern): Likewise.
6401 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6402 (vect_mask_constant_operand_p): Likewise.
6403 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
6404 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
6405 (get_group_load_store_type, get_load_store_type): Likewise.
6406 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
6407 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
6408 (vectorizable_conversion, vectorizable_assignment): Likewise.
6409 (vectorizable_shift, vectorizable_operation): Likewise.
6410 (vectorizable_store, vect_is_simple_cond): Likewise.
6411 (vectorizable_condition, vectorizable_comparison): Likewise.
6412 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
6413 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
6414 and move it to the end. Cope with null def_stmt_outs.
6416 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6418 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
6420 2018-06-29 Jeff Law <law@redhat.com>
6422 * config/v850/v850.c (v850_legitimate_address_p): Handle large
6423 displacements for TARGET_V850E2V3 and newer.
6424 (TARGET_LRA_P): Remove. Defaults to LRA now.
6425 * config/v850/v850.md (sign23byte_load): Remove.
6426 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
6427 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
6429 2018-06-29 Martin Liska <mliska@suse.cz>
6432 * coverage.c (coverage_init): Mangle full path name.
6433 * doc/invoke.texi: Document the change.
6434 * gcov-io.c (mangle_path): New.
6435 * gcov-io.h (mangle_path): Likewise.
6436 * gcov.c (mangle_name): Use mangle_path for path mangling.
6438 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6440 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
6441 if starting source register is not even.
6443 2018-06-29 Martin Liska <mliska@suse.cz>
6445 PR tree-optimization/86263
6446 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
6447 Make edge redirection.
6449 2018-06-29 David Malcolm <dmalcolm@redhat.com>
6451 * dumpfile.c (dump_loc): Add indentation based on scope depth.
6452 (dump_scope_depth): New variable.
6453 (get_dump_scope_depth): New function.
6454 (dump_begin_scope): New function.
6455 (dump_end_scope): New function.
6456 * dumpfile.h (get_dump_scope_depth): New declaration.
6457 (dump_begin_scope): New declaration.
6458 (dump_end_scope): New declaration.
6459 (class auto_dump_scope): New class.
6460 (AUTO_DUMP_SCOPE): New macro.
6461 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
6464 2018-06-29 Richard Biener <rguenther@suse.de>
6466 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
6467 compute_all_dependences succeeds.
6468 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
6469 exceed --param loop-max-datarefs-for-datadeps.
6471 2018-06-29 Jakub Jelinek <jakub@redhat.com>
6473 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
6475 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
6478 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
6479 alternative 0 in preferred_for_speed attribute.
6481 2018-06-28 Paul Koning <ni1d@arrl.net>
6483 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
6484 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
6485 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
6486 * config/pdp11/pdp11.md: Correct "length" attribute calculation
6487 for shift insn patterns.
6489 2018-06-28 David Malcolm <dmalcolm@redhat.com>
6491 * cgraph.c (cgraph_node::get_body): Replace assignments to
6492 "dump_file" with calls to set_dump_file.
6493 * dumpfile.c (alt_dump_file): Make static, and group with...
6494 (alt_flags): ...this definition.
6495 (dumps_are_enabled): New variable.
6496 (refresh_dumps_are_enabled): New function.
6497 (set_dump_file): New function.
6498 (set_alt_dump_file): New function.
6499 (gcc::dump_manager::dump_start): Replace assignments to
6500 "dump_file" and "alt_dump_file" with calls to set_dump_file and
6502 (gcc::dump_manager::dump_finish): Likewise.
6503 * dumpfile.h (alt_dump_file): Delete decl.
6504 (dumps_are_enabled): New variable decl.
6505 (set_dump_file): New function decl.
6506 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
6508 * tree-nested.c (lower_nested_functions): Replace assignments to
6509 "dump_file" with calls to set_dump_file.
6511 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6513 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
6514 goto_locus of each outgoing edge of each basic block.
6516 2018-06-28 Richard Biener <rguenther@suse.de>
6518 * dwarf2out.c (decl_scope_table): Remove.
6519 (push_decl_scope): Likewise.
6520 (pop_decl_scope): Likewise.
6521 (gen_type_die_for_member): Do not call push/pop_decl_scope.
6522 (gen_struct_or_union_type_die): Likewise.
6523 (gen_tagged_type_die): Likewise.
6524 (dwarf2out_init): Do not initialize decl_scope_table.
6525 (dwarf2out_c_finalize): Do not free it.
6527 2018-06-28 Richard Biener <rguenther@suse.de>
6529 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
6530 deciding whether to not re-use a DIE.
6532 2018-06-28 Richard Biener <rguenther@suse.de>
6534 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
6535 DW_AT_abstract_origin attribute.
6537 2018-06-28 Martin Liska <mliska@suse.cz>
6539 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6540 Use newly introduced constants.
6541 * tree-switch-conversion.h (struct jump_table_cluster):
6542 Define max_ratio_for_size and max_ratio_for_speed.
6544 2018-06-28 Martin Liska <mliska@suse.cz>
6546 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6547 Add new checking assert to catch invalid state.
6548 (jump_table_cluster::can_be_handled): Handle single case
6550 (jump_table_cluster::is_beneficial): Bail out for such case.
6551 (bit_test_cluster::find_bit_tests):
6552 Add new checking assert to catch invalid state.
6553 (bit_test_cluster::can_be_handled): Handle single case
6555 (bit_test_cluster::is_beneficial): Bail out for such case.
6556 (switch_decision_tree::analyze_switch_statement):
6559 2018-06-28 Martin Liska <mliska@suse.cz>
6561 * common.opt: Introduce -completion option.
6562 * gcc.c (driver_handle_option): Handle it.
6563 (driver::main): Print completions if completion
6565 * opt-suggestions.c (option_proposer::get_completions):
6567 (option_proposer::suggest_completion): Likewise.
6568 (option_proposer::find_param_completions): Likewise.
6569 (verify_autocompletions): Likewise.
6570 (test_completion_valid_options): Likewise.
6571 (test_completion_valid_params): Likewise.
6572 (in_completion_p): Likewise.
6573 (empty_completion_p): Likewise.
6574 (test_completion_partial_match): Likewise.
6575 (test_completion_garbage): Likewise.
6576 (opt_proposer_c_tests): Likewise.
6577 * opt-suggestions.h: Declare new functions.
6578 * opts.c (common_handle_option): Handle OPT__completion_.
6579 * selftest-run-tests.c (selftest::run_tests): Add
6580 opt_proposer_c_tests.
6581 * selftest.c (assert_str_startswith): New.
6582 * selftest.h (assert_str_startswith): Likewise.
6583 (opt_proposer_c_tests): New.
6584 (ASSERT_STR_STARTSWITH): Likewise.
6586 2018-06-28 Martin Liska <mliska@suse.cz>
6588 * Makefile.in: Add opt-suggestions.o.
6589 * gcc-main.c: Include opt-suggestions.h.
6590 * gcc.c (driver::driver): Likewise.
6591 (driver::~driver): Remove m_option_suggestions.
6592 (driver::build_option_suggestions): Moved to option_proposer.
6593 (driver::suggest_option): Likewise.
6594 (driver::handle_unrecognized_options): Use option_proposer.
6595 * gcc.h (class driver): Add new memver m_option_proposer.
6596 * opt-suggestions.c: New file.
6597 * opt-suggestions.h: New file.
6599 2018-06-28 Martin Liska <mliska@suse.cz>
6601 * vec.h (class auto_string_vec): New (moved from auto_argvec).
6602 (auto_string_vec::~auto_string_vec): Likewise.
6604 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
6606 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
6607 prevent_decl_creation_for_types fields up and add reset_location field.
6608 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
6609 statement if id->reset_location is true.
6610 (copy_edges_for_bb): Do not set goto_locus on the new edges if
6611 id->reset_location is true.
6612 (copy_phis_for_bb): Force input_location on the arguments if
6613 id->reset_location is true.
6614 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
6615 is set on the function to be inlined.
6617 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
6619 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
6621 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
6623 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
6624 registers for Pmode.
6625 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
6626 hard registers for the clobbered pseudo.
6628 2018-06-27 Paul Koning <ni1d@arrl.net>
6630 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
6631 mutually exclusive options.
6632 * config/pdp11/constraints.md (h): New constraint.
6633 (O): Update definition to match shift code generation.
6634 (D): New constraint.
6635 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
6637 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
6639 (output_jump): Change arguments.
6640 (pdp11_fixed_cc_regs): New function.
6641 (pdp11_cc_mode): Ditto.
6642 (pdp11_expand_shift): Ditto.
6643 (pdp11_assemble_shift): Ditto.
6644 (pdp11_small_shift): Ditto.
6645 (pdp11_branch_cost): Remove.
6646 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
6648 (pdp11_register_move_cost): Update for CC registers.
6649 (pdp11_rtx_costs): Add case for LSHIFTRT.
6650 (pdp11_output_jump): Add CCNZ mode conditional branches.
6651 (notice_update_cc_on_set): Remove.
6652 (pdp11_cc_mode): New function.
6653 (simple_memory_operand): Correct pre/post decrement case.
6654 (no_side_effect_operand): New function.
6655 (pdp11_regno_reg_class): Add CC_REGS class.
6656 (pdp11_fixed_cc_regs): New function.
6657 (pdp11_small_shift): New function.
6658 (pdp11_expand_shift): New function to expand shift insns.
6659 (pdp11_assemble_shift): New function to output shifts.
6660 (pdp11_branch_cost): Remove.
6661 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
6662 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
6663 (WCHAR_TYPE): Ditto.
6664 (PTRDIFF_TYPE): Ditto.
6665 (ADJUST_INSN_LENGTH): New macro.
6666 (FIXED_REGISTERS): Add CC registers.
6667 (CALL_USED_REGISTERS): Ditto.
6669 (REG_CLASS_NAMES): Ditto.
6670 (REG_CLASS_CONTENTS): Ditto.
6671 (SELECT_CC_MODE): Use new function.
6672 (TARGET_FLAGS_REGNUM): New macro.
6673 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
6674 (cc0_reg_rtx): Remove.
6675 (CC_STATUS_MDEP): Remove.
6676 (CC_STATUS_MDEFP_INIT): Remove.
6677 (CC_IN_FPU): Remove.
6678 (NOTICE_UPDATE_CC): Remove.
6679 (REGISTER_NAMES): Add CC registers.
6680 (BRANCH_COST): Change to constant 1.
6681 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
6683 * config/pdp11/pdp11.opt (mbcopy): Remove.
6684 (mbcopy-builtin): Remove.
6685 (mbranch-cheap): Remove.
6686 (mbranch-expensive): Remove.
6687 * config/pdp11/predicates.md (expand_shift_operand): Update to
6688 match shift code generation.
6689 (ccnz_operator): New predicate.
6690 * doc/invoke.texi (PDP-11 Options): Remove deleted options
6691 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
6692 Remove non-existent option -mabshi, -mno-abshi. Document mutually
6694 * doc/md.texi (PDP-11): Document new D and h constraints. Update
6695 description of O constraint.
6697 2018-06-27 Jeff Law <law@redhat.com>
6698 Austin Law <austinklaw@gmail.com>
6700 * config/v850/v850.md (addsi3_set_flags): New pattern.
6701 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
6702 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
6703 (zero_extendhisi2_v850_set_flags): Likewise.
6704 (zero_extendqisi2_v850_set_flags): Likewise.
6705 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
6706 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
6707 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
6709 * config/v850/v850-protos.h (notice_update_cc): Remove.
6710 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
6711 (v850_print_operand): Handle 'D' and "d".
6712 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
6713 Add handling of arithmetic/logical operations compared against zero.
6714 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
6715 Do not look at v850_compare_op, instead get mode from last argument.
6716 (v850_gen_compare): Remove
6717 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
6718 after reload for prologue insns.
6719 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
6721 (construct_save_jarl): Likewise.
6722 (TARGET_FLAGS_REGNUM): Define.
6723 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
6724 (NOTICE_UPDATE_CC): Remove.
6725 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
6726 than cc0. Conditionalize on reload_completed.
6727 (cmpsi_insn, setfcc_insn): Likewise.
6728 (tst1 splitter): Turn into define_and_split which sets the flags
6730 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
6731 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
6732 (cstoresf4, cstoredf4): Clobber the flags.
6733 (cmpsi, cmpsf, cmpdf): Remove expanders.
6734 (setf_insn): Remove pattern.
6735 (addsi3): Turn into define_and_split which clobbers the flags after
6736 reload and a suitable pattern (addsi3_clobber_flags) for use after
6738 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
6739 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
6740 (ashrsi3, ashrsi3_v850e2): Likewise.
6741 (bins): Clobber the flags.
6742 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
6743 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
6744 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
6745 (call_value_internal_short, call_value_internal_long): Likewise.
6746 (callt_save_interrupt, callt_return_interrupt): Likewise.
6747 (save_interrupt, return_interrupt): Likewise.
6748 (callt_save_all_interrupt, save_all_interrupt): Likewise.
6749 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
6750 (restore_all_interrupt, _restore_all_interrupt): Likewise.
6751 (All FP comparisons): Only allow after reload has completed.
6753 (divh, divhu): Tweak output template.
6754 (branch_z_normal, branch_z_invert): Remove
6755 (branch_nz_normal, branch_nz_invert): Likewise.
6756 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
6758 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
6759 * config/v850/v850.c (notice_update_cc): Remove.
6760 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
6761 (CC_NO_CARRY): Likewise.
6762 (NOTICE_UPDATE_CC): Define to nothing.
6763 * config/v850/v850.md: Remove block comment on cc0 handling
6764 Remove "cc" attribute from all patterns. Remove cc_status handling
6765 from all patterns. Minor formatting fixes.
6767 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6769 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
6770 (cortex-a76.cortex-a55): Likewise.
6771 * config/aarch64/aarch64-tune.md: Regenerate.
6772 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
6773 cortex-a76.cortex-a55.
6775 2018-06-27 Jeff Law <law@redhat.com>
6777 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
6778 (MULTILIB_DIRNAMES): Similarly.
6780 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
6782 * gimple.h (gimple_return_retbnd): Delete.
6783 (gimple_return_set_retbnd): Likewise.
6784 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
6785 gimple_return_set_retbnd.
6786 * gimple-pretty-print.c (dump_gimple_return): Remove call to
6787 gimple_return_retbnd and adjust.
6788 * tree-inline.h (struct copy_body_data): Remove retbnd field.
6789 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
6790 Explicitly return NULL in a couple more cases. Move assertion
6791 on debug statements and remove unreachable code.
6792 (reset_debug_binding): Do not test id->retbnd.
6793 (expand_call_inline): Do not set it.
6795 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
6797 * configure.ac: Add --disable-gcov option.
6798 * configure: Regenerate.
6799 * Makefile.in: Honour @enable_gcov@.
6800 * doc/install.texi: Document --disable-gcov.
6802 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6804 * config/arm/arm-cpus.in (cortex-a76): New entry.
6805 (cortex-a76.cortex-a55): Likewise.
6806 * config/arm/arm-tables.opt: Regenerate.
6807 * config/arm/arm-tune.md: Likewise.
6808 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
6809 * doc/invoke.texi (ARM Options): Document cortex-a76 and
6810 cortex-a76.cortex-a55.
6812 2018-06-27 Tamar Christina <tamar.christina@arm.com>
6815 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
6817 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
6819 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
6821 (EPILOGUE_USES): Likewise.
6823 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
6825 * tree-inline.c (remap_location): New function extracted from...
6826 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
6827 (copy_phis_for_bb): ...here. Call remap_location.
6828 (copy_cfg_body): Adjust call to copy_edges_for_bb.
6830 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
6832 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
6833 unaligned vsx for 16B memset.
6835 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
6838 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
6839 ieee128_float_type_node to long_double_type_node unless
6840 TARGET_LONG_DOUBLE_128 is set.
6842 2018-06-26 David Malcolm <dmalcolm@redhat.com>
6844 * cfgloop.c (get_loop_location): Convert return type from
6845 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
6846 by implicit construction from rtx_insn *, and using
6847 dump_user_location_t::from_function_decl for the fallback case.
6848 * cfgloop.h (get_loop_location): Convert return type from
6849 location_t to dump_user_location_t.
6850 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
6851 dump_printf_loc to pass in a dump_location_t rather than a
6852 location_t, via the gimple stmt.
6853 * coverage.c (get_coverage_counts): Update calls to
6854 dump_printf_loc to pass in dump_location_t rather than a
6856 * doc/optinfo.texi (Dump types): Convert example of
6857 dump_printf_loc from taking "locus" to taking "insn". Update
6858 description of the "_loc" calls to cover dump_location_t.
6859 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
6861 (dump_user_location_t::dump_user_location_t): New constructors,
6862 from gimple *stmt and rtx_insn *.
6863 (dump_user_location_t::from_function_decl): New function.
6864 (dump_loc): Make static.
6865 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
6866 const dump_location_t &.
6867 (dump_generic_expr_loc): Delete.
6868 (dump_printf_loc): Convert param "loc" from location_t to
6869 const dump_location_t &.
6870 (selftest::test_impl_location): New function.
6871 (selftest::dumpfile_c_tests): New function.
6872 * dumpfile.h: Include "profile-count.h".
6873 (class dump_user_location_t): New class.
6874 (struct dump_impl_location_t): New struct.
6875 (class dump_location_t): New class.
6876 (dump_printf_loc): Convert 2nd param from source_location to
6877 const dump_location_t &.
6878 (dump_generic_expr_loc): Delete.
6879 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
6880 const dump_location_t &.
6881 * gimple-fold.c (fold_gimple_assign): Update call to
6882 dump_printf_loc to pass in a dump_location_t rather than a
6883 location_t, via the gimple stmt.
6884 (gimple_fold_call): Likewise.
6885 * gimple-loop-interchange.cc
6886 (loop_cand::analyze_iloop_reduction_var): Update for change to
6887 check_reduction_path.
6888 (tree_loop_interchange::interchange): Update for change to
6890 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
6891 change in return-type of find_loop_location.
6892 (graphite_regenerate_ast_isl): Likewise.
6893 * graphite-optimize-isl.c (optimize_isl): Likewise.
6894 * graphite.c (graphite_transform_loops): Likewise.
6895 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
6896 pass in a dump_location_t rather than a location_t, via the
6898 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6899 * ipa.c (walk_polymorphic_call_targets): Likewise.
6900 * loop-unroll.c (report_unroll): Convert "locus" param from
6901 location_t to dump_location_t.
6902 (decide_unrolling): Update for change to get_loop_location's
6904 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
6905 location_t to dump_user_location_t.
6906 (grid_find_single_omp_among_assignments_1): Updates calls to
6907 dump_printf_loc to pass in a dump_location_t rather than a
6908 location_t, via the gimple stmt.
6909 (grid_parallel_clauses_gridifiable): Convert "tloc" from
6910 location_t to dump_location_t. Updates calls to dump_printf_loc
6911 to pass in a dump_location_t rather than a location_t, via the
6913 (grid_inner_loop_gridifiable_p): Likewise.
6914 (grid_dist_follows_simple_pattern): Likewise.
6915 (grid_gfor_follows_tiling_pattern): Likewise.
6916 (grid_target_follows_gridifiable_pattern): Likewise.
6917 (grid_attempt_target_gridification): Convert initialization
6918 of local "grid" from memset to zero-initialization; FIXME: does
6919 this require C++11? Update call to dump_printf_loc to pass in a
6920 optinfo_location rather than a location_t, via the gimple stmt.
6921 * profile.c (read_profile_edge_counts): Updates call to
6922 dump_printf_loc to pass in a dump_location_t rather than a
6924 (compute_branch_probabilities): Likewise.
6925 * selftest-run-tests.c (selftest::run_tests): Call
6927 * selftest.h (dumpfile_c_tests): New decl.
6928 * tree-loop-distribution.c (pass_loop_distribution::execute):
6929 Update for change in return type of find_loop_location.
6930 * tree-parloops.c (parallelize_loops): Likewise.
6931 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
6932 "locus" from location_t to dump_user_location_t.
6933 (canonicalize_loop_induction_variables): Likewise.
6934 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
6935 for change in return type of find_loop_location.
6936 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
6937 to dump_printf_loc to pass in a dump_location_t rather than a
6938 location_t, via the stmt.
6939 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
6941 * tree-vect-loop-manip.c (find_loop_location): Convert return
6942 type from source_location to dump_user_location_t.
6943 (vect_do_peeling): Update for above change.
6944 (vect_loop_versioning): Update for change in type of
6946 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
6947 from location_t to dump_user_location_t.
6948 (vect_estimate_min_profitable_iters): Update for change in type
6950 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
6951 location_t to dump_location_t.
6952 (vect_slp_bb): Update for change in type of vect_location.
6953 * tree-vectorizer.c (vect_location): Convert from source_location
6954 to dump_user_location_t.
6955 (try_vectorize_loop_1): Update for change in vect_location's type.
6956 (vectorize_loops): Likewise.
6957 (increase_alignment): Likewise.
6958 * tree-vectorizer.h (vect_location): Convert from source_location
6959 to dump_user_location_t.
6960 (find_loop_location): Convert return type from source_location to
6961 dump_user_location_t.
6962 (check_reduction_path): Convert 1st param from location_t to
6963 dump_user_location_t.
6964 * value-prof.c (check_counter): Update call to dump_printf_loc to
6965 pass in a dump_user_location_t rather than a location_t; update
6966 call to error_at for change in type of "locus".
6967 (check_ic_target): Update call to dump_printf_loc to
6968 pass in a dump_user_location_t rather than a location_t, via the
6971 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
6973 * config/s390/s390.h (enum processor_flags): Do not use
6974 default tune parameter when -march was specified.
6976 2018-06-26 Jakub Jelinek <jakub@redhat.com>
6979 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
6980 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
6983 2018-06-26 Richard Biener <rguenther@suse.de>
6985 PR tree-optimization/86287
6987 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
6988 (vect_analyze_loop): Initialize n_stmts.
6990 2018-06-26 Richard Biener <rguenther@suse.de>
6993 * fold-const.c (fold_convertible_p): Pointer extension
6996 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
6999 * dwarf2out.c (loc_list_has_views): Adjust comments.
7000 (dw_loc_list): Split single cross-partition range with
7003 2018-06-25 Jeff Law <law@redhat.com>
7005 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
7006 on -mbig-switch by default.
7008 * config/v850/predicates.md (const_float_1_operand): Fix match_code
7010 (const_float_0_operand): Remove unused predicate.
7011 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
7012 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
7013 (recipsf2): New expander. Original pattern now called
7015 (recipdf2, recipdf2_insn): Similarly.
7016 (rsqrtsf2, rsqrtsf2_insn): Similarly
7017 (rsqrtdf2, rsqrtdf2_insn): Similarly
7019 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
7021 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
7022 Simplify logic for FreeBSD (twice).
7024 2018-06-25 Martin Sebor <msebor@redhat.com>
7026 PR tree-optimization/86204
7027 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
7028 a strnlen result if it's less than the length of the string.
7030 2018-06-25 Martin Sebor <msebor@redhat.com>
7032 PR tree-optimization/85700
7033 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
7034 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
7035 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
7037 2018-06-25 Martin Sebor <msebor@redhat.com>
7039 * doc/extend.texi (Zero-length arrays): Update and clarify.
7041 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
7043 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
7044 added IEEE/IBM long double multilib support on PowerPC little
7045 endian Linux systems.
7046 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
7047 (MULTILIB_DEFAULTS): Likewise.
7048 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7050 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
7051 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
7052 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
7054 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
7057 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
7058 (REORDER_45): Likewise.
7060 2018-06-25 Jeff Law <law@redhat.com>
7062 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
7063 dividend to 32 bits. Adjust length.
7064 (udivmodhi4): Cleanup output template. Fix length.
7066 2018-06-25 Carl Love <cel@us.ibm.com>
7068 * config/rs6000/vsx.md: Change word selector to prefered location.
7070 2018-06-25 Richard Biener <rguenther@suse.de>
7072 PR tree-optimization/86304
7073 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
7074 epilogue-if-converted loops as well.
7076 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
7078 * lto-section-out.c (lto_begin_section): Do not print section
7079 name for noaddr and unnumbered dumps.
7081 2018-06-25 Richard Biener <rguenther@suse.de>
7083 * tree-vectorizer.h (struct vec_info_shared): New structure
7084 with parts split out from struct vec_info and loop_nest from
7085 struct _loop_vec_info.
7086 (struct vec_info): Adjust accordingly.
7087 (struct _loop_vec_info): Likewise.
7088 (LOOP_VINFO_LOOP_NEST): Adjust.
7089 (LOOP_VINFO_DATAREFS): Likewise.
7090 (LOOP_VINFO_DDRS): Likewise.
7091 (struct _bb_vec_info): Likewise.
7092 (BB_VINFO_DATAREFS): Likewise.
7093 (BB_VINFO_DDRS): Likewise.
7094 (struct _stmt_vec_info): Add dr_aux member.
7095 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
7096 (DR_MISALIGNMENT_UNINITIALIZED): New.
7097 (set_dr_misalignment): Adjust.
7098 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
7099 (vect_analyze_loop): Adjust prototype.
7100 (vect_analyze_loop_form): Likewise.
7101 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7102 Compute dependences lazily.
7103 (vect_record_base_alignments): Use shared datarefs/ddrs.
7104 (vect_verify_datarefs_alignment): Likewise.
7105 (vect_analyze_data_refs_alignment): Likewise.
7106 (vect_analyze_data_ref_accesses): Likewise.
7107 (vect_analyze_data_refs): Likewise.
7108 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
7109 constructor parameter for shared part.
7110 (vect_analyze_loop_form): Pass in shared part and adjust.
7111 (vect_analyze_loop_2): Pass in storage for the number of
7112 stmts. Move loop nest finding to the caller. Compute
7114 (vect_analyze_loop): Pass in shared part.
7115 (vect_transform_loop): Verify shared datarefs are unchanged.
7116 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
7117 constructor parameter for shared part.
7118 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
7119 (vect_slp_bb): Verify shared datarefs are unchanged before
7121 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
7123 (new_stmt_vec_info): Initialize DR_AUX misalignment to
7124 DR_MISALIGNMENT_UNINITIALIZED.
7125 * tree-vectorizer.c (vec_info::vec_info): Add constructor
7126 parameter for shared part.
7127 (vec_info::~vec_info): Adjust.
7128 (vec_info_shared::vec_info_shared): New.
7129 (vec_info_shared::~vec_info_shared): Likewise.
7130 (vec_info_shared::save_datarefs): Likewise.
7131 (vec_info_shared::check_datarefs): Likewise.
7132 (try_vectorize_loop_1): Construct shared part live for analyses
7133 of a single loop for multiple vector sizes.
7134 * tree-parloops.c (gather_scalar_reductions): Adjust.
7136 2018-06-25 Richard Biener <rguenther@suse.de>
7138 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
7139 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
7140 (vect_analyze_data_refs): Remove similar code from here and
7141 simplify accordingly.
7143 2018-06-25 Richard Biener <rguenther@suse.de>
7145 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
7146 for reverse storage order accesses rather than asserting
7147 they cannot happen here.
7149 2018-06-25 Tom de Vries <tdevries@suse.de>
7152 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
7153 Use data16 instead of .byte for insn prefix.
7155 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
7158 * parser.c (make_char_string_pack): Pass this literal chars
7159 through cpp_interpret_string.
7160 (cp_parser_userdef_numeric_literal): Check the result of
7161 make_char_string_pack.
7163 2018-06-24 Maya Rashish <coypu@sdf.org>
7165 * ginclude/stddef.h: Simplify conditions around avoiding
7166 re-definition of __size_t.
7168 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7170 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
7171 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
7173 2018-06-22 Maya Rashish <coypu@sdf.org>
7175 * doc/invoke.texi (mno-fancy-math-387): Update for changes
7176 made to OpenBSD and NetBSD through the years.
7178 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7180 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7181 behavior of vec_pack (vector double, vector double) to match
7182 behavior of vec_float2 (vector double, vector double).
7184 2018-06-22 Olivier Hainque <hainque@adacore.com>
7186 * gimplify.c (gimplify_function_tree): Prevent creation
7187 of a trampoline for the address of the current function
7188 passed to entry/exit instrumentation hooks.
7190 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
7193 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
7196 2018-06-22 Martin Liska <mliska@suse.cz>
7198 PR tree-optimization/86263
7199 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7200 Bail out if is_enabled is false.
7201 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
7203 (jump_table_cluster::is_enabled): New function.
7205 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7207 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7208 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
7209 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
7210 (lto_input_ts_binfo_tree_pointers): Likewise.
7211 * tree-streamer-out.c (streamer_write_tree_bitfields,
7212 write_ts_binfo_tree_pointers): Likewise.
7213 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
7215 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
7217 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
7219 2018-06-22 Martin Liska <mliska@suse.cz>
7221 * symbol-summary.h (get): Make it pure and inline move
7222 functionality from ::get function.
7223 (get): Remove and inline into ::get and ::get_create.
7224 (get_create): Move code from ::get function.
7226 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7229 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
7230 -x assembler-with-cpp.
7232 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7234 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
7235 _FILE_OFFSET_BITS=64 for C++.
7237 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
7239 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
7240 conversion insn that shows up when pr85657-3.c is compiled using
7241 IEEE 128-bit long double.
7242 (neg<mode>2_internal): Use the correct mode to check whether the
7243 mode is IBM extended.
7244 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
7245 multiply and divide external functions from being created more
7248 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7250 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
7252 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
7253 the edge can be forwarded.
7254 (cfg_layout_merge_blocks): Likewise.
7256 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7258 * except.c (finish_eh_generation): Commit edge insertions only after
7259 the EH edges have been redirected from post-landing to landing pads.
7261 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7263 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
7264 create_tmp_var_for to create the FRAME decl.
7265 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
7267 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
7269 * tree-inline.c (copy_edges_for_bb): Minor tweak.
7270 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
7271 debug statement when resetting its value.
7272 (expand_call_inline): Copy the locus of the call onto the assignment
7273 of the return value, if any. Use local variable in more cases.
7275 2018-06-21 Martin Liska <mliska@suse.cz>
7277 * ipa-pure-const.c (propagate_nothrow): Use
7278 funct_state_summaries->get.
7279 (dump_malloc_lattice): Likewise.
7280 (propagate_malloc): Likewise.
7282 2018-06-21 Richard Biener <rguenther@suse.de>
7284 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
7285 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
7286 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
7288 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
7289 BLOCK_ABSTRACT_ORIGIN unconditionally.
7291 2018-06-21 David Malcolm <dmalcolm@redhat.com>
7293 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
7295 * ipa-reference.c (ipa_reference_c_finalize): Delete
7296 ipa_ref_opt_sum_summaries and set it to NULL.
7298 2018-06-21 Tom de Vries <tdevries@suse.de>
7300 PR tree-optimization/85859
7301 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
7302 test with comment from bb_no_side_effects_p.
7304 2018-06-21 Richard Biener <rguenther@suse.de>
7306 PR tree-optimization/86232
7307 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
7308 max for constant niter.
7310 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7312 * config/aarch64/aarch64-simd.md
7313 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
7315 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
7317 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
7318 Make opernads of the unspec commutative.
7320 2018-06-21 Richard Biener <rguenther@suse.de>
7322 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
7323 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
7324 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
7325 (vect_analyze_data_ref_dependence): Re-order checks to deal with
7327 (vect_record_base_alignments): Do not record base alignment
7328 for gathers or scatters.
7329 (vect_compute_data_ref_alignment): Drop return value that is always
7330 true. Bail out early for gathers or scatters.
7331 (vect_enhance_data_refs_alignment): Bail out early for gathers
7333 (vect_find_same_alignment_drs): Likewise.
7334 (vect_analyze_data_refs_alignment): Remove dead code.
7335 (vect_slp_analyze_and_verify_node_alignment): Likewise.
7336 (vect_analyze_data_refs): For possible gathers or scatters do
7337 not create an alternate DR, just check their possible validity
7338 and mark them. Adjust DECL_NONALIASED handling to not rely
7340 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
7341 update inits of gathers or scatters.
7342 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
7343 Also copy gather/scatter flag to pattern vinfo.
7345 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
7347 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
7348 behavior of vec_packsu (vector unsigned long long, vector unsigned
7349 long long) to match behavior of vec_packs with same signature.
7351 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
7352 Thomas Schwinge <thomas@codesourcery.com>
7353 Cesar Philippidis <cesar@codesourcery.com>
7355 * gimplify.c (gimplify_scan_omp_clauses): Add support for
7356 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7357 (gimplify_adjust_omp_clauses): Likewise.
7358 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
7359 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
7360 (gimplify_omp_target_update): Update handling of acc update and
7362 * omp-low.c (install_var_field): Remove unused parameter
7363 base_pointers_restrict.
7364 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
7365 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
7367 (omp_target_base_pointers_restrict_p): Delete.
7368 (scan_omp_target): Update call to scan_sharing_clauses.
7369 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
7371 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
7372 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
7373 (convert_local_omp_clauses): Likewise.
7374 * tree-pretty-print.c (dump_omp_clause): Likewise.
7375 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
7377 (omp_clause_code_name): Likewise.
7379 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7382 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
7385 PR tree-optimization/86231
7386 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
7387 anti-range don't overwrite *vr0min before using it to compute *vr0max.
7389 2018-06-20 Tom de Vries <tdevries@suse.de>
7391 PR tree-optimization/86097
7392 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
7393 iv type if signedness of iv type is not the same as that of *nit.
7395 2018-06-20 Jakub Jelinek <jakub@redhat.com>
7397 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
7398 EDGE_EH edges, verify they are all EDGE_EH.
7400 2018-06-20 Maya Rashish <coypu@sdf.org>
7402 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
7404 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7406 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
7407 * config/aarch64/aarch64.c (xgene1_tunings): Add
7408 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
7409 (aarch64_mode_valid_for_sched_fusion_p):
7410 Allow 16-byte modes.
7411 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
7412 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
7414 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
7416 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
7417 * config/aarch64/iterators.md (VQ2): New mode iterator.
7419 2018-06-20 Martin Liska <mliska@suse.cz>
7421 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7422 Change default ratio from 10 to 8.
7424 2018-06-20 Martin Liska <mliska@suse.cz>
7426 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7428 (bit_test_cluster::find_bit_tests): Likewise.
7429 (switch_decision_tree::analyze_switch_statement): Find clusters.
7430 * tree-switch-conversion.h (struct jump_table_cluster): Document
7433 2018-06-20 Martin Liska <mliska@suse.cz>
7435 * tree-switch-conversion.c (switch_conversion::collect):
7436 Record m_uniq property.
7437 (switch_conversion::expand): Bail out for special conditions.
7438 (group_cluster::~group_cluster): New.
7439 (group_cluster::group_cluster): Likewise.
7440 (group_cluster::dump): Likewise.
7441 (jump_table_cluster::emit): New.
7442 (switch_decision_tree::fix_phi_operands_for_edges): New.
7443 (struct case_node): Remove struct.
7444 (jump_table_cluster::can_be_handled): New.
7445 (case_values_threshold): Moved to header.
7446 (reset_out_edges_aux): Likewise.
7447 (jump_table_cluster::is_beneficial): New.
7448 (bit_test_cluster::can_be_handled): Likewise.
7449 (add_case_node): Remove.
7450 (bit_test_cluster::is_beneficial): New.
7451 (case_bit_test::cmp): New.
7452 (bit_test_cluster::emit): New.
7453 (expand_switch_as_decision_tree_p): Remove.
7454 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
7455 (fix_phi_operands_for_edge): Likewise.
7456 (switch_decision_tree::analyze_switch_statement): New.
7457 (compute_cases_per_edge): Move ...
7458 (switch_decision_tree::compute_cases_per_edge): ... here.
7459 (try_switch_expansion): Likewise.
7460 (switch_decision_tree::try_switch_expansion): Likewise.
7461 (record_phi_operand_mapping): Likewise.
7462 (switch_decision_tree::record_phi_operand_mapping): Likewise.
7463 (emit_case_decision_tree): Likewise.
7464 (switch_decision_tree::emit): Likewise.
7465 (balance_case_nodes): Likewise.
7466 (switch_decision_tree::balance_case_nodes): Likewise.
7467 (dump_case_nodes): Likewise.
7468 (switch_decision_tree::dump_case_nodes): Likewise.
7469 (emit_jump): Likewise.
7470 (switch_decision_tree::emit_jump): Likewise.
7471 (emit_cmp_and_jump_insns): Likewise.
7472 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
7473 (emit_case_nodes): Likewise.
7474 (switch_decision_tree::emit_case_nodes): Likewise.
7475 (conditional_probability): Remove.
7476 * tree-switch-conversion.h (enum cluster_type): New.
7478 (struct cluster): Likewise.
7479 (cluster::cluster): Likewise.
7480 (struct simple_cluster): Likewise.
7481 (simple_cluster::simple_cluster): Likewise.
7482 (struct group_cluster): Likewise.
7483 (struct jump_table_cluster): Likewise.
7484 (struct bit_test_cluster): Likewise.
7485 (struct min_cluster_item): Likewise.
7486 (struct case_tree_node): Likewise.
7487 (case_tree_node::case_tree_node): Likewise.
7488 (jump_table_cluster::case_values_threshold): Likewise.
7489 (struct case_bit_test): Likewise.
7490 (struct switch_decision_tree): Likewise.
7491 (struct switch_conversion): Likewise.
7492 (switch_decision_tree::reset_out_edges_aux): Likewise.
7494 2018-06-20 Martin Liska <mliska@suse.cz>
7496 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
7497 (hoist_edge_and_branch_if_true): Likewise.
7498 (expand_switch_using_bit_tests_p): Likewise.
7499 (struct case_bit_test): Likewise.
7500 (case_bit_test_cmp): Likewise.
7501 (emit_case_bit_tests): Likewise.
7502 (switch_conversion::switch_conversion): New class.
7503 (struct switch_conv_info): Remove old struct.
7504 (collect_switch_conv_info): More to ...
7505 (switch_conversion::collect): ... this.
7506 (check_range): Likewise.
7507 (switch_conversion::check_range): Likewise.
7508 (check_all_empty_except_final): Likewise.
7509 (switch_conversion::check_all_empty_except_final): Likewise.
7510 (check_final_bb): Likewise.
7511 (switch_conversion::check_final_bb): Likewise.
7512 (create_temp_arrays): Likewise.
7513 (switch_conversion::create_temp_arrays): Likewise.
7514 (free_temp_arrays): Likewise.
7515 (gather_default_values): Likewise.
7516 (switch_conversion::gather_default_values): Likewise.
7517 (build_constructors): Likewise.
7518 (switch_conversion::build_constructors): Likewise.
7519 (constructor_contains_same_values_p): Likewise.
7520 (switch_conversion::contains_same_values_p): Likewise.
7521 (array_value_type): Likewise.
7522 (switch_conversion::array_value_type): Likewise.
7523 (build_one_array): Likewise.
7524 (switch_conversion::build_one_array): Likewise.
7525 (build_arrays): Likewise.
7526 (switch_conversion::build_arrays): Likewise.
7527 (gen_def_assigns): Likewise.
7528 (switch_conversion::gen_def_assigns): Likewise.
7529 (prune_bbs): Likewise.
7530 (switch_conversion::prune_bbs): Likewise.
7531 (fix_phi_nodes): Likewise.
7532 (switch_conversion::fix_phi_nodes): Likewise.
7533 (gen_inbound_check): Likewise.
7534 (switch_conversion::gen_inbound_check): Likewise.
7535 (process_switch): Use the newly created class.
7536 (switch_conversion::expand): New.
7537 (switch_conversion::~switch_conversion): New.
7538 * tree-switch-conversion.h: New file.
7540 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7542 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
7543 tree-vect-patterns.c.
7544 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
7545 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
7546 (vect_recog_sad_pattern): Likewise.
7547 (vect_recog_widen_sum_pattern): Likewise.
7548 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
7549 (vect_recog_widen_shift_pattern): Remove the type_in argument.
7550 (vect_recog_rotate_pattern): Likewise.
7551 (vect_recog_mult_pattern): Likewise.
7552 (vect_recog_vector_vector_shift_pattern): Likewise.
7553 (vect_recog_divmod_pattern): Likewise.
7554 (vect_recog_mixed_size_cond_pattern): Likewise.
7555 (vect_recog_bool_pattern): Likewise.
7556 (vect_recog_mask_conversion_pattern): Likewise.
7557 (vect_try_gather_scatter_pattern): Likewise.
7558 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
7559 (vect_recog_over_widening_pattern): Likewise.
7560 (vect_recog_gather_scatter_pattern): Likewise.
7561 (vect_recog_func_ptr): Move from tree-vectorizer.h
7562 (vect_vect_recog_func_ptrs): Move further down the file.
7563 (vect_recog_func): Likewise. Remove the third argument.
7564 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
7565 (vect_pattern_recog_1): Expect the pattern function to do any
7566 necessary target tests. Also expect it to provide a vector type.
7567 Remove the type_in handling.
7569 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7571 * tree-vect-patterns.c (vect_pattern_detected): New function.
7572 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
7573 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
7574 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
7575 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
7576 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7577 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7578 (vect_recog_mask_conversion_pattern)
7579 (vect_try_gather_scatter_pattern): Likewise.
7581 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7583 * tree-vect-patterns.c (vect_get_internal_def): New function.
7584 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7585 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
7586 (search_type_for_mask_1): Use it.
7588 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7590 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7591 redundant WIDEN_SUM_EXPR handling.
7592 (vect_recog_sad_pattern): Likewise.
7594 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7596 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
7597 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
7598 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
7599 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
7600 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
7602 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7604 * tree-vect-stmts.c (vectorizable_call): Make sure that we
7605 use the stmt_vec_info of the original bb statement for the
7606 new zero assignment, even if the call is part of a pattern.
7608 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7610 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
7611 that the sequence is attached to the original statement rather
7612 than the pattern statement.
7613 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
7614 PATTERN_DEF_SEQ from the original statement rather than
7615 the main pattern statement.
7616 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
7617 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
7618 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
7620 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
7622 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
7623 definition statements before the early exit for statements that aren't
7625 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
7627 (vect_transform_loop): ...here. Process pattern definition
7628 statements without first checking whether the main pattern
7629 statement is live or relevant.
7631 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7633 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
7634 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
7636 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
7638 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
7639 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
7640 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
7641 (expand_block_compare): Change select_block_compare_mode call.
7642 (expand_strncmp_align_check): Use new functions, fix comment.
7643 (emit_final_str_compare_gpr): New function.
7644 (expand_strn_compare): Refactor and clean up code.
7645 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
7647 2018-06-19 Tony Reix <tony.reix@atos.com>
7648 Damien Bergamini <damien.bergamini@atos.com>
7649 David Edelsohn <dje.gcc@gmail.com>
7651 * collect2.c (static_obj): New variable.
7652 (static_libs): New variable.
7653 (is_in_list): Uncomment declaration.
7654 (main): Track AIX libraries linked statically.
7655 (is_in_list): Uncomment definition.
7656 (scan_prog_file): Don't add AIX shared libraries initializer
7657 to constructor list if linking statically.
7659 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
7661 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
7663 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
7665 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
7667 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
7670 2018-06-19 Martin Liska <mliska@suse.cz>
7672 * config/i386/i386.c (ix86_can_inline_p): Do not use
7673 ipa_fn_summaries::get_create.
7674 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
7676 (devirtualization_time_bonus): Likewise.
7677 (ipcp_propagate_stage): Likewise.
7678 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7679 (edge_set_predicate): Likewise.
7680 (evaluate_conditions_for_known_args): Likewise.
7681 (evaluate_properties_for_edge): Likewise.
7682 (ipa_call_summary::reset): Tranform to ...
7683 (ipa_call_summary::~ipa_call_summary): ... this.
7684 (ipa_fn_summary::reset): Transform to ...
7685 (ipa_fn_summary::~ipa_fn_summary): ... this.
7686 (ipa_fn_summary_t::remove): Rename to ...
7687 (ipa_fn_summary_t::remove_callees): ... this.
7688 (ipa_fn_summary_t::duplicate): Use placement new
7689 instead of memory copy.
7690 (ipa_call_summary_t::duplicate): Likewise.
7691 (ipa_call_summary_t::remove): Remove.
7692 (dump_ipa_call_summary): Change get_create to get.
7693 (ipa_dump_fn_summary): Dump only when summary exists.
7694 (analyze_function_body): Use symbol_summary::get instead
7696 (compute_fn_summary): Likewise.
7697 (estimate_edge_devirt_benefit): Likewise.
7698 (estimate_edge_size_and_time): Likewise.
7699 (inline_update_callee_summaries): Likewise.
7700 (remap_edge_change_prob): Likewise.
7701 (remap_edge_summaries): Likewise.
7702 (ipa_merge_fn_summary_after_inlining): Likewise.
7703 (write_ipa_call_summary): Likewise.
7704 (ipa_fn_summary_write): Likewise.
7705 (ipa_free_fn_summary): Likewise.
7706 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
7707 (struct ipa_call_summary): Likewise.
7708 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
7710 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
7711 (estimate_size_after_inlining): Likewise.
7712 (estimate_growth): Likewise.
7713 (growth_likely_positive): Likewise.
7714 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7715 (inline_call): Likewise.
7716 * ipa-inline.c (caller_growth_limits): Likewise.
7717 (can_inline_edge_p): Likewise.
7718 (can_inline_edge_by_limits_p): Likewise.
7719 (compute_uninlined_call_time): Likewise.
7720 (compute_inlined_call_time): Likewise.
7721 (want_inline_small_function_p): Likewise.
7722 (edge_badness): Likewise.
7723 (update_caller_keys): Likewise.
7724 (update_callee_keys): Likewise.
7725 (inline_small_functions): Likewise.
7726 (inline_to_all_callers_1): Likewise.
7727 (dump_overall_stats): Likewise.
7728 (early_inline_small_functions): Likewise.
7729 (early_inliner): Likewise.
7730 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7731 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7732 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7733 * ipa-split.c (execute_split_functions): Likewise.
7734 * symbol-summary.h: Likewise.
7735 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7737 2018-06-19 Richard Biener <rguenther@suse.de>
7739 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
7740 (vectorize_loops): ... here. Fix dbgcnt handling.
7741 (try_vectorize_loop): Wrap try_vectorize_loop_1.
7743 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
7746 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
7747 ieee128 argument takes up only one (vector) register, not two (floating
7750 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
7752 * gimplify.c (gimplify_init_constructor): Really never clear for an
7753 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
7755 2018-06-19 Richard Biener <rguenther@suse.de>
7757 PR tree-optimization/86179
7758 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
7759 after failed recognition.
7761 2018-06-18 Martin Sebor <msebor@redhat.com>
7764 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
7765 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
7766 Handle integer subtraction.
7767 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
7768 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
7770 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7772 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
7773 param from rtx to rtx_insn *.
7774 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
7776 (frv_ifcvt_modify_insn): Likwise.
7777 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
7778 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
7779 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
7780 as_a <rtx_insn *> cast to local "unprotected_region" once
7781 it's been established that it's not NULL or pc_rtx.
7782 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
7783 param "sethi" from rtx to rtx_insn *.
7784 (nds32_group_float_insns): Likewise for param "insn".
7785 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
7787 (vax_output_int_subtract): Likewise.
7788 * config/vax/vax.c (vax_output_int_add): Likewise for param
7790 (vax_output_int_subtract): Likewise.
7791 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
7792 (emit_pattern_after): Likewise for param "after".
7793 (emit_insn_after): Likewise.
7794 (emit_jump_insn_after): Likewise.
7795 (emit_call_insn_after): Likewise.
7796 (emit_debug_insn_after): Likewise.
7797 (emit_pattern_before): Likewise for param "before".
7798 (emit_insn_before): Likewise.
7799 (emit_jump_insn_before): Likewise.
7800 * final.c (get_insn_template): Likewise for param "insn", removing
7802 * output.h (get_insn_template): Likewise for 2nd param.
7803 * rtl.h (emit_insn_before): Likewise.
7804 (emit_jump_insn_before): Likewise.
7805 (emit_debug_insn_before_noloc): Likewise.
7806 (emit_insn_after): Likewise.
7807 (emit_jump_insn_after): Likewise.
7808 (emit_call_insn_after): Likewise.
7809 (emit_debug_insn_after): Likewise.
7810 (set_insn_deleted): Likewise for param.
7812 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
7815 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
7816 floating point modes, so that IFmode is numerically greater than
7817 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
7818 to declare the ordering. This prevents IFmode from being
7819 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
7820 machine. Include rs6000-modes.h to share the fractional values
7821 between genmodes* and the rest of the compiler.
7825 * config/rs6000/rs6000-modes.h: New file.
7826 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
7827 meaning of rs6000_long_double_size so that 126..128 selects an
7828 appropriate 128-bit floating point type.
7829 (rs6000_option_override_internal): Likewise.
7830 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
7831 (TARGET_LONG_DOUBLE_128): Change the meaning of
7832 rs6000_long_double_size so that 126..128 selects an appropriate
7833 128-bit floating point type.
7834 (LONG_DOUBLE_TYPE_SIZE): Update comment.
7835 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
7836 source and destination to match the standard usage.
7837 (truncifkf2): Likewise.
7838 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
7839 ISA 2.07 to use an explicit clobber, instead of passing in a
7841 (copysign<mode>3_soft): Likewise.
7843 2018-06-18 David Malcolm <dmalcolm@redhat.com>
7845 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
7846 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
7847 (vect_slp_analyze_instance_dependence): Likewise.
7848 (vect_enhance_data_refs_alignment): Likewise.
7849 (vect_analyze_data_refs_alignment): Likewise.
7850 (vect_slp_analyze_and_verify_instance_alignment
7851 (vect_analyze_data_ref_accesses): Likewise.
7852 (vect_prune_runtime_alias_test_list): Likewise.
7853 (vect_analyze_data_refs): Likewise.
7854 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
7855 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7856 (vect_analyze_scalar_cycles_1): Likewise.
7857 (vect_get_loop_niters): Likewise.
7858 (vect_analyze_loop_form_1): Likewise.
7859 (vect_update_vf_for_slp): Likewise.
7860 (vect_analyze_loop_operations): Likewise.
7861 (vect_analyze_loop): Likewise.
7862 (vectorizable_induction): Likewise.
7863 (vect_transform_loop): Likewise.
7864 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
7865 * tree-vect-slp.c (vect_analyze_slp): Likewise.
7866 (vect_make_slp_decision): Likewise.
7867 (vect_detect_hybrid_slp): Likewise.
7868 (vect_slp_analyze_operations): Likewise.
7869 (vect_slp_bb): Likewise.
7870 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
7871 (vectorizable_bswap): Likewise.
7872 (vectorizable_call): Likewise.
7873 (vectorizable_simd_clone_call): Likewise.
7874 (vectorizable_conversion): Likewise.
7875 (vectorizable_assignment): Likewise.
7876 (vectorizable_shift): Likewise.
7877 (vectorizable_operation): Likewise.
7878 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
7880 2018-06-18 Martin Sebor <msebor@redhat.com>
7882 PR tree-optimization/81384
7883 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
7884 * builtins.c (expand_builtin_strnlen): New function.
7885 (expand_builtin): Call it.
7886 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
7887 * builtins.def (BUILT_IN_STRNLEN): New.
7888 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
7889 Warn for bounds in excess of maximum object size.
7890 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
7891 single-value ranges. Handle strnlen.
7892 (handle_builtin_strlen): Handle strnlen.
7893 (strlen_check_and_optimize_stmt): Same.
7894 * doc/extend.texi (Other Builtins): Document strnlen.
7896 2018-06-18 Maya Rashish <coypu@sdf.org>
7898 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
7899 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
7900 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
7902 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
7904 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
7906 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7908 * tree.c (escaped_string::escape): Replace cast to char * by
7909 const_cast<char *> (unescaped).
7911 2018-06-18 Nick Clifton <nickc@redhat.com>
7914 * tree.c (escaped_string): New class. Converts an unescaped
7915 string into its escaped equivalent.
7916 (warn_deprecated_use): Use the new class to convert the
7917 deprecation message, if present.
7918 (test_escaped_strings): New self test.
7919 (test_c_tests): Add test_escaped_strings.
7920 * doc/extend.texi (deprecated): Add a note that the
7921 deprecation message is affected by the -fmessage-length
7922 option, and that control characters will be escaped.
7923 (#pragma GCC error): Document this pragma.
7924 (#pragma GCC warning): Likewise.
7925 * doc/invoke.texi (-fmessage-length): Document this option's
7926 effect on the #warning and #error preprocessor directives and
7927 the deprecated attribute.
7929 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
7931 * tree.c (decl_value_expr_lookup): Revert latest change.
7932 (decl_value_expr_insert): Likewise.
7934 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
7936 * gimplify.c (nonlocal_vlas): Delete.
7937 (nonlocal_vla_vars): Likewise.
7938 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
7940 (gimplify_body): Do not create and destroy nonlocal_vlas.
7941 * tree-nested.c: Include diagnostic.h.
7942 (use_pointer_in_frame): Tweak.
7943 (lookup_field_for_decl): Add assertion and declare the transformation.
7944 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
7945 internal error when the reference is in a wrong context. Do not
7946 create a debug decl by default.
7947 (note_nonlocal_block_vlas): Delete.
7948 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
7949 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
7950 create a debug decl by default.
7951 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
7952 call is in a wrong context.
7953 (fixup_vla_decls): New function.
7954 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
7955 debug variables were created.
7956 * tree.c (decl_value_expr_lookup): Add checking assertion.
7957 (decl_value_expr_insert): Likewise.
7959 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
7962 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
7963 * tree-scalar-evolution.c (interpret_expr): Likewise.
7964 (expression_expensive_p): Likewise.
7965 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
7966 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
7967 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
7968 (ssa_defined_by_minus_one_stmt_p): New.
7970 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
7973 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
7974 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
7975 * dojump.c (do_jump): Likewise.
7976 * expr.c (expand_expr_real_2): Check operand type's sign.
7977 * fold-const.c (const_unop): Handle ABSU_EXPR.
7978 (fold_abs_const): Likewise.
7979 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
7980 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
7981 (strip_sign_op_1): Likesise.
7982 * match.pd: Add new pattern to generate ABSU_EXPR.
7983 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
7984 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7985 * tree-eh.c (operation_could_trap_helper_p): Likewise.
7986 * tree-inline.c (estimate_operator_cost): Likewise.
7987 * tree-pretty-print.c (dump_generic_node): Likewise.
7988 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
7989 * tree.def (ABSU_EXPR): New.
7991 2018-06-16 Jakub Jelinek <jakub@redhat.com>
7994 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
7995 documented as preserved for backward compatibility only.
7996 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
7998 PR rtl-optimization/86108
7999 * bb-reorder.c (create_forwarder_block): Renamed to ...
8000 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
8001 jump from new landing pad to the second part.
8002 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
8005 2018-06-15 Jakub Jelinek <jakub@redhat.com>
8008 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
8009 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
8010 Only call store_expr for halves if the mode is the same.
8013 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
8014 Fix up comment formatting.
8016 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
8018 * typed-splay-tree.h (typed_splay_tree::remove): New function.
8019 (typed_splay_tree::closure,
8020 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
8021 (typed_splay_tree::typed_splay_tree,
8022 typed_splay_tree::operator =): Declared private.
8023 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
8024 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
8025 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
8026 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
8027 typed_splay_tree::splay_tree_splay,
8028 typed_splay_tree::splay_tree_foreach_helper,
8029 typed_splay_tree::splay_tree_insert,
8030 typed_splay_tree::splay_tree_remove,
8031 typed_splay_tree::splay_tree_lookup,
8032 typed_splay_tree::splay_tree_predecessor,
8033 typed_splay_tree::splay_tree_successor,
8034 typed_splay_tree::splay_tree_min,
8035 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
8036 (typed_splay_tree::root, typed_splay_tree::comp,
8037 typed_splay_tree::delete_key,
8038 typed_splay_tree::delete_value): New data members.
8039 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
8040 typed_splay_tree::remove.
8042 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
8044 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
8045 -mginv and -mno-ginv to the assembler.
8046 * config/mips/mips.opt (-mcrc): New option.
8048 * doc/invoke.text (-mcrc): Document.
8051 2018-06-15 Nick Clifton <nickc@redhat.com>
8054 * tree.c (escaped_string): New class. Converts an unescaped
8055 string into its escaped equivalent.
8056 (warn_deprecated_use): Use the new class to convert the
8057 deprecation message, if present.
8058 (test_escaped_strings): New self test.
8059 (test_c_tests): Add test_escaped_strings.
8060 * doc/extend.texi (deprecated): Add a note that the
8061 deprecation message is affected by the -fmessage-length
8062 option, and that control characters will be escaped.
8063 (#pragma GCC error): Document this pragma.
8064 (#pragma GCC warning): Likewise.
8065 * doc/invoke.texi (-fmessage-length): Document this option's
8066 effect on the #warning and #error preprocessor directives and
8067 the deprecated attribute.
8069 2018-06-15 Richard Biener <rguenther@suse.de>
8071 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
8072 here, also noting vector size used.
8073 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
8074 size used in MSG_OPTIMIZED_LOCATIONS dump.
8075 (pass_slp_vectorize::execute): Adjust.
8077 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
8080 * config/arc/arc.c (arc_return_address_register): Fix
8083 2018-06-15 Richard Biener <rguenther@suse.de>
8086 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
8087 leave useless conversion stripping to force_gimple_operand_gsi.
8088 (gimplify_build2): Likewise.
8089 (gimplify_build1): Likewise.
8091 2018-06-15 Richard Biener <rguenther@suse.de>
8094 * tree-cfg.c (move_stmt_op): unshare invariant addresses
8095 before adjusting their block.
8097 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
8099 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
8100 multilibs for *-*-rtems*.
8101 * config/riscv/t-rtems: New file.
8103 2018-06-14 Jakub Jelinek <jakub@redhat.com>
8106 * match.pd ((A +- CST1) +- CST2): Punt if last resort
8107 unsigned_type_for returns NULL.
8110 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
8111 subregs of multi-word pseudos unless the float mode has word size.
8113 2018-06-14 Richard Biener <rguenther@suse.de>
8116 * tree-vect-generic.c (build_word_mode_vector_type): Remove
8117 duplicate and harmful type_hash_canon.
8118 * tree.c (type_hash_canon): Assert we didn't find ourselves.
8120 2018-06-14 Richard Biener <rguenther@suse.de>
8123 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
8126 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
8128 * config/rtems.h (STDINT_LONG32): Define.
8130 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
8131 Prachi Godbole <prachi.godbole@imgtec.com>
8133 * config/mips/mips-cpus.def: Define P6600.
8134 * config/mips/mips-tables.opt: Regenerate.
8135 * config/mips/mips.c (mips_ucbranch_type): New enum.
8136 (mips_rtx_cost_data): Add support for P6600.
8137 (mips_issue_rate): Likewise.
8138 (mips_multipass_dfa_lookahead): Likewise.
8139 (mips_avoid_hazard): Likewise.
8140 (mips_reorg_process_insns): Likewise.
8141 (mips_classify_branch_p6600): New function.
8142 * config/mips/mips.h (TUNE_P6600): New define.
8143 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
8144 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
8145 * config/mips/mips.md: Include p6600.md.
8146 (processor): Add p6600.
8147 * config/mips/p6600.md: New file.
8148 * doc/invoke.texi: Add p6600 to supported architectures.
8150 2018-06-13 Martin Sebor <msebor@redhat.com>
8152 PR tree-optimization/86114
8153 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
8155 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
8157 2018-06-13 Richard Biener <rguenther@suse.de>
8159 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
8160 Properly set vector type of the intermediate stmt.
8161 * tree-vect-stmts.c (vectorizable_operation): The destination
8162 var always has vectype_out type.
8164 2018-06-13 Jeff Law <law@redhat.com>
8166 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
8167 integer 0 for argument to print_rtl_with_bb.
8168 (rl78_reorg): Likewise.
8170 2018-06-13 David Malcolm <dmalcolm@redhat.com>
8172 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
8173 from rtx to rtx_insn *.
8174 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
8176 (add_sched_insns_for_speculation): Likewise for local "target",
8177 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8178 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
8179 from rtx_insn ** to rtx_code_label **.
8180 (reorg_emit_nops): Likewise.
8181 (c6x_reorg): Likewise for local "call_labels".
8182 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
8184 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
8185 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
8186 the loops over LABEL_REFS.
8187 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
8189 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
8190 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
8191 (split_branches): Strengthen local "olabel" from rtx to
8192 rtx_insn *, adding a safe_as_a cast.
8193 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
8195 (add_insn_after): Likewise for first two params.
8196 (add_insn_before): Likewise.
8197 (remove_insn): Likewise for param.
8198 (emit_pattern_before_noloc): Likewise for second and third params.
8199 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
8200 (emit_call_insn_before_noloc): Likewise.
8201 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
8203 (emit_barrier_before): Likewise.
8204 (emit_label_before): Strengthen "label" param from "rtx" to
8205 "rtx_code_label *". Strengthen "before" param from "rtx" to
8207 (emit_insn_after_1): Strengthen "after" param from "rtx" to
8209 (emit_pattern_after_noloc): Likewise.
8210 (emit_insn_after_noloc): Likewise.
8211 (emit_jump_insn_after_noloc): Likewise.
8212 (emit_call_insn_after_noloc): Likewise.
8213 (emit_debug_insn_after_noloc): Likewise.
8214 (emit_barrier_after): Likewise.
8215 (emit_label_after): Likewise for both params.
8216 (emit_pattern_after_setloc): Likewise for "after" param. Convert
8217 "loc" param from "int" to "location_t".
8218 (emit_insn_after_setloc): Likewise.
8219 (emit_jump_insn_after_setloc): Likewise.
8220 (emit_call_insn_after_setloc): Likewise.
8221 (emit_debug_insn_after_setloc): Likewise.
8222 (emit_pattern_before_setloc): Likewise for "before" param. Convert
8223 "loc" param from "int" to "location_t".
8224 (emit_pattern_before): Convert NULL_RTX to NULL.
8225 (emit_insn_before_setloc): Convert "loc" param from "int" to
8227 (emit_jump_insn_before_setloc): Likewise.
8228 (emit_call_insn_before_setloc): Likewise.
8229 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
8230 rtx_insn *. Convert "loc" param from "int" to "location_t".
8231 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
8232 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
8233 Convert 3rd param from "int" to "location_t".
8234 (emit_barrier_before, emit_barrier_after, next_real_insn):
8235 Strengthen param from rtx to rtx_insn *.
8236 (emit_label_before): Strengthen 1st param from "rtx" to
8237 "rtx_code_label *". Strengthen 2nd param from "rtx" to
8239 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
8240 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
8241 Strengthen 2nd param from "rtx" to "rtx_insn *".
8242 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
8243 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
8244 Likewise. Convert 3rd param from "int" to "location_t".
8245 (emit_label_after): Strengthen 1st param from "rtx" to
8247 (next_real_insn, remove_insn): Strengthen param from "rtx" to
8249 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
8250 from "rtx" to "rtx_insn *".
8252 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
8254 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
8255 bodies streamed in with -Q.
8256 * dumpfile.c (dump_files): Add lto-stream-out dump file.
8257 * dumpfile.h (tree_dump_index): Add lto_stream_out.
8258 * gimple-streamer-out.c: Include gimple-pretty-print.h
8259 (output_bb): Dump stmts streamed.
8260 * lto-section-out.c: Include print-tree.h
8261 (lto_begin_section): Dump sections created.
8262 (lto_output_decl_index): Dump decl encoded.
8263 * lto-streamer-out.c: Include print-tree.h
8264 (create_output_block): Dump output block created.
8265 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
8266 (output_function): Dump function output.
8267 (output_constructor): Dump constructor streamed.
8268 (write_global_stream): Output indexes encoded.
8269 (produce_asm_for_decls): Dump streams encoded.
8270 * lto-streamer.c (streamer_dump_file): New global var.
8271 * lto-streamer.h (streamer_dump_file): Declare.
8272 * passes.c (ipa_write_summaries): Initialize streamer dump.
8273 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
8276 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
8279 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
8280 offsets for register save directives. Emit a second batch of save
8281 directives, if need be, when the function accesses prior frames.
8283 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8285 * config/arc/fpu.md (fmasf4): Force operand to register.
8286 (fnmasf4): Likewise.
8288 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8290 * config/arc/arc-protos.h (arc_pad_return): Remove.
8291 * config/arc/arc.c (machine_function): Remove force_short_suffix
8293 (arc_print_operand): Adjust printing of '&'.
8294 (arc_verify_short): Remove conditional printing of short suffix.
8295 (arc_final_prescan_insn): Remove reference to size_reason.
8296 (pad_return): New function.
8297 (arc_reorg): Call pad_return.
8298 (arc_pad_return): Remove.
8299 (arc_init_machine_status): Remove reference to force_short_suffix.
8300 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
8301 (attr length): When attribute iscompact is true force to 2
8302 regardless; in the case of maybe check if we want to force the
8303 instruction to have 4 bytes length.
8304 (nopv): Change it to generate 4 byte long nop as well.
8305 (blockage): New pattern.
8306 (simple_return): Remove call to arc_pad_return.
8307 (p_return_i): Likewise.
8309 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8311 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
8313 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8315 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
8318 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8320 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
8321 for ARC700 and ARCv2.
8323 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
8326 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
8327 operands[2] instead of operands[1].
8330 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8332 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
8333 case, check whether the outer register overlaps an unallocatable
8334 register, not just whether it fits the required class.
8336 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
8338 * poly-int.h (can_div_trunc_p): Add new overload in which all values
8340 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
8341 (memrefs_conflict_p): Likewise.
8342 (init_alias_analysis): Likewise.
8343 * cfgexpand.c (expand_debug_expr): Likewise.
8344 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
8345 * cse.c (fold_rtx): Likewise.
8346 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
8347 * expr.c (emit_block_move_hints): Likewise.
8348 (clear_storage_hints, push_block, emit_push_insn): Likewise.
8349 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
8350 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
8351 (emit_group_store): Likewise.
8352 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
8353 to read the PRE/POST_MODIFY increment.
8354 * calls.c (store_one_arg): Use strip_offset.
8355 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
8357 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
8359 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
8360 (simplify_binary_operation_1): Extend CONST_INT handling to
8362 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
8363 than a HOST_WIDE_INT.
8364 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
8366 (adjust_mems, add_stores): Update accodingly.
8367 (vt_canonicalize_addr): Track polynomial offsets.
8368 (emit_note_insn_var_location): Likewise.
8369 (vt_add_function_parameter): Likewise.
8370 (vt_initialize): Likewise.
8372 2018-06-12 Jeff Law <law@redhat.com>
8374 * config.gcc (alpha*-*-freebsd*): Remove.
8375 * config/alpha/freebsd.h: Remove.
8377 2018-06-12 David Malcolm <dmalcolm@redhat.com>
8380 * spellcheck-tree.c (levenshtein_distance): Rename to...
8381 (get_edit_distance): ...this, and update for underlying renaming.
8382 * spellcheck-tree.h (levenshtein_distance): Rename to...
8383 (get_edit_distance): ...this.
8384 * spellcheck.c (levenshtein_distance): Rename to...
8385 (get_edit_distance): ...this. Convert from Levenshtein distance
8386 to Damerau-Levenshtein distance by supporting transpositions of
8387 adjacent characters. Rename "v1" to "v_next" and "v0" to
8389 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
8390 (selftest::test_edit_distance_unit_test_oneway): ...this, and
8391 update for underlying renaming.
8392 (selftest::levenshtein_distance_unit_test): Rename to...
8393 (selftest::test_get_edit_distance_unit): ...this, and update for
8394 underlying renaming.
8395 (selftest::test_find_closest_string): Add example from PR 69968
8396 where transposition helps
8397 (selftest::test_metric_conditions): Update for renaming.
8398 (selftest::test_metric_conditions): Likewise.
8399 (selftest::spellcheck_c_tests): Likewise.
8400 * spellcheck.h (levenshtein_distance): Rename both overloads to...
8401 (get_edit_distance): ...this.
8402 (best_match::consider): Update for renaming.
8404 2018-06-12 Martin Sebor <msebor@redhat.com>
8406 PR tree-optimization/85259
8407 * builtins.c (compute_objsize): Handle constant offsets.
8408 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
8409 true iff a warning has been issued.
8410 * gimple.h (gimple_nonartificial_location): New function.
8411 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
8412 gimple_nonartificial_location and handle -Wno-system-headers.
8413 (handle_builtin_stxncpy): Same.
8415 2018-06-12 Martin Sebor <msebor@redhat.com>
8418 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
8420 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
8422 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
8423 BUILTIN_VEC_XST entries for pointer to double and long long.
8425 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
8428 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8430 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
8433 2018-06-12 Martin Liska <mliska@suse.cz>
8435 * doc/options.texi: Document IntegerRange.
8437 2018-06-12 Martin Liska <mliska@suse.cz>
8439 * config/i386/i386.opt: Make MPX-related options as Deprecated.
8440 * opt-functions.awk: Handle Deprecated flag.
8441 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
8443 (read_cmdline_option): Report warning for a deprecated option.
8444 * opts.h (struct cl_option): Add new field cl_deprecated.
8445 (CL_ERR_DEPRECATED): New.
8447 2018-06-12 Martin Liska <mliska@suse.cz>
8449 * doc/options.texi: Document Deprecated option flag.
8451 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8453 * config/arc/arc-arch.h (arc_extras): New enum.
8454 (arc_cpu_t):Add field extra.
8455 (arc_cpu_types): Consider the extras.
8456 * config/arc/arc-cpus.def: Add extras info.
8457 * config/arc/arc-opts.h (processor_type): Consider extra field.
8458 * config/arc/arc.c (arc_override_options): Handle extra field.
8460 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
8462 * config/arc/arc-arch.h: Update ARC_OPTX macro.
8463 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
8465 * config/arc/arc.c (arc_init): Update pic warning.
8466 (irq_range): Update irq range parsing warnings.
8467 (arc_override_options): Update various warning messages.
8468 (arc_handle_aux_attribute): Likewise.
8470 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
8472 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
8474 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8476 * doc/sourcebuild.texi: Document usage of line number 0 in verify
8477 compiler messages directives.
8479 2018-06-12 Matthew Fortune <mfortune@gmail.com>
8481 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
8482 * config/mips/mips-tables.opt: Regenerate.
8483 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
8485 * doc/invoke.texi: Document -march=i6500.
8487 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
8489 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
8490 (i6400_gpmul): Add cpu_unit.
8491 (i6400_gpdiv): Likewise.
8492 (i6400_msa_add_d): Update reservations.
8493 (i6400_msa_int_add) Likewise.
8494 (i6400_msa_short_logic3) Likewise.
8495 (i6400_msa_short_logic2) Likewise.
8496 (i6400_msa_short_logic) Likewise.
8497 (i6400_msa_move) Likewise.
8498 (i6400_msa_cmp) Likewise.
8499 (i6400_msa_short_float2) Likewise.
8500 (i6400_msa_div_d) Likewise.
8501 (i6400_msa_long_logic1) Likewise.
8502 (i6400_msa_long_logic2) Likewise.
8503 (i6400_msa_mult) Likewise.
8504 (i6400_msa_long_float2) Likewise.
8505 (i6400_msa_long_float4) Likewise.
8506 (i6400_msa_long_float5) Likewise.
8507 (i6400_msa_long_float8) Likewise.
8508 (i6400_fpu_fadd): Include frint type.
8509 (i6400_fpu_store): New define_insn_reservation.
8510 (i6400_fpu_load): Likewise.
8511 (i6400_fpu_move): Likewise.
8512 (i6400_fpu_fcmp): Likewise.
8513 (i6400_fpu_fmadd): Likewise.
8514 (i6400_int_mult): Include imul3nc type and update reservation.
8515 (i6400_int_div): Include idiv3 type and update reservation.
8516 (i6400_int_load): Update to check type not move_type.
8517 (i6400_int_store): Likewise.
8518 (i6400_int_prefetch): Set zero latency.
8520 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
8522 * gcc.c: Document new %@{...} sequence.
8523 (LINK_COMMAND_SPEC): Use it for the -L switches.
8524 (cpp_unique_options): Use it for the -I switches.
8525 (at_file_argbuf): New global variable.
8526 (in_at_file): Likewise.
8527 (alloc_args): Create at_file_argbuf.
8528 (clear_args): Truncate at_file_argbuf.
8529 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
8530 (open_at_file): New function.
8531 (close_at_file): Likewise.
8532 (create_at_file): Delete.
8533 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
8536 (validate_switches_from_spec): Deal with %@{...} sequence.
8537 (validate_switches): Likewise.
8538 (driver::finalize): Call clear_args.
8540 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
8542 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
8544 2018-06-11 Martin Sebor <msebor@redhat.com>
8546 * doc/invoke.texi (-Wall): List -Wc++17-compat.
8547 (Wno-class-memaccess): Add @opindex.
8548 (Wno-templates, Wno-multiple-inheritance): Same.
8549 (Wno-virtual-inheritance, Wno-namespaces): Same.
8550 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
8551 (Wno-format-overflow, Wno-format-truncation): Same.
8552 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
8553 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
8554 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
8555 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
8556 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
8557 (Wno-misspelled-isr): Same.
8559 2018-06-11 Martin Sebor <msebor@redhat.com>
8561 * PR tree-optimization/86083
8562 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
8564 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
8566 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
8568 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
8571 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
8572 on the correct operand.
8573 (*movdi_internal64): Ditto.
8575 2018-06-11 Martin Liska <mliska@suse.cz>
8577 PR tree-optimization/86089
8578 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
8580 2018-06-11 Julia Koval <julia.koval@intel.com>
8582 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
8583 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
8584 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
8586 2018-06-11 Olivier Hainque <hainque@adacore.com>
8588 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
8589 for Ada with strict dwarf2.
8591 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
8594 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
8597 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
8599 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
8601 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
8603 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
8606 2018-06-08 Martin Liska <mliska@suse.cz>
8608 * tree-cfg.h (debug_function): Fix argument type to match
8611 2018-06-08 Martin Liska <mliska@suse.cz>
8613 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
8614 Remove usage of MPX-related (and removed) fields.
8615 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
8617 2018-06-08 David Malcolm <dmalcolm@redhat.com>
8619 * cfg.c (debug): Use TDF_NONE rather than 0.
8620 * cfghooks.c (debug): Likewise.
8621 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
8622 (struct dump_option_value_info): Convert to...
8623 (struct kv_pair): ...this template type.
8624 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
8626 (optinfo_verbosity_options): Likewise.
8627 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
8629 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
8630 than int for "optgroup_flags" param.
8631 (dump_generic_expr_loc): Use dump_flags_t rather than int for
8633 (dump_dec): Likewise.
8634 (dump_finish): Use TDF_NONE rather than 0.
8635 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
8636 rather than int for "optgroup_flags" param. Use TDF_NONE rather
8637 than 0. Update for change to option_ptr.
8638 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
8639 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
8640 0. Update for changes to optinfo_verbosity_options and
8642 (opt_info_switch_p): Convert optgroup_flags from int to
8644 (dump_basic_block): Use dump_flags_t rather than int
8645 for "dump_kind" param.
8646 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
8647 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
8648 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
8649 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
8650 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
8651 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
8652 TDF_NONE): Convert from macros to...
8653 (enum dump_flag): ...this new enum.
8654 (dump_flags_t): Update to use enum.
8655 (operator|, operator&, operator~, operator|=, operator&=):
8656 Implement for dump_flags_t.
8657 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
8658 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
8659 Convert from macros to...
8660 (enum optgroup_flag): ...this new enum.
8661 (optgroup_flags_t): New typedef.
8662 (operator|, operator|=): Implement for optgroup_flags_t.
8663 (struct dump_file_info): Convert field "alt_flags" to
8664 dump_flags_t. Convert field "optgroup_flags" to
8666 (dump_basic_block): Use dump_flags_t rather than int for param.
8667 (dump_generic_expr_loc): Likewise.
8668 (dump_dec): Likewise.
8669 (dump_register): Convert param "optgroup_flags" to
8671 (opt_info_enable_passes): Likewise.
8672 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
8674 * gimple-pretty-print.c (debug): Likewise.
8675 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
8676 (merged_store_group::apply_stores): Likewise.
8677 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
8678 * gimple.c (verify_gimple_pp): Likewise.
8679 * graphite-poly.c (print_pbb_body): Likewise.
8680 * passes.c (pass_manager::register_one_dump_file): Convert
8681 local "optgroup_flags" to optgroup_flags_t.
8682 * print-tree.c (print_node): Use TDF_NONE rather than 0.
8684 (debug_body): Likewise.
8685 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
8686 to optgroup_flags_t.
8687 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
8689 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
8690 (convert_mult_to_fma): Likewise.
8691 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
8692 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
8693 * tree-vect-data-refs.c (dump_lower_bound): Convert param
8694 "dump_kind" to dump_flags_t.
8696 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8698 * config/rs6000/rs6000.c (min, max): Delete.
8700 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
8702 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
8703 -mabi=spe and -mabi=no-spe.
8705 2018-06-08 Martin Liska <mliska@suse.cz>
8707 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
8708 where we expect an existing summary.
8710 2018-06-08 Martin Liska <mliska@suse.cz>
8712 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
8713 * ipa-inline.h (estimate_edge_growth): Likewise.
8715 2018-06-08 Martin Liska <mliska@suse.cz>
8717 * cgraph.c (function_version_hasher::hash): Use
8718 cgraph_node::get_uid ().
8719 (function_version_hasher::equal):
8720 * cgraph.h (cgraph_node::get_uid): New method.
8721 * ipa-inline.c (update_caller_keys): Use
8722 cgraph_node::get_uid ().
8723 (update_callee_keys): Likewise.
8724 * ipa-utils.c (searchc): Likewise.
8725 (ipa_reduced_postorder): Likewise.
8726 * lto-cgraph.c (input_node): Likewise.
8727 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
8728 * symbol-summary.h (symtab_insertion): Likewise.
8729 (symtab_removal): Likewise.
8730 (symtab_duplication): Likewise.
8731 * tree-pretty-print.c (dump_function_header): Likewise.
8732 * tree-sra.c (convert_callers_for_node): Likewise.
8734 2018-06-08 Martin Liska <mliska@suse.cz>
8736 * cgraph.c (symbol_table::create_edge): Always assign a new
8738 (symbol_table::free_edge): Do not recycle numbers.
8739 * cgraph.h (cgraph_edge::get): New method.
8740 * symbol-summary.h (symtab_removal): Use it.
8741 (symtab_duplication): Likewise.
8742 (call_summary::hashable_uid): Remove.
8744 2018-06-08 Martin Liska <mliska@suse.cz>
8746 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
8747 (initialize_growth_caches): Remove.
8748 (free_growth_caches): Likewise.
8749 (do_estimate_edge_time): Use edge_growth_cache.
8750 (do_estimate_edge_size): Likewise.
8751 (do_estimate_edge_hints): Likewise.
8752 * ipa-inline.c (reset_edge_caches): Likewise.
8753 (recursive_inlining): Likewise.
8754 (inline_small_functions): Likewise.
8755 * ipa-inline.h (initialize_growth_caches): Remove.
8756 (estimate_edge_size): Likewise.
8757 (estimate_edge_time): Likewise.
8758 (estimate_edge_hints): Likewise.
8759 (reset_edge_growth_cache): Likewise.
8760 * symbol-summary.h (call_summary::remove): New method.
8762 2018-06-08 Martin Liska <mliska@suse.cz>
8764 * ipa-cp.c (class edge_clone_summary): New summary.
8765 (grow_edge_clone_vectors): Remove.
8766 (ipcp_edge_duplication_hook): Remove.
8767 (class edge_clone_summary_t): New call_summary class.
8768 (ipcp_edge_removal_hook): Remove.
8769 (edge_clone_summary_t::duplicate): New function.
8770 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
8771 (create_specialized_node): Likewise.
8772 (ipcp_driver): Initialize edge_clone_summaries and do not
8775 2018-06-08 Martin Liska <mliska@suse.cz>
8777 * symbol-summary.h (get): New function.
8778 (call_summary::m_initialize_when_cloning): New class member.
8780 2018-06-08 Martin Liska <mliska@suse.cz>
8782 * cgraph.c (cgraph_node::remove): Do not recycle uid.
8783 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
8784 (symbol_table::allocate_cgraph_symbol): Do not set uid.
8785 * passes.c (uid_hash_t): Record removed_nodes by their uids.
8786 (remove_cgraph_node_from_order): Use the removed_nodes set.
8787 (do_per_function_toporder): Likwise.
8788 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
8789 instead of summary_uid.
8790 (symtab_removal): Likewise.
8791 (symtab_duplication): Likewise.
8793 2018-06-08 Martin Liska <mliska@suse.cz>
8795 * ipa-cp.c (ipcp_store_bits_results): Use
8796 ipcp_transformation_sum.
8797 (ipcp_store_vr_results): Likewise.
8798 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
8800 (ipcp_transformation_initialize): ... this.
8801 (ipa_set_node_agg_value_chain):
8802 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
8803 (write_ipcp_transformation_info): Likewise.
8804 (read_ipcp_transformation_info): Likewise.
8805 (ipcp_update_bits): Likewise.
8806 (ipcp_update_vr): Likewise.
8807 (ipcp_transform_function): Likewise.
8808 * ipa-prop.h: Rename ipcp_transformation_summary to
8809 ipcp_transformation.
8810 (class ipcp_transformation_t): New function summary.
8811 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
8812 (ipa_get_agg_replacements_for_node): Likewise.
8814 2018-06-08 Martin Liska <mliska@suse.cz>
8816 * ipa-pure-const.c (struct funct_state_d): Do it class instead
8818 (class funct_state_summary_t): New function_summary class.
8819 (has_function_state): Remove.
8820 (get_function_state): Likewise.
8821 (set_function_state): Likewise.
8822 (add_new_function): Likewise.
8823 (funct_state_summary_t::insert): New function.
8824 (duplicate_node_data): Remove.
8825 (remove_node_data): Remove.
8826 (funct_state_summary_t::duplicate): New function.
8827 (register_hooks): Create new funct_state_summaries.
8828 (pure_const_generate_summary): Use it.
8829 (pure_const_write_summary): Likewise.
8830 (pure_const_read_summary): Likewise.
8831 (propagate_pure_const): Likewise.
8832 (propagate_nothrow): Likewise.
8833 (dump_malloc_lattice): Likewise.
8834 (propagate_malloc): Likewise.
8835 (execute): Do not register hooks, just remove summary
8837 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
8840 2018-06-08 Martin Liska <mliska@suse.cz>
8842 * ipa-reference.c (remove_node_data): Remove.
8843 (duplicate_node_data): Likewise.
8844 (class ipa_ref_var_info_summary_t): New class.
8845 (class ipa_ref_opt_summary_t): Likewise.
8846 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
8847 (get_reference_optimization_summary): Use
8848 ipa_ref_opt_sum_summaries.
8849 (set_reference_vars_info): Remove.
8850 (set_reference_optimization_summary): Likewise.
8851 (ipa_init): Create summaries.
8852 (init_function_info): Use function summary.
8853 (ipa_ref_opt_summary_t::duplicate): New function.
8854 (ipa_ref_opt_summary_t::remove): New function.
8855 (get_read_write_all_from_node): Fix GNU coding style.
8856 (propagate): Use function summary.
8857 (write_node_summary_p): Fix GNU coding style.
8858 (stream_out_bitmap): Likewise.
8859 (ipa_reference_read_optimization_summary): Use function summary.
8860 (ipa_reference_c_finalize): Do not release hooks.
8862 2018-06-08 Martin Liska <mliska@suse.cz>
8864 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
8865 (analyze_function_body): Extract multiple calls of get_create.
8866 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
8867 * ipa-inline.c (recursive_inlining): Use ::get method.
8868 * ipa-inline.h (estimate_edge_growth): Likewise.
8870 2018-06-08 Martin Liska <mliska@suse.cz>
8872 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
8874 (hsa_function_summary::hsa_function_summary): Use the new enum
8876 (hsa_gpu_implementation_p): Use hsa_summaries::get.
8877 * hsa-gen.c (hsa_get_host_function): Likewise.
8878 (get_brig_function_name): Likewise.
8879 * ipa-hsa.c (process_hsa_functions): Likewise.
8880 (ipa_hsa_write_summary): Likewise.
8881 * symbol-summary.h (symtab_duplication): Use ::get function/
8882 (get): New function.
8884 2018-06-08 Martin Liska <mliska@suse.cz>
8886 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
8888 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
8889 (hsa_register_kernel): Likewise.
8890 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
8891 * hsa-gen.c (hsa_get_host_function): Likewise.
8892 (get_brig_function_name): Likewise.
8893 (generate_hsa): Likewise.
8894 (pass_gen_hsail::execute): Likewise.
8895 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
8896 (devirtualization_time_bonus): Likewise.
8897 (ipcp_propagate_stage): Likewise.
8898 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8899 (edge_set_predicate): Likewise.
8900 (evaluate_conditions_for_known_args): Likewise.
8901 (evaluate_properties_for_edge): Likewise.
8902 (ipa_fn_summary::reset): Likewise.
8903 (ipa_fn_summary_t::duplicate): Likewise.
8904 (dump_ipa_call_summary): Likewise.
8905 (ipa_dump_fn_summary): Likewise.
8906 (analyze_function_body): Likewise.
8907 (compute_fn_summary): Likewise.
8908 (estimate_edge_devirt_benefit): Likewise.
8909 (estimate_edge_size_and_time): Likewise.
8910 (estimate_calls_size_and_time): Likewise.
8911 (estimate_node_size_and_time): Likewise.
8912 (inline_update_callee_summaries): Likewise.
8913 (remap_edge_change_prob): Likewise.
8914 (remap_edge_summaries): Likewise.
8915 (ipa_merge_fn_summary_after_inlining): Likewise.
8916 (ipa_update_overall_fn_summary): Likewise.
8917 (read_ipa_call_summary): Likewise.
8918 (inline_read_section): Likewise.
8919 (write_ipa_call_summary): Likewise.
8920 (ipa_fn_summary_write): Likewise.
8921 (ipa_free_fn_summary): Likewise.
8922 * ipa-hsa.c (process_hsa_functions): Likewise.
8923 (ipa_hsa_write_summary): Likewise.
8924 (ipa_hsa_read_section): Likewise.
8925 * ipa-icf.c (sem_function::merge): Likewise.
8926 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
8927 (do_estimate_edge_time): Likewise.
8928 (estimate_size_after_inlining): Likewise.
8929 (estimate_growth): Likewise.
8930 (growth_likely_positive): Likewise.
8931 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8932 (inline_call): Likewise.
8933 * ipa-inline.c (caller_growth_limits): Likewise.
8934 (can_inline_edge_p): Likewise.
8935 (can_inline_edge_by_limits_p): Likewise.
8936 (compute_uninlined_call_time): Likewise.
8937 (compute_inlined_call_time): Likewise.
8938 (want_inline_small_function_p): Likewise.
8939 (edge_badness): Likewise.
8940 (update_caller_keys): Likewise.
8941 (update_callee_keys): Likewise.
8942 (recursive_inlining): Likewise.
8943 (inline_small_functions): Likewise.
8944 (inline_to_all_callers_1): Likewise.
8945 (dump_overall_stats): Likewise.
8946 (early_inline_small_functions): Likewise.
8947 (early_inliner): Likewise.
8948 * ipa-inline.h (estimate_edge_growth): Likewise.
8949 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8950 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8951 * ipa-prop.h (IPA_NODE_REF): Likewise.
8952 (IPA_EDGE_REF): Likewise.
8953 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8954 (propagate_malloc): Likewise.
8955 * ipa-split.c (execute_split_functions): Likewise.
8956 * symbol-summary.h: Rename get to get_create.
8958 (get_create): Likewise.
8959 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8961 2018-06-08 Martin Liska <mliska@suse.cz>
8963 * symbol-summary.h (release): Move definition out of class
8965 (symtab_removal): Likewise.
8966 (symtab_duplication): Likewise.
8968 2018-06-08 Martin Liska <mliska@suse.cz>
8970 * symbol-summary.h (function_summary): Move constructor
8971 implementation out of class declaration.
8972 (release): Likewise.
8973 (symtab_insertion): Likewise.
8974 (symtab_removal): Likewise.
8975 (symtab_duplication): Likewise.
8978 2018-06-08 Martin Liska <mliska@suse.cz>
8980 * Makefile.in: Remove support for MPX (macros, related functions,
8981 fields in cgraph_node, ...).
8982 * builtin-types.def (BT_BND): Likewise.
8983 (BT_FN_BND_CONST_PTR): Likewise.
8984 (BT_FN_CONST_PTR_BND): Likewise.
8985 (BT_FN_VOID_PTR_BND): Likewise.
8986 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
8987 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
8988 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
8989 (expand_builtin_mempcpy_with_bounds): Likewise.
8990 (expand_builtin_memset_with_bounds): Likewise.
8991 (expand_builtin_memset_args): Likewise.
8992 (std_expand_builtin_va_start): Likewise.
8993 (expand_builtin): Likewise.
8994 (expand_builtin_with_bounds): Likewise.
8995 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
8996 (DEF_LIB_BUILTIN_CHKP): Likewise.
8997 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
8998 (DEF_CHKP_BUILTIN): Likewise.
8999 (BUILT_IN_MEMCPY): Likewise.
9000 (BUILT_IN_MEMMOVE): Likewise.
9001 (BUILT_IN_MEMPCPY): Likewise.
9002 (BUILT_IN_MEMSET): Likewise.
9003 (BUILT_IN_STPCPY): Likewise.
9004 (BUILT_IN_STRCAT): Likewise.
9005 (BUILT_IN_STRCHR): Likewise.
9006 (BUILT_IN_STRCPY): Likewise.
9007 (BUILT_IN_STRLEN): Likewise.
9008 (BUILT_IN_MEMCPY_CHK): Likewise.
9009 (BUILT_IN_MEMMOVE_CHK): Likewise.
9010 (BUILT_IN_MEMPCPY_CHK): Likewise.
9011 (BUILT_IN_MEMSET_CHK): Likewise.
9012 (BUILT_IN_STPCPY_CHK): Likewise.
9013 (BUILT_IN_STRCAT_CHK): Likewise.
9014 (BUILT_IN_STRCPY_CHK): Likewise.
9015 * calls.c (store_bounds): Likewise.
9016 (emit_call_1): Likewise.
9017 (special_function_p): Likewise.
9018 (maybe_warn_nonstring_arg): Likewise.
9019 (initialize_argument_information): Likewise.
9020 (finalize_must_preallocate): Likewise.
9021 (compute_argument_addresses): Likewise.
9022 (expand_call): Likewise.
9023 * cfgexpand.c (expand_call_stmt): Likewise.
9024 (expand_return): Likewise.
9025 (expand_gimple_stmt_1): Likewise.
9026 (pass_expand::execute): Likewise.
9027 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
9028 (cgraph_node::remove): Likewise.
9029 (cgraph_node::dump): Likewise.
9030 (cgraph_node::verify_node): Likewise.
9031 * cgraph.h (chkp_function_instrumented_p): Likewise.
9032 (symtab_node::get_alias_target): Likewise.
9033 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
9034 (cgraph_local_p): Likewise.
9035 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
9036 (cgraph_edge::rebuild_references): Likewise.
9037 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
9038 (walk_polymorphic_call_targets): Likewise.
9039 (cgraph_node::expand_thunk): Likewise.
9040 (symbol_table::output_weakrefs): Likewise.
9041 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
9042 (ix86_handle_option): Likewise.
9043 * config/i386/constraints.md: Likewise.
9044 * config/i386/i386-builtin-types.def (BND): Likewise.
9048 * config/i386/i386-builtin.def (BDESC_END): Likewise.
9049 (BDESC_FIRST): Likewise.
9051 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
9052 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
9053 * config/i386/i386.c (enum reg_class): Likewise.
9054 (ix86_target_string): Likewise.
9055 (ix86_option_override_internal): Likewise.
9056 (ix86_conditional_register_usage): Likewise.
9057 (ix86_valid_target_attribute_inner_p): Likewise.
9058 (ix86_set_indirect_branch_type): Likewise.
9059 (ix86_set_current_function): Likewise.
9060 (ix86_function_arg_regno_p): Likewise.
9061 (init_cumulative_args): Likewise.
9062 (ix86_function_arg_advance): Likewise.
9063 (ix86_function_arg): Likewise.
9064 (ix86_pass_by_reference): Likewise.
9065 (ix86_function_value_regno_p): Likewise.
9066 (ix86_function_value_1): Likewise.
9067 (ix86_function_value_bounds): Likewise.
9068 (ix86_return_in_memory): Likewise.
9069 (ix86_setup_incoming_vararg_bounds): Likewise.
9070 (ix86_va_start): Likewise.
9071 (indirect_thunk_need_prefix): Likewise.
9072 (print_reg): Likewise.
9073 (ix86_print_operand): Likewise.
9074 (ix86_expand_call): Likewise.
9075 (ix86_output_function_return): Likewise.
9076 (reg_encoded_number): Likewise.
9077 (BDESC_VERIFYS): Likewise.
9078 (ix86_init_mpx_builtins): Likewise.
9079 (ix86_init_builtins): Likewise.
9080 (ix86_emit_cmove): Likewise.
9081 (ix86_emit_move_max): Likewise.
9082 (ix86_expand_builtin): Likewise.
9083 (ix86_builtin_mpx_function): Likewise.
9084 (ix86_get_arg_address_for_bt): Likewise.
9085 (ix86_load_bounds): Likewise.
9086 (ix86_store_bounds): Likewise.
9087 (ix86_load_returned_bounds): Likewise.
9088 (ix86_store_returned_bounds): Likewise.
9089 (ix86_class_likely_spilled_p): Likewise.
9090 (ix86_hard_regno_mode_ok): Likewise.
9091 (x86_order_regs_for_local_alloc): Likewise.
9092 (ix86_mitigate_rop): Likewise.
9093 (ix86_bnd_prefixed_insn_p): Likewise.
9094 (ix86_mpx_bound_mode): Likewise.
9095 (ix86_make_bounds_constant): Likewise.
9096 (ix86_initialize_bounds): Likewise.
9097 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
9098 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
9099 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
9100 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
9101 (TARGET_CHKP_BOUND_MODE): Likewise.
9102 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
9103 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
9104 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
9105 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
9106 * config/i386/i386.h (TARGET_MPX): Likewise.
9107 (TARGET_MPX_P): Likewise.
9108 (VALID_BND_REG_MODE): Likewise.
9109 (FIRST_BND_REG): Likewise.
9110 (LAST_BND_REG): Likewise.
9111 (enum reg_class): Likewise.
9112 (BND_REG_P): Likewise.
9113 (BND_REGNO_P): Likewise.
9114 (BNDmode): Likewise.
9115 (ADJUST_INSN_LENGTH): Likewise.
9116 * config/i386/i386.md: Likewise.
9117 * config/i386/i386.opt: Likewise.
9118 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
9119 (defined): Likewise.
9120 (LINK_MPX): Likewise.
9121 (MPX_SPEC): Likewise.
9122 (LIBMPX_SPEC): Likewise.
9123 (LIBMPXWRAPPERS_SPEC): Likewise.
9124 (CHKP_SPEC): Likewise.
9125 * config/i386/predicates.md: Likewise.
9126 * dbxout.c (dbxout_type): Likewise.
9127 * doc/extend.texi: Likewise.
9128 * doc/invoke.texi: Likewise.
9129 * doc/md.texi: Likewise.
9130 * doc/tm.texi: Likewise.
9131 * doc/tm.texi.in: Likewise.
9132 * dwarf2out.c (is_base_type): Likewise.
9133 (gen_formal_types_die): Likewise.
9134 (gen_subprogram_die): Likewise.
9135 (gen_type_die_with_usage): Likewise.
9136 (gen_decl_die): Likewise.
9137 (dwarf2out_late_global_decl): Likewise.
9138 * expr.c (expand_assignment): Likewise.
9139 (emit_storent_insn): Likewise.
9140 (store_expr_with_bounds): Likewise.
9141 (store_expr): Likewise.
9142 (expand_expr_real_1): Likewise.
9143 * expr.h (store_expr_with_bounds): Likewise.
9144 * function.c (use_register_for_decl): Likewise.
9145 (struct bounds_parm_data): Likewise.
9146 (assign_parms_augmented_arg_list): Likewise.
9147 (assign_parm_find_entry_rtl): Likewise.
9148 (assign_parm_is_stack_parm): Likewise.
9149 (assign_parm_load_bounds): Likewise.
9150 (assign_bounds): Likewise.
9151 (assign_parms): Likewise.
9152 (expand_function_start): Likewise.
9153 * gcc.c (CHKP_SPEC): Likewise.
9154 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
9155 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
9156 (wrestrict_dom_walker::check_call): Likewise.
9157 * gimple.c (gimple_build_call_from_tree): Likewise.
9158 * gimple.h (enum gf_mask): Likewise.
9159 (gimple_call_with_bounds_p): Likewise.
9160 (gimple_call_set_with_bounds): Likewise.
9161 * gimplify.c (gimplify_init_constructor): Likewise.
9162 * ipa-cp.c (initialize_node_lattices): Likewise.
9163 (propagate_constants_across_call): Likewise.
9164 (find_more_scalar_values_for_callers_subset): Likewise.
9165 * ipa-hsa.c (process_hsa_functions): Likewise.
9166 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
9167 * ipa-icf.c (sem_function::merge): Likewise.
9168 * ipa-inline.c (early_inliner): Likewise.
9169 * ipa-pure-const.c (warn_function_noreturn): Likewise.
9170 (warn_function_cold): Likewise.
9171 (propagate_pure_const): Likewise.
9172 * ipa-ref.h (enum GTY): Likewise.
9173 * ipa-split.c (find_retbnd): Likewise.
9174 (consider_split): Likewise.
9175 (split_function): Likewise.
9176 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9177 * ipa.c (walk_polymorphic_call_targets): Likewise.
9178 (symbol_table::remove_unreachable_nodes): Likewise.
9179 (process_references): Likewise.
9180 (cgraph_build_static_cdtor_1): Likewise.
9181 * lto-cgraph.c (lto_output_node): Likewise.
9182 (output_refs): Likewise.
9183 (compute_ltrans_boundary): Likewise.
9184 (input_overwrite_node): Likewise.
9185 (input_node): Likewise.
9186 (input_cgraph_1): Likewise.
9187 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
9188 * passes.c (pass_manager::execute_early_local_passes): Likewise.
9189 (class pass_chkp_instrumentation_passes): Likewise.
9190 (make_pass_chkp_instrumentation_passes): Likewise.
9191 * passes.def: Likewise.
9192 * rtl.h (struct GTY): Likewise.
9193 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
9194 * stor-layout.c (layout_type): Likewise.
9195 * symtab.c: Likewise.
9196 * target.def: Likewise.
9197 * targhooks.c (default_chkp_bound_type): Likewise.
9198 (default_chkp_bound_mode): Likewise.
9199 (default_builtin_chkp_function): Likewise.
9200 (default_chkp_function_value_bounds): Likewise.
9201 (default_chkp_make_bounds_constant): Likewise.
9202 (default_chkp_initialize_bounds): Likewise.
9203 * targhooks.h (default_chkp_bound_type): Likewise.
9204 (default_chkp_bound_mode): Likewise.
9205 (default_builtin_chkp_function): Likewise.
9206 (default_chkp_function_value_bounds): Likewise.
9207 (default_chkp_make_bounds_constant): Likewise.
9208 (default_chkp_initialize_bounds): Likewise.
9209 * toplev.c (compile_file): Likewise.
9210 (process_options): Likewise.
9211 * tree-core.h (DEF_BUILTIN): Likewise.
9212 (DEF_BUILTIN_CHKP): Likewise.
9213 * tree-inline.c (declare_return_variable): Likewise.
9214 (remap_gimple_stmt): Likewise.
9215 (copy_bb): Likewise.
9216 (initialize_inlined_parameters): Likewise.
9217 (expand_call_inline): Likewise.
9218 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
9219 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
9220 (make_pass_ipa_chkp_produce_thunks): Likewise.
9221 (make_pass_chkp): Likewise.
9222 (make_pass_chkp_opt): Likewise.
9223 (make_pass_chkp_instrumentation_passes): Likewise.
9224 * tree-pretty-print.c (dump_generic_node): Likewise.
9225 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
9226 * tree-ssa-dce.c (propagate_necessity): Likewise.
9227 (eliminate_unnecessary_stmts): Likewise.
9228 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
9229 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
9230 * tree-ssa-sccvn.h: Likewise.
9231 * tree-ssa-strlen.c (get_string_length): Likewise.
9232 (valid_builtin_call): Likewise.
9233 (adjust_last_stmt): Likewise.
9234 (handle_builtin_strchr): Likewise.
9235 (handle_builtin_strcpy): Likewise.
9236 (handle_builtin_stxncpy): Likewise.
9237 (handle_builtin_memcpy): Likewise.
9238 (handle_builtin_strcat): Likewise.
9239 (strlen_check_and_optimize_stmt): Likewise.
9240 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
9241 * tree-streamer-in.c: Likewise.
9242 * tree-streamer.c (record_common_node): Likewise.
9243 * tree.c (tree_code_size): Likewise.
9244 (wide_int_to_tree_1): Likewise.
9245 (type_contains_placeholder_1): Likewise.
9246 (build_common_tree_nodes): Likewise.
9247 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
9248 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
9249 (POINTER_BOUNDS_P): Likewise.
9250 (BOUNDED_TYPE_P): Likewise.
9251 (BOUNDED_P): Likewise.
9252 (CALL_WITH_BOUNDS_P): Likewise.
9253 (pointer_bounds_type_node): Likewise.
9254 * value-prof.c (gimple_ic): Likewise.
9255 * var-tracking.c (vt_add_function_parameters): Likewise.
9256 * varasm.c (make_decl_rtl): Likewise.
9257 (assemble_start_function): Likewise.
9258 (output_constant): Likewise.
9259 (maybe_assemble_visibility): Likewise.
9260 * varpool.c (ctor_for_folding): Likewise.
9261 * chkp-builtins.def: Remove.
9262 * ipa-chkp.c: Remove.
9263 * ipa-chkp.h: Remove.
9264 * rtl-chkp.c: Remove.
9265 * rtl-chkp.h: Remove.
9266 * tree-chkp-opt.c: Remove.
9267 * tree-chkp.c: Remove.
9268 * tree-chkp.h: Remove.
9270 2018-06-07 Carl Love <cel@us.ibm.com>
9272 * config/rs6000/vsx.md (vextract_fp_from_shorth,
9273 vextract_fp_from_shortl): Add BE support.
9275 2018-06-07 Paul Koning <ni1d@arrl.net>
9277 * compare-elim.c (try_merge_compare): Don't merge compare if
9278 address contains a side effect.
9279 (try_eliminate_compare): Likewise.
9281 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
9283 * config.gcc: Support "tremont".
9284 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
9285 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9287 * config/i386/i386.c (m_TREMONT): Define.
9288 (processor_target_table): Add "tremont".
9289 (PTA_TREMONT): Define.
9290 (ix86_lea_outperforms): Add TARGET_TREMONT.
9291 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
9292 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
9293 and M_INTEL_GOLDMONT_PLUS.
9294 (fold_builtin_cpu): Add "tremont".
9295 (ix86_add_stmt_cost): Add TARGET_TREMONT.
9296 (ix86_option_override_internal): Add "tremont".
9297 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
9298 (processor_type): Add PROCESSOR_TREMONT.
9299 * config/i386/x86-tune.def: Add m_TREMONT.
9300 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
9302 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9304 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
9305 symbol defined for msp430i* devices to be lower case.
9307 2018-06-07 Richard Biener <rguenther@suse.de>
9309 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
9310 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
9311 Properly wrap signed arithmetic if overflow wraps.
9313 2018-06-07 Jakub Jelinek <jakub@redhat.com>
9315 PR tree-optimization/69615
9316 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
9317 of a cast from a same precision integral SSA_NAME in a bb dominated
9318 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
9319 cast to utype if rhs2 has already a compatible type.
9321 2018-06-07 Richard Biener <rguenther@suse.de>
9323 PR tree-optimization/85935
9324 * graphite-scop-detection.c (find_params_in_bb): Analyze
9325 condition operands with respect to the correct loop. Assert
9326 the analysis doesn't fail.
9328 2018-06-04 Carl Love <cel@us.ibm.com>
9330 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
9331 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
9332 as it is slightly cheaper.
9333 (first_match_or_eos_index_<mode>):
9334 Calculate index using natural element order.
9335 (first_match_index_<mode>):
9336 Calculate index using natural element order.
9337 (first_match_or_eos_index_<mode>):
9338 Calculate index using natural order.
9339 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
9340 for BE and LE modes.
9341 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
9342 P9V_BUILTIN_VCLZLSBB_V16QI.
9343 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
9346 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9348 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
9349 indentation and line wrap for many prototypes. Add missing
9350 @smallexample directives around block of prototypes for vec_xl and
9353 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
9355 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
9356 track if we pass or return IEEE 128-bit floating point.
9357 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
9358 C++ mangling that is compatible with GCC 8.1.
9359 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
9360 (init_cumulative_args): Note if we pass or return IEEE 128-bit
9361 floating point types.
9362 (rs6000_function_arg_advance_1): Likewise.
9363 (rs6000_mangle_type): Optionally generate mangled names that match
9364 what GCC 8.1 generated for IEEE 128-bit floating point types.
9365 (rs6000_globalize_decl_name): If we have an external function that
9366 passes or returns IEEE 128-bit types, generate a weak reference
9367 from the mangled name used in GCC 8.1 to the current mangled
9369 (rs6000_init_builtins): Make __ibm128 use the long double type if
9370 long double is IBM extended double. Make __float128 use the long
9371 double type if long double is IEEE 128-bit.
9374 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
9375 macro for __ibm128 built-in functions.
9376 (PACK_IF): Add __ibm128 pack/unpack functions.
9377 (UNPACK_IF): Likewise.
9378 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
9379 enable long double built-in functions if long double is IEEE
9380 128-bit floating point.
9381 (rs6000_invalid_builtin): Update long double built-in function
9383 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
9384 functions, adjust the built-in function to use the long double
9385 built-in function if __ibm128 and long double are the same type.
9386 * doc/extend.texi (PowerPC builtins): Update documention for
9387 __builtin_{,un}pack_longdouble. Add documentation for
9388 __builtin_{,un}pack_ibm128.
9390 2018-06-06 Jim Wilson <jimw@sifive.com>
9392 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
9393 (struct machine_function): New field interrupt_mode.
9394 (riscv_handle_type_attribute): New function. Add forward declaration.
9395 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
9396 (riscv_expand_epilogue): Check interrupt_mode field.
9397 (riscv_set_current_function): Check interrupt attribute args and
9398 set interrupt_mode field.
9399 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
9400 (riscv_sret, riscv_uret): New.
9401 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
9402 new arguments to interrupt attribute.
9404 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
9407 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
9408 Don't handle -mcpu=power8 if -mpower9-vector is also used.
9410 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
9412 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
9413 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
9414 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9415 several redundant entries.
9417 2018-06-06 David Malcolm <dmalcolm@redhat.com>
9419 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
9420 type from "rtx" to "rtx_insn *".
9421 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
9422 for local "call_insn", removing cast.
9423 (ix86_expand_call): Likewise, introducing a "call_insn" local.
9425 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
9427 PR tree-optimization/86066
9428 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
9429 for BIT_INSERT_EXPR stores.
9431 2018-06-06 Richard Biener <rguenther@suse.de>
9433 PR tree-optimization/86062
9434 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9435 component refs ontop
9436 of to be offsetted base.
9438 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9440 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
9441 to be static and remove check on interrupt attribute name.
9443 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
9445 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9446 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
9448 2018-06-05 Steve Ellcey <sellcey@cavium.com>
9451 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
9453 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
9454 Remove second argument, change how error is called.
9455 (aarch64_layout_arg): Remove second argument from
9456 aarch64_err_no_fpadvsimd call.
9457 (aarch64_init_cumulative_args): Ditto.
9458 (aarch64_gimplify_va_arg_expr): Ditto.
9459 * config/aarch64/aarch64.md (mov<mode>): Ditto.
9461 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
9463 * config/i386/i386.md (simple_return_indirect_internal): New expander.
9464 (*simple_return_indirect_internal<mode>): Rename from
9465 simple_return_indirect_internal. Use W mode iterator.
9466 (rstorssp): New expander.
9467 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
9468 (clrssbsy): New expander.
9469 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
9471 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
9473 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
9475 (cmse_check_pointed_object): Likewise.
9477 2018-06-05 Martin Liska <mliska@suse.cz>
9479 PR gcov-profile/47618
9480 * doc/invoke.texi: Document how -fprofile-dir format
9483 2018-06-05 Richard Biener <rguenther@suse.de>
9485 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
9486 removal pretend DOM info isn't available so we do not update
9487 it and only remove edges, not dominated blocks. Actually free
9488 DOM info in case we removed something. Remove unreachable blocks.
9489 (mfb_keep_latches): Work with either DOM info or marked backedges.
9490 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9491 first. Mark backedges if DOM info isn't available.
9492 (Re-)compute DOM info after cleanup_control_flow_pre.
9494 2018-06-05 Richard Biener <rguenther@suse.de>
9496 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
9497 (locus_discrim_hasher::hash): Adjust.
9498 (locus_discrim_hasher::equal): Likewise.
9499 (next_discriminator_for_locus): Work on line directly.
9500 (same_line_p): Pass in expanded locus1 as well.
9501 (assign_discriminators): Avoid redundant location expansions.
9503 2018-06-05 Richard Biener <rguenther@suse.de>
9505 PR tree-optimization/86046
9506 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
9507 if required after clearing TREE_ADDRESSABLE.
9509 2018-06-05 Richard Biener <rguenther@suse.de>
9511 PR tree-optimization/86047
9512 * tree-ssa-loop.c (for_each_index): Glob handling of all
9513 decls and constants and really handle all of them.
9515 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9518 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
9519 qualifier_void_pointer and qualifier_const_void_pointer.
9520 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
9521 (arm_init_builtins): Handle the above.
9522 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9523 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
9524 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
9527 2018-06-05 Martin Liska <mliska@suse.cz>
9529 * auto-profile.c (read_autofdo_file): Do not use
9530 gcov_ctr_summary struct.
9531 (afdo_callsite_hot_enough_for_early_inline): Likewise.
9532 * coverage.c (struct counts_entry): Likewise.
9533 (read_counts_file): Read just single summary entry.
9534 (get_coverage_counts): Use gcov_summary struct.
9535 * coverage.h (get_coverage_counts): Likewise.
9536 * gcov-dump.c (dump_working_sets): Likewise.
9537 (tag_summary): Dump just single summary.
9538 * gcov-io.c (gcov_write_summary): Write just histogram
9540 (gcov_read_summary): Read just single summary.
9541 (compute_working_sets): Use gcov_summary struct.
9542 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
9543 of GCOV_COUNTERS_SUMMABLE.
9544 (GCOV_COUNTERS_SUMMABLE): Remove.
9545 (GCOV_FIRST_VALUE_COUNTER): Replace with
9546 GCOV_COUNTER_V_INTERVAL.
9547 (struct gcov_ctr_summary): Remove.
9548 (struct gcov_summary): Directly use fields of former
9550 (compute_working_sets): Use gcov_summary struct.
9551 * gcov.c (read_count_file): Do not use ctrs fields.
9552 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
9554 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
9556 * profile.c: Likewise.
9557 * profile.h: Likewise.
9559 2018-06-05 Martin Liska <mliska@suse.cz>
9561 PR gcov-profile/84846
9562 * gcov.c (output_lines): Print working directory only
9563 in intermediate format.
9565 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
9567 * config/s390/s390-builtin-types.def: Add void function type.
9568 * config/s390/s390-builtins.def: Use the function type for the
9571 2018-06-04 Jim Wilson <jimw@sifive.com>
9573 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
9575 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
9576 and maybe_eh_return. Change regno to unsigned int. Use new args to
9577 handle EH_RETURN_DATA_REGNO registers properly.
9578 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9579 (riscv_expand_epilogue): Update comment. Change argument name and
9580 type. Update code to use new name and type. Pass new args to
9581 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
9583 * config/riscv/riscv.md (NORMAL_RETURN): New.
9584 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
9585 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
9586 (eh_return): Call gen_eh_return_internal and emit barrier.
9587 (eh_return_internal): Call riscv_expand_epilogue.
9589 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
9591 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
9592 bit_insertion field and declare can_be_merged_into method.
9593 (merged_store_group::can_be_merged_into): New method.
9594 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
9595 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
9596 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
9598 2018-06-04 Richard Biener <rguenther@suse.de>
9600 PR tree-optimization/85955
9601 * builtins.c (fold_builtin_sincos): Convert pointers to
9602 destination to appropriate type before dereferencing.
9604 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
9606 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
9608 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
9610 * expr.c (expand_expr_real_1): Force the operand into memory if
9611 its TYPE_MODE is BLKmode and if there is no integer mode for
9612 the number of bits being extracted.
9614 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9618 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9619 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
9621 2018-06-04 Richard Biener <rguenther@suse.de>
9623 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
9624 (cleanup_tree_cfg_noloop): ... single caller. Do
9625 start_recording_case_labels later.
9627 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9629 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
9630 to _IMMINTRIN_H_INCLUDED.
9631 * config/i386/pconfigintrin.h: Ditto.
9632 * config/i386/waitpkgintrin.h: Ditto.
9633 * config/i386/immintrin.h: Add includes for sgxintrin.h,
9634 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
9635 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
9636 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
9637 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
9638 waitpkgintrin.h and cldemoteintrin.h.
9640 2018-06-04 Richard Biener <rguenther@suse.de>
9642 PR tree-optimization/86038
9643 * tracer.c (find_best_successor): Check probability for
9644 being initialized, bail out if not.
9646 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
9649 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
9650 of bits to ignore when comparing architectures.
9652 2018-06-04 Jakub Jelinek <jakub@redhat.com>
9654 PR tree-optimization/69615
9655 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
9656 maximum or minimum of the type, try to merge it also as if
9657 range1 is + [-, x - 1] or + [x + 1, -].
9660 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
9662 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
9664 PR tree-optimization/86034
9665 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
9666 the unsigned bitfield type in a bit insertion sequence if it does not
9667 have a larger precision than the bitfield size.
9668 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
9670 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
9672 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
9674 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9676 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
9677 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
9678 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
9679 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
9681 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
9683 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
9684 Disable -fdelete-null-pointer-checks for ELF toolchain.
9686 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9687 Kito Cheng <kito.cheng@gmail.com>
9689 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
9690 (nds32le-*-*, nds32be-*-*): Integrate checking process.
9691 (nds32*-*-*): Add glibc and uclibc conditions.
9692 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
9693 (TARGET_EXCEPT_UNWIND_INFO): Define.
9694 * config/nds32/elf.h: New file.
9695 * config/nds32/linux.h: New file.
9696 * config/nds32/nds32-elf.opt: New file.
9697 * config/nds32/nds32-linux.opt: New file.
9698 * config/nds32/nds32-fp-as-gp.c
9699 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
9700 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
9702 (nds32_asm_file_end): Ditto.
9703 (nds32_print_operand): Ditto.
9704 (nds32_insert_attributes): Ditto.
9705 (nds32_init_libfuncs): New function.
9706 (TARGET_HAVE_TLS): Define.
9707 (TARGET_INIT_LIBFUNCS): Define.
9708 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
9710 (TARGET_ELF): Apply different mcmodel setting.
9711 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
9712 been migrated into elf.h and linux.h files.
9713 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
9714 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
9715 (mcmodel): The content has been migrated into nds32-elf.opt and
9716 nds32-linux.opt files.
9717 * config/nds32/t-elf: New file.
9718 * config/nds32/t-linux: New file.
9720 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
9721 Shiva Chen <shiva0217@gmail.com>
9723 * config/nds32/constants.md (unspec_volatile_element): Add
9724 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
9725 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
9727 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
9728 (make_pass_nds32_fp_as_gp): Declare.
9729 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
9731 (nds32_asm_function_end_prologue): Remove unused asm output.
9732 (nds32_asm_function_begin_epilogue): Remove unused asm output.
9733 (nds32_asm_file_start): Output necessary fp_as_gp information.
9734 (nds32_option_override): Adjust register usage.
9735 (nds32_expand_prologue): Consider fp_as_gp situation.
9736 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
9737 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
9740 (simple_return): Ditto.
9741 (omit_fp_begin): Output special directive for fp_as_gp.
9742 (omit_fp_end): Output special directive for fp_as_gp.
9743 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
9744 mforbid-fp-as-gp): New options.
9746 2018-06-01 Mark Wielaard <mark@klomp.org>
9748 * dwarf2out.c (dwarf2out_finish): Remove generation of
9749 DW_AT_loclists_base.
9751 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
9753 * gimple-ssa-store-merging.c: Include gimple-fold.h.
9754 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
9755 (struct merged_store_group): Add bit_insertion field.
9756 (dump_char_array): Use standard hexadecimal format.
9757 (merged_store_group::merged_store_group): Set bit_insertion to false.
9758 (merged_store_group::apply_stores): Use optimal buffer size. Deal
9759 with BIT_INSERT_EXPR stores. Move up code updating the mask and
9760 also print the mask in the dump file.
9761 (pass_store_merging::gate): Minor tweak.
9762 (imm_store_chain_info::coalesce_immediate): Fix wrong association
9763 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
9764 stores with INTEGER_CST stores.
9765 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
9766 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
9767 and use it throughout. Generate bit insertion sequences if need be.
9768 (pass_store_merging::process_store): Remove redundant condition.
9769 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
9771 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
9773 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
9774 the 128-bit floating point types. Fix function comment.
9776 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9778 * config/aarch64/aarch64-simd.md
9779 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
9781 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
9784 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9786 PR tree-optimization/85989
9787 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
9789 (backprop::intersect_uses): Check it when deciding whether this
9790 is a backedge reference.
9791 (backprop::process_block): Add each phi to m_visited_phis
9792 after visiting it, then clear it at the end.
9794 2018-06-01 Richard Biener <rguenther@suse.de>
9796 * tree-vectorizer.h (vect_dr_stmt): New function.
9797 (vect_get_load_cost): Adjust.
9798 (vect_get_store_cost): Likewise.
9799 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
9800 Use vect_dr_stmt instead of DR_SMTT.
9801 (vect_record_base_alignments): Likewise.
9802 (vect_calculate_target_alignment): Likewise.
9803 (vect_compute_data_ref_alignment): Likewise and make static.
9804 (vect_update_misalignment_for_peel): Likewise.
9805 (vect_verify_datarefs_alignment): Likewise.
9806 (vector_alignment_reachable_p): Likewise.
9807 (vect_get_data_access_cost): Likewise. Pass down
9808 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
9809 (vect_get_peeling_costs_all_drs): Likewise.
9810 (vect_peeling_hash_get_lowest_cost): Likewise.
9811 (vect_enhance_data_refs_alignment): Likewise.
9812 (vect_find_same_alignment_drs): Likewise.
9813 (vect_analyze_data_refs_alignment): Likewise.
9814 (vect_analyze_group_access_1): Likewise.
9815 (vect_analyze_group_access): Likewise.
9816 (vect_analyze_data_ref_access): Likewise.
9817 (vect_analyze_data_ref_accesses): Likewise.
9818 (vect_vfa_segment_size): Likewise.
9819 (vect_small_gap_p): Likewise.
9820 (vectorizable_with_step_bound_p): Likewise.
9821 (vect_prune_runtime_alias_test_list): Likewise.
9822 (vect_analyze_data_refs): Likewise.
9823 (vect_supportable_dr_alignment): Likewise.
9824 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
9825 (vect_gen_prolog_loop_niters): Likewise.
9826 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
9827 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
9829 (vect_recog_mask_conversion_pattern): Likewise.
9830 (vect_try_gather_scatter_pattern): Likewise.
9831 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
9832 to vect_get_store_cost.
9833 (vect_get_store_cost): Get stmt_info instead of DR.
9834 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
9835 (vect_get_load_cost): Get stmt_info instead of DR.
9837 2018-06-01 Richard Biener <rguenther@suse.de>
9840 * gimple-fold.c (var_decl_component_p): Also allow offsetted
9841 vars wrapped in MEM_REFs.
9843 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
9845 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9846 Fix subreg tests so that we only return a choice between
9847 GENERAL_REGS and FP_REGS if the original classes included both.
9849 2018-06-01 Richard Biener <rguenther@suse.de>
9852 * tree-ssa-structalias.c (get_function_part_constraint):
9853 Handle NULL fi->decl.
9854 (find_func_aliases_for_call): Properly handle indirect
9855 fi from direct call.
9856 (find_func_clobbers): Likewise.
9857 (ipa_pta_execute): Likewise.
9858 (create_variable_info_for): For functions that are ifunc_resolver
9859 resolve to a varinfo that contains the result of the resolver call.
9860 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
9863 2018-05-31 Michael Collison <michael.collison@arm.com>
9865 * config/aarch64/aarch64.md:
9866 (*fix_to_zero_extenddfdi2): New pattern.
9867 * gcc.target/aarch64/fix_extend1.c: New testcase.
9869 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
9873 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
9874 and BUILT_IN_STRNCMP_EQ.
9875 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
9876 BUILT_IN_STRNCMP_EQ.
9877 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
9878 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
9879 (gimple_fold_builtin): Likewise.
9880 * tree-ssa-strlen.c (compute_string_length): New function.
9881 (determine_min_obsize): New function.
9882 (handle_builtin_string_cmp): New function to handle calls to
9883 string compare functions.
9884 (strlen_optimize_stmt): Add handling to builtin string compare
9886 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
9887 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
9888 * tree.c (build_common_builtin_nodes): Add new defines of
9889 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
9891 2018-05-31 Jakub Jelinek <jakub@redhat.com>
9894 * bb-reorder.c (pass_partition_blocks::gate): Return false for
9895 functions with naked attribute.
9897 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
9899 * config/i386/sse.md (avx_vec_concat<mode>):
9900 Substitute concat_tg_mode mode attribute with xtg_mode.
9901 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
9902 (concat_tg_mode): Remove mode attribute.
9904 2018-05-31 Martin Sebor <msebor@redhat.com>
9907 * calls.c (alloc_max_size): Correct a logic error/typo.
9908 Treat excessive arguments as infinite. Warn for invalid arguments.
9909 * doc/invoke.texi (-Walloc-size-larger-than): Update.
9911 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
9914 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
9915 and movx for Haswell.
9917 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
9918 Cesar Philippidis <cesar@codesourcery.com>
9921 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
9922 when emitting error on private/firstprivate reductions.
9923 * omp-low.c (lower_omp_target): Avoid reference-type processing
9924 on pointers for firstprivate clause.
9926 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
9928 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
9931 * config/aarch64/aarch64-simd.md
9932 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
9933 (aarch64_ld1_x3_<mode>): Likewise
9934 (aarch64_st1x2<VALLDIF:mode>): Likewise
9935 (aarch64_st1_x2_<mode>): Likewise
9936 (aarch64_st1x3<VALLDIF:mode>): Likewise
9937 (aarch64_st1_x3_<mode>): Likewise
9938 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
9939 (vld1_s8_x3): Likewise.
9940 (vld1_u16_x3): Likewise.
9941 (vld1_s16_x3): Likewise.
9942 (vld1_u32_x3): Likewise.
9943 (vld1_s32_x3): Likewise.
9944 (vld1_u64_x3): Likewise.
9945 (vld1_s64_x3): Likewise.
9946 (vld1_f16_x3): Likewise.
9947 (vld1_f32_x3): Likewise.
9948 (vld1_f64_x3): Likewise.
9949 (vld1_p8_x3): Likewise.
9950 (vld1_p16_x3): Likewise.
9951 (vld1_p64_x3): Likewise.
9952 (vld1q_u8_x3): Likewise.
9953 (vld1q_s8_x3): Likewise.
9954 (vld1q_u16_x3): Likewise.
9955 (vld1q_s16_x3): Likewise.
9956 (vld1q_u32_x3): Likewise.
9957 (vld1q_s32_x3): Likewise.
9958 (vld1q_u64_x3): Likewise.
9959 (vld1q_s64_x3): Likewise.
9960 (vld1q_f16_x3): Likewise.
9961 (vld1q_f32_x3): Likewise.
9962 (vld1q_f64_x3): Likewise.
9963 (vld1q_p8_x3): Likewise.
9964 (vld1q_p16_x3): Likewise.
9965 (vld1q_p64_x3): Likewise.
9966 (vst1_s64_x2): Likewise.
9967 (vst1_u64_x2): Likewise.
9968 (vst1_f64_x2): Likewise.
9969 (vst1_s8_x2): Likewise.
9970 (vst1_p8_x2): Likewise.
9971 (vst1_s16_x2): Likewise.
9972 (vst1_p16_x2): Likewise.
9973 (vst1_s32_x2): Likewise.
9974 (vst1_u8_x2): Likewise.
9975 (vst1_u16_x2): Likewise.
9976 (vst1_u32_x2): Likewise.
9977 (vst1_f16_x2): Likewise.
9978 (vst1_f32_x2): Likewise.
9979 (vst1_p64_x2): Likewise.
9980 (vst1q_s8_x2): Likewise.
9981 (vst1q_p8_x2): Likewise.
9982 (vst1q_s16_x2): Likewise.
9983 (vst1q_p16_x2): Likewise.
9984 (vst1q_s32_x2): Likewise.
9985 (vst1q_s64_x2): Likewise.
9986 (vst1q_u8_x2): Likewise.
9987 (vst1q_u16_x2): Likewise.
9988 (vst1q_u32_x2): Likewise.
9989 (vst1q_u64_x2): Likewise.
9990 (vst1q_f16_x2): Likewise.
9991 (vst1q_f32_x2): Likewise.
9992 (vst1q_f64_x2): Likewise.
9993 (vst1q_p64_x2): Likewise.
9994 (vst1_s64_x3): Likewise.
9995 (vst1_u64_x3): Likewise.
9996 (vst1_f64_x3): Likewise.
9997 (vst1_s8_x3): Likewise.
9998 (vst1_p8_x3): Likewise.
9999 (vst1_s16_x3): Likewise.
10000 (vst1_p16_x3): Likewise.
10001 (vst1_s32_x3): Likewise.
10002 (vst1_u8_x3): Likewise.
10003 (vst1_u16_x3): Likewise.
10004 (vst1_u32_x3): Likewise.
10005 (vst1_f16_x3): Likewise.
10006 (vst1_f32_x3): Likewise.
10007 (vst1_p64_x3): Likewise.
10008 (vst1q_s8_x3): Likewise.
10009 (vst1q_p8_x3): Likewise.
10010 (vst1q_s16_x3): Likewise.
10011 (vst1q_p16_x3): Likewise.
10012 (vst1q_s32_x3): Likewise.
10013 (vst1q_s64_x3): Likewise.
10014 (vst1q_u8_x3): Likewise.
10015 (vst1q_u16_x3): Likewise.
10016 (vst1q_u32_x3): Likewise.
10017 (vst1q_u64_x3): Likewise.
10018 (vst1q_f16_x3): Likewise.
10019 (vst1q_f32_x3): Likewise.
10020 (vst1q_f64_x3): Likewise.
10021 (vst1q_p64_x3): Likewise.
10023 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10025 * config/msp430/msp430.c (msp430_output_labelref): Prepend
10026 user_label_prefix to name.
10028 * tree-core.h: Update comment about the format of NAME string
10029 passed to handler in attribute_spec.
10031 * config/msp430/msp430.md: Remove erroneous subreg expression from
10032 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
10033 zero_extend{q,h}isi2.
10035 2018-05-30 Borislav Petkov <bp@suse.de>
10037 * doc/extend.texi: Document some architecture specific
10038 constraints and sort entries.
10040 2018-05-30 Martin Sebor <msebor@redhat.com>
10042 PR middle-end/85369
10043 * builtins.c (expand_builtin_stpcpy_1): New function.
10044 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
10045 only if the former succeeds.
10047 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10049 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
10052 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10054 * doc/invoke.texi (-flinker-output): Document
10056 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10058 * passes.c (ipa_write_summaries): Only modify statements if body
10060 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
10061 incrementally linking.
10062 (ipa_passes): Likewise.
10063 * lto-cgraph.c (lto_output_node): When incrementally linking do not
10064 pass down resolution info.
10065 * common.opt (flag_incremental_link): Update info.
10066 * gcc.c (plugin specs): Turn flinker-output=* to
10067 -plugin-opt=-linker-output-known
10068 * toplev.c (compile_file): Also cut compilation when doing incremental
10070 * flag-types. (enum lto_partition_model): Add
10071 LTO_LINKER_OUTPUT_NOLTOREL.
10072 (invoke.texi): Add -flinker-output docs.
10073 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
10074 link same way as WPA; do not stream in dead initializers.
10076 * dwarf2out.c (dwarf2out_die_ref_for_decl,
10077 darf2out_register_external_decl): Support incremental link.
10079 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10081 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
10083 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
10085 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
10086 it down to simple_object_copy_lto_debug_sections.
10087 (run_gcc): Determine incremental LTO link time and configure
10088 lto1 into non-wpa mode, disable renaming of debug sections.
10090 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
10092 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10093 descriptions of various incorrectly documented functions.
10095 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
10098 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10099 address check not strict.
10101 2018-05-30 Richard Biener <rguenther@suse.de>
10103 PR tree-optimization/85964
10104 * tracer.c (better_p): Drop initialized count check, we only
10105 call the function with initialized counts now.
10106 (find_best_successor): Do find a best edge if one
10107 has uninitialized count.
10108 (find_best_predecessor): Likewise. Do BB frequency check only
10109 if count is initialized.
10111 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
10113 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
10114 (aarch64_ldrstr_offset_compare): New.
10115 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
10116 load/store orderings.
10117 (aarch64_gen_adjusted_ldpstp): Likewise.
10119 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
10121 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10122 Check for subset of GENERAL_REGS and FP_REGS.
10123 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
10126 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10128 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
10129 and wi::to_poly_offset. Add the current offset and then check
10130 whether the sum fits, rather than using an unchecked addition of
10131 a checked term. Check for a shwi rather than a uhwi.
10132 * expr.c (get_bit_range): Use tree_to_poly_uint64.
10133 (store_constructor): Use poly_int_tree_p.
10134 (expand_expr_real_1): Likewise.
10135 * function.c (assign_temp): Likewise.
10136 * fold-const.c (const_binop): Use poly_int_tree_p and
10137 wi::to_poly_offset.
10138 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
10140 * ipa-icf-gimple.c (func_checker::compare_operand): Use
10141 to_poly_offset for MEM offsets.
10142 * ipa-icf.c (sem_variable::equals): Likewise.
10143 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
10144 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
10145 wi::to_poly_offset for BIT_FIELD_REF offsets.
10146 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
10147 wi::to_poly_offset.
10148 * var-tracking.c (emit_note_insn_var_location): Use
10149 tree_to_poly_uint64.
10151 2018-05-29 Jim Wilson <jimw@sifive.com>
10153 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
10155 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
10158 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
10159 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
10161 (sse4_1_round<mode>2): Use nonimmediate_operand
10162 for operand 1 predicate.
10164 2018-05-29 Martin Sebor <msebor@redhat.com>
10165 Richard Biener <rguenther@suse.de>
10168 * calls.c (get_size_range): Call determine_value_range instead
10169 of get_value_range..
10170 * tree-vrp.h (determine_value_range): Declared new function.
10171 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
10173 2018-05-29 Richard Biener <rguenther@suse.de>
10175 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
10176 sure to use non-pattern stmts for get_earlier_stmt arguments.
10177 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
10178 called on pattern stmts.
10179 (get_later_stmt): Likewise.
10181 2018-05-29 Martin Liska <mliska@suse.cz>
10183 PR gcov-profile/85759
10184 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
10187 2018-05-29 Jakub Jelinek <jakub@redhat.com>
10189 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
10191 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
10195 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10196 VEC_PACK_FLOAT_EXPR): New tree codes.
10197 * tree-pretty-print.c (op_code_prio): Handle
10198 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10199 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10200 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10201 * tree-inline.c (estimate_operator_cost): Likewise.
10202 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
10203 * fold-const.c (const_binop): Likewise.
10204 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
10205 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
10206 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10207 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
10208 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
10209 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
10210 * expr.c (expand_expr_real_2): Likewise.
10211 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
10212 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
10213 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
10215 * optabs.c (expand_widen_pattern_expr): For
10216 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
10217 sign from result type rather than operand's type.
10218 (expand_binop_directly): For vec_packu_float_optab and
10219 vec_packs_float_optab allow result type to be different from operand's
10221 * optabs-tree.c (optab_for_tree_code): Handle
10222 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10223 VEC_PACK_FLOAT_EXPR. Formatting fixes.
10224 * tree-vect-generic.c (expand_vector_operations_1): Handle
10225 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
10226 VEC_PACK_FLOAT_EXPR.
10227 * tree-vect-stmts.c (supportable_widening_operation): Handle
10229 (supportable_narrowing_operation): Handle FLOAT_EXPR.
10230 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
10231 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
10232 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
10233 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
10235 (vec_pack<floatprefix>_float_<mode>): New expander.
10236 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
10238 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
10239 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
10240 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
10241 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
10242 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
10244 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
10245 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
10246 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
10247 VEC_PACK_FLOAT_EXPR): Document.
10249 2018-05-29 Richard Biener <rguenther@suse.de>
10251 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
10253 (stmt_vec_info_vec): Make pointer.
10254 (init_stmt_vec_info_vec): Remove.
10255 (free_stmt_vec_info_vec): Likewise.
10256 (set_stmt_vec_info_vec): New function.
10257 (free_stmt_vec_infos): Likewise.
10258 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
10259 (set_vinfo_for_stmt): Likewise.
10260 (get_earlier_stmt): Likewise.
10261 (get_later_stmt): Likewise.
10262 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
10263 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
10264 (vec_info::~vec_info): Free stmt_vec_infos.
10265 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
10266 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
10267 (pass_slp_vectorize::execute): Likewise.
10268 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
10269 (free_stmt_vec_info_vec): Likewise.
10270 (set_stmt_vec_info_vec): New function.
10271 (free_stmt_vec_infos): Likewise.
10272 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
10273 the global stmt_vec_info_vec.
10274 * tree-parloops.c (gather_scalar_reductions): Use
10275 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
10278 2018-05-29 Richard Biener <rguenther@suse.de>
10280 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
10282 2018-05-29 Martin Liska <mliska@suse.cz>
10283 David Malcolm <dmalcolm@redhat.com>
10285 * vec.c (test_reverse): New.
10286 (vec_c_tests): Add new test.
10287 * vec.h (vl_ptr>::reverse): New function.
10289 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
10291 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
10293 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
10296 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
10298 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
10300 2018-05-28 Richard Biener <rguenther@suse.de>
10302 PR tree-optimization/85933
10303 * tree-vect-data-refs.c (vect_record_base_alignments): Only
10304 look at stmts marked as vectorizable.
10306 2018-05-28 Richard Biener <rguenther@suse.de>
10308 PR tree-optimization/85934
10309 * tree-vect-generic.c (expand_vector_operations_1): Hoist
10310 vector boolean check before scalar optimization.
10312 2018-05-28 Jakub Jelinek <jakub@redhat.com>
10314 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
10317 2018-05-28 Mark Wielaard <mark@klomp.org>
10319 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
10320 if it is an expression containing a minus sign.
10322 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
10324 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
10326 2018-05-27 Paul Koning <ni1d@arrl.net>
10328 * config/pdp11/pdp11.md (truncsihi2): Remove.
10330 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
10331 Chung-Ju Wu <jasonwucj@gmail.com>
10333 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
10335 (unaligned_store_dw): Ditto.
10336 * config/nds32/nds32-memory-manipulation.c
10337 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
10338 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
10339 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
10340 (emit_setmem_word_loop): Rename to ...
10341 (emit_setmem_doubleword_loop): ... this.
10342 (nds32_gen_dup_4_byte_to_word_value): New function.
10343 (nds32_gen_dup_8_byte_to_double_word_value): New function.
10344 (nds32_expand_setmem_loop): Refine implementation.
10345 (nds32_expand_setmem_loop_v3m): Ditto.
10346 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
10349 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10351 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
10353 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
10355 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
10356 (nds32_init_machine_status): Initialize machine->attr_naked_p and
10357 machine->attr_no_prologue_p.
10358 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
10359 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
10360 (nds32_expand_epilogue): Consider attr_naked_p.
10361 (nds32_expand_epilogue_v3pop): Likewise.
10362 (nds32_can_use_return_insn): Likewise.
10363 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
10364 attr_no_prologue_p fields.
10365 * config/nds32/nds32.opt (mret-in-naked-func): New option.
10367 2018-05-27 Jakub Jelinek <jakub@redhat.com>
10370 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
10372 * config/i386/sse.md
10373 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
10375 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
10377 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
10379 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
10381 (*<floatsuffix>floatv2div2sf2): Rename to ...
10382 (*float<floatunssuffix>v2div2sf2): ... this.
10383 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
10384 (float<floatunssuffix>v2div2sf2_mask): ... this.
10385 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
10386 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
10387 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
10389 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
10391 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10393 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
10395 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10397 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
10399 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
10400 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
10401 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
10402 gen_ufix_truncv8dfv8si2.
10403 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
10404 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
10405 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
10406 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
10407 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
10408 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
10409 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
10410 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
10412 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10416 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
10418 2018-05-25 Jim Wilson <jimw@sifive.com>
10420 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
10421 * config/riscv/riscv.c (struct machine_function): Add
10422 interrupt_handler_p and attribute_checked_p fields.
10423 (riscv_attribute_table): Add interrupt.
10424 (riscv_interrupt_type_p): New.
10425 (riscv_save_reg_p): Save extra regs for interrupt handler.
10426 (riscv_use_save_libcall): Return false for interrupt handler.
10427 (riscv_first_stack_step): Add forward declaration.
10428 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
10429 for interrupt handler with large frame. Use it for saved reg list.
10430 (riscv_expand_prologue): Move flag_stack_usage_info support to
10431 eliminate duplication.
10432 (riscv_expand_epilogue): Generate mret for interrupt handler.
10433 (riscv_epilogue_uses): New.
10434 (riscv_can_use_return_insn): Return false for interrupt handler.
10435 (riscv_function_ok_for_sibcall): Likewise.
10436 (riscv_set_current_function): Add interrupt handler support.
10437 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
10438 * config/riscv/riscv.md (UNSPECV_MRET): New.
10440 (riscv_frflags, riscv_fsflags): Use tab after opcode.
10442 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
10444 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
10446 PR tree-optimization/85712
10447 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
10448 this candidate has already been replaced in-situ by a copy.
10450 2018-05-25 Jason Merrill <jason@redhat.com>
10452 PR c++/80485 - inline function non-zero address.
10453 * symtab.c (nonzero_address): Check DECL_COMDAT.
10455 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
10458 * config/alpha/alpha.md (ashlsi3): New insn pattern.
10459 (*ashlsi_se): Rename from *ashldi_se. Define as sign
10460 extension of SImode operation. Use const123_operand predicate.
10461 (*saddsi_1): Remove.
10462 (*saddl_se_1): Ditto.
10463 (*ssubsi_1): Ditto.
10464 (*ssubl_se_1): Ditto.
10465 * config/alpha/predicates.md (const123_operand): New predicate.
10466 * config/alpha/constraints.md (P): Use IN_RANGE.
10468 2018-05-25 Richard Biener <rguenther@suse.de>
10470 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
10472 (ref_maybe_used_by_stmt_p): Likewise.
10473 (stmt_may_clobber_ref_p): Likewise.
10474 (stmt_may_clobber_ref_p_1): Likewise.
10475 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
10477 (ref_maybe_used_by_stmt_p): Likewise.
10478 (stmt_may_clobber_ref_p): Likewise.
10479 (stmt_may_clobber_ref_p_1): Likewise.
10480 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
10481 the alias oracle to disambiguate DRs with stmts DR analysis
10483 (vect_analyze_data_refs): Do not give up on not analyzable
10484 DRs for BB vectorization. Remove code truncating the dataref
10487 2018-05-25 Jakub Jelinek <jakub@redhat.com>
10490 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10491 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
10492 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
10494 2018-05-25 Richard Biener <rguenther@suse.de>
10496 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
10497 function, combining stmt data ref gathering and fatal analysis
10499 (vect_analyze_data_refs): Remove now redudnant code and simplify.
10500 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
10501 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
10502 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
10503 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
10505 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10507 PR tree-optimization/85720
10508 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
10509 SCC if all partitions are builtins.
10510 (version_loop_by_alias_check): New parameter. Generate cancelable
10511 runtime alias check if all partitions are builtins.
10512 (distribute_loop): Update call to above function.
10514 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10516 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
10517 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
10518 (parm_default_def_partition_arg): Ditto.
10519 (set_parm_default_def_partition): Ditto.
10520 (get_parm_default_def_partitions): Ditto and make it static.
10521 (get_undefined_value_partitions): Ditto and make it static.
10522 (remove_ssa_form): Refactor call to init_var_map here.
10523 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
10524 computation for loop region.
10525 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
10526 (register_default_def): Delete.
10527 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
10528 (parm_default_def_partition_arg): Ditto.
10529 (set_parm_default_def_partition): Ditto.
10530 (get_parm_default_def_partitions): Ditto and make it static.
10531 (get_undefined_value_partitions): Ditto and make it static.
10532 (coalesce_with_default, coalesce_with_default): Update comment.
10533 (create_coalesce_list_for_region): New func factored out from
10534 create_outofssa_var_map.
10535 (populate_coalesce_list_for_outofssa): New func factored out from
10536 create_outofssa_var_map and coalesce_ssa_name.
10537 (create_outofssa_var_map): Delete.
10538 (coalesce_ssa_name): Refactor to support live range computation.
10539 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
10540 (get_parm_default_def_partitions): Delete.
10541 (get_undefined_value_partitions): Ditto.
10542 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
10543 computation for loop region.
10544 (new_tree_live_info, loe_visit_block): Ditto.
10545 (live_worklist, set_var_live_on_entry): Ditto.
10546 (calculate_live_on_exit, verify_live_on_entry): Ditto.
10547 * tree-ssa-live.h (struct _var_map): New fields.
10548 (init_var_map): Change decl.
10549 (region_contains_p): New.
10551 2018-05-25 Bin Cheng <bin.cheng@arm.com>
10553 * tree-ssa-live.h (live_merge_and_clear): Delete.
10555 2018-05-25 Richard Biener <rguenther@suse.de>
10558 * tree-dump.c (dequeue_and_dump): Remove access to removed
10559 operand 2 of a SWITCH_EXPR.
10561 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10563 * doc/sourcebuild.texi (vect_double_cond_arith): Include
10564 multiplication and division.
10565 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
10566 (cond_udiv@var{m}, cond_umod@var{m}): Document.
10567 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
10568 (cond_udiv_optab, cond_umod_optab): New optabs.
10569 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
10570 (IFN_COND_RDIV): New internal functions.
10571 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
10572 TRUNC_MOD_EXPR and RDIV_EXPR.
10573 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
10574 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
10576 (SVE_INT_BINARY): Include mult.
10577 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
10578 (optab, sve_int_op): Handle mult.
10579 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
10581 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
10582 for SVE_INT_BINARY_SD.
10584 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10586 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
10587 (optab, sve_int_op): Handle div and udiv.
10588 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
10589 for SVE_INT_BINARY_SD.
10590 (*<optab><mode>3): New insn for the same.
10592 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10594 * tree-vect-patterns.c: Include predict.h.
10595 (vect_recog_divmod_pattern): Restrict check for division support
10596 to when optimizing for size.
10598 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10600 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
10601 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
10602 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
10603 (gimple_match_op::set_op): Likewise.
10604 (gimple_resimplify4): Declare.
10605 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
10606 (expr::gen_transform): Likewise.
10607 (decision_tree::gen): Generate a simplification routine for 4 operands.
10608 * gimple-match-head.c (gimple_simplify): Add an overload for
10609 4 operands. In the top-level function, handle up to 4 call
10610 arguments and call gimple_resimplify4.
10611 (gimple_resimplify4): New function.
10612 (build_call_internal): Pass a fourth operand.
10613 (maybe_push_to_seq): Likewise.
10614 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
10615 Fold VEC_COND_EXPRs of an operation and a default value into
10616 an IFN_COND_* function if possible.
10617 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
10619 (SVE_COND_FP_BINARY): Include them.
10620 (optab, sve_fp_op): Handle them.
10621 (SVE_INT_BINARY_REV): New code iterator.
10622 (SVE_COND_FP_BINARY_REV): New int iterator.
10623 (commutative): New int attribute.
10624 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
10626 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
10628 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
10629 (*cond_<optab><mode>): New patterns for reversed operands.
10631 2018-05-25 Richard Biener <rguenther@suse.de>
10633 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
10634 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
10635 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
10636 (STMT_VINFO_GROUPED_ACCESS): Adjust.
10637 * tree-vect-data-refs.c (everywhere): Adjust users.
10638 * tree-vect-loop.c (everywhere): Likewise.
10639 * tree-vect-slp.c (everywhere): Likewise.
10640 * tree-vect-stmts.c (everywhere): Likewise.
10641 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10643 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10645 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
10647 (gcc_cv_as_section_exclude): ... this.
10648 Try Solaris as #exclude syntax.
10649 * configure: Regenerate.
10650 * config.in: Regenerate.
10651 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
10653 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
10654 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
10656 * varasm.c (default_elf_asm_named_section): Don't check if
10657 HAVE_GAS_SECTION_EXCLUDE is defined.
10659 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10661 * doc/md.texi: Update the documentation of the cond_* optabs
10662 to mention the new final operand. Fix GET_MODE_NUNITS call.
10663 Describe the scalar case too.
10664 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
10665 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
10667 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
10668 (get_conditional_internal_fn): Update comment.
10669 * tree-vect-loop.c (vectorizable_reduction): Pass the original
10670 accumulator value as a final argument to conditional functions.
10671 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
10672 a define_expand and add an "else" operand. Assert for now that
10673 the else operand is equal to operand 2. Use SVE_INT_BINARY and
10674 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
10675 (*cond_<optab><mode>): New patterns.
10676 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
10677 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10678 (UNSPEC_COND_EOR): Delete.
10679 (optab): Remove associated mappings.
10680 (SVE_INT_BINARY): New code iterator.
10681 (sve_int_op): Remove int attribute and add "minus" to the code
10683 (SVE_COND_INT_OP): Delete.
10684 (SVE_COND_FP_OP): Rename to...
10685 (SVE_COND_FP_BINARY): ...this.
10687 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
10689 * optabs.c (can_reuse_operands_p): New function.
10690 (maybe_legitimize_operands): Try to reuse the results for
10693 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10695 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
10696 Add {q} suffix to insn mnemonic.
10698 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10700 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
10701 (msp430_warn_func_return): New.
10703 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
10705 * fold-const.c (tree_nonzero_bits): New function.
10706 * fold-const.h (tree_nonzero_bits): Likewise.
10707 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
10708 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
10710 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
10714 * varasm.c (assemble_alias): Check ifunc_resolver only on
10717 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
10720 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
10721 when memory input operand is handled.
10723 2018-05-24 Luis Machado <luis.machado@linaro.org>
10725 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
10727 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
10729 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10731 * match.pd: Delay FMA folds until after vectorization.
10733 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
10736 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
10737 address check not strict.
10739 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
10741 * gimple-match.h (gimple_match_op): New class.
10742 (mprts_hook): Replace parameters with a gimple_match_op *.
10743 (maybe_build_generic_op): Likewise.
10744 (gimple_simplified_result_is_gimple_val): Replace parameters with
10745 a const gimple_match_op *.
10746 (gimple_simplify): Replace code_helper * and tree * parameters with
10747 a gimple_match_op * parameter.
10748 (gimple_resimplify1): Replace code_helper *, tree and tree *
10749 parameters with a gimple_match_op * parameter.
10750 (gimple_resimplify2): Likewise.
10751 (gimple_resimplify3): Likewise.
10752 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
10753 parameters with a gimple_match_op * parameter.
10754 * gimple-match-head.c (gimple_simplify): Change prototypes of
10755 auto-generated functions to take a gimple_match_op * instead of
10756 separate code_helper * and tree * parameters. Make the same
10757 change in the top-level overload and update calls to the
10758 gimple_resimplify routines. Update calls to the auto-generated
10759 functions and to maybe_push_res_to_seq in the publicly-facing
10760 operation-specific gimple_simplify overloads.
10761 (gimple_match_op::MAX_NUM_OPS): Define.
10762 (gimple_resimplify1): Replace rcode and ops with a single res_op
10763 parameter. Update call to gimple_simplify.
10764 (gimple_resimplify2): Likewise.
10765 (gimple_resimplify3): Likewise.
10766 (mprts_hook): Replace parameters with a gimple_match_op *.
10767 (maybe_build_generic_op): Likewise.
10768 (build_call_internal): Replace type, nargs and ops with
10769 a gimple_match_op *.
10770 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
10771 with a single gimple_match_op *. Update calls to mprts_hook,
10772 build_call_internal and gimple_simplified_result_is_gimple_val.
10773 Factor out code that is common to the tree_code and combined_fn cases.
10774 * genmatch.c (expr::gen_transform): Replace tem_code and
10775 tem_ops with a gimple_match_op called tem_op. Update calls
10776 to the gimple_resimplify functions and maybe_push_res_to_seq.
10777 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
10778 res_ops. Update call to the gimple_resimplify functions.
10779 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
10780 (decision_tree::gen): Make the functions take a gimple_match_op *
10781 called res_op instead of separate res_code and res_ops parameters.
10782 Update call accordingly.
10783 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
10784 and ops with a single res_op parameter. Update calls to
10785 maybe_build_generic_op and maybe_push_res_to_seq.
10786 (fold_stmt_1): Update calls to gimple_simplify and
10787 replace_stmt_with_simplification.
10788 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
10789 and gimple_simplified_result_is_gimple_val.
10790 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
10792 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
10793 with a gimple_match_op *.
10794 (vn_nary_build_or_lookup): Likewise. Update call to
10795 vn_nary_build_or_lookup_1.
10796 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
10797 gimple_match_op *. Update calls to the gimple_resimplify routines
10798 and to gimple_simplified_result_is_gimple_val.
10799 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
10800 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
10801 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
10802 (visit_nary_op): Likewise.
10803 (visit_reference_op_load): Likewise.
10805 2018-05-23 Luis Machado <luis.machado@linaro.org>
10807 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
10808 modifier for printing the step amount.
10810 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
10813 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
10816 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
10818 * doc/sourcebuild.texi (Endianness): New subsubsection.
10820 2018-05-23 Luis Machado <luis.machado@linaro.org>
10822 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10823 <prefetch_dynamic_strides>: New const bool field.
10824 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10825 prefetch_dynamic_strides.
10826 (exynosm1_prefetch_tune): Likewise.
10827 (thunderxt88_prefetch_tune): Likewise.
10828 (thunderx_prefetch_tune): Likewise.
10829 (thunderx2t99_prefetch_tune): Likewise.
10830 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
10832 (aarch64_override_options_internal): Update to set
10833 PARAM_PREFETCH_DYNAMIC_STRIDES.
10834 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
10835 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
10836 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
10837 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
10838 prefetch-dynamic-strides setting.
10840 2018-05-23 Luis Machado <luis.machado@linaro.org>
10842 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
10843 <minimum_stride>: New const int field.
10844 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
10845 minimum_stride field defaulting to -1.
10846 (exynosm1_prefetch_tune): Likewise.
10847 (thunderxt88_prefetch_tune): Likewise.
10848 (thunderx_prefetch_tune): Likewise.
10849 (thunderx2t99_prefetch_tune): Likewise.
10850 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
10851 <default_opt_level>: Set to 3.
10852 (aarch64_override_options_internal): Update to set
10853 PARAM_PREFETCH_MINIMUM_STRIDE.
10854 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
10855 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
10856 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
10857 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
10858 stride is constant and is below the minimum stride threshold.
10860 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10862 * config/arm/arm-cpus.in (mode26): Delete.
10863 (armv4): Delete mode26 reference.
10864 * config/arm/arm.c (arm_configure_build_target): Delete use of
10867 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10869 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
10871 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
10872 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
10874 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
10875 (floatunsdidf2): Ditto.
10877 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
10879 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
10880 (fixuns_trunc<mode>si2_avx512f): Ditto.
10881 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
10882 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
10883 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
10885 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
10887 PR rtl-optimization/79985
10888 * df-scan.c (df_insn_refs_collect): Remove special case for
10889 global registers and asm statements.
10891 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
10893 * extend.texi (Global Register Variables): Rewrite the bullet list.
10894 Note that the register is available for allocation. Note that access
10895 via inline asm must use constraints. Add note about async-signal
10896 handlers. Remove paragraph about automagic register selection.
10898 2018-05-23 Richard Biener <rguenther@suse.de>
10900 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
10901 of fixed offset from memset VN.
10903 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
10905 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
10906 first_interp field.
10907 (alloc_cand_and_find_basis): Initialize first_interp field.
10908 (slsr_process_mul): Modify first_interp field.
10909 (slsr_process_add): Likewise.
10910 (slsr_process_cast): Modify first_interp field for each new
10912 (slsr_process_copy): Likewise.
10913 (dump_candidate): Dump first_interp field.
10914 (replace_mult_candidate): Process all interpretations, not just
10916 (replace_rhs_if_not_dup): Likewise.
10917 (replace_one_candidate): Likewise.
10919 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
10921 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
10923 (aarch64_needs_frame_chain): New function.
10924 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
10926 2018-05-23 Sudakshina Das <sudi.das@arm.com>
10929 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
10930 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
10931 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
10932 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
10933 as true for strict-align.
10934 (aarch64_can_inline_p): Perform checks even when callee has no
10935 attributes to check for strict alignment.
10936 * doc/extend.texi (AArch64 Function Attributes): Document
10938 * doc/invoke.texi: (AArch64 Options): Likewise.
10940 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
10942 PR tree-optimization/85853
10943 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
10944 the handling of the root of the node to...
10945 (vect_slp_analyze_node_operations_1): ...this new function,
10946 and run the whole thing with the child nodes' def types
10947 set according to their SLP node's def type.
10949 2018-05-23 Richard Biener <rguenther@suse.de>
10951 PR middle-end/85874
10952 * tree-data-ref.c (create_runtime_alias_checks): Defer
10953 and ignore overflow warnings.
10955 2018-05-23 Yury Gribov <tetra2005@gmail.com>
10957 PR tree-optimization/85822
10958 * tree-vrp.c (is_masked_range_test): Fix handling of negative
10961 2018-05-23 Richard Biener <rguenther@suse.de>
10963 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10964 memset constants via native_interpret_expr.
10966 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
10969 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
10971 (cgraph_node::create_alias): Likewise.
10972 (cgraph_node::get_availability): Check ifunc_resolver instead
10973 of looking up ifunc attribute.
10974 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
10975 * varasm.c (do_assemble_alias): Likewise.
10976 (assemble_alias): Likewise.
10977 (default_binds_local_p_3): Likewise.
10978 * cgraph.h (cgraph_node): Add ifunc_resolver.
10979 (cgraph_node::only_called_directly_or_aliased_p): Return false
10980 for IFUNC resolver.
10981 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
10983 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
10984 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
10985 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
10986 instead of looking up ifunc attribute.
10988 2018-05-22 Luis Machado <luis.machado@linaro.org>
10990 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
10992 2018-05-22 Martin Sebor <msebor@redhat.com>
10994 PR middle-end/85359
10995 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
10996 only when expasion succeeds.
10997 (expand_builtin_strcmp): Same.
10998 (expand_builtin_strncmp): Same.
11000 2018-05-22 Martin Sebor <msebor@redhat.com>
11002 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
11004 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11005 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11007 * config/aarch64/aarch64-ldpstp.md: Replace uses of
11008 aarch64_mem_pair_operand with memory_operand and delete operand swapping
11010 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11011 Add check for legitimate_address.
11012 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
11013 (aarch64_swap_ldrstr_operands): New.
11014 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
11017 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
11018 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11020 * config/aarch64/aarch64.md: New patterns to generate stp
11022 (store_pair_sw, store_pair_dw): New patterns to generate stp for
11023 single words and double words.
11024 (load_pair_sw, load_pair_dw): Likewise.
11025 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
11027 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
11029 * config/aarch64/aarch64-ldpstp.md: Modify peephole
11030 for different mode ldpstp and add peephole for merged zero stores.
11031 Likewise for loads.
11032 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
11034 (aarch64_gen_store_pair): Rename calls to match new patterns.
11035 (aarch64_gen_load_pair): Rename calls to match new patterns.
11036 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
11037 (load_pair<DREG:mode><DREG2:mode>): ... This.
11038 (store_pair<mode>): Rename to...
11039 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
11040 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
11041 New mode iterators.
11042 (V_INT_EQUIV): Handle SImode.
11043 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
11046 2018-05-22 Martin Sebor <msebor@redhat.com>
11049 * calls.c (maybe_warn_nonstring_arg): Use string length to set
11050 or ajust the presumed bound on an operation to avoid unnecessary
11053 2018-05-22 Martin Sebor <msebor@redhat.com>
11055 PR tree-optimization/85826
11056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
11057 assuming that a DECL necesarily has a constant size.
11059 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
11061 PR middle-end/85862
11062 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
11064 2018-05-22 Richard Biener <rguenther@suse.de>
11066 PR tree-optimization/85834
11067 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
11068 non-constant and non-zero memset arguments.
11070 2018-05-22 Martin Liska <mliska@suse.cz>
11073 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
11075 2018-05-22 Richard Biener <rguenther@suse.de>
11077 PR tree-optimization/85863
11078 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
11079 comparisons when vectype is specified.
11080 (vectorizable_condition): Do not specify vectype for
11081 vect_is_simple_cond when SLP vectorizing.
11083 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
11086 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
11087 define __ibm128 as long double.
11088 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
11089 as a distinct type when IEEE 128-bit support is enabled.
11090 (init_float128_ieee): Fix up conversions between IFmode and IEEE
11091 128-bit types to use the correct functions.
11092 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
11093 convert between 128-bit floating point types that have different
11094 modes but the same representation, instead of using gen_lowpart to
11096 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
11098 (IFKF_reg): New attributes to give the register constraints for
11100 (extend<mode>tf2_internal): New insns to mark an explicit
11101 conversion between 128-bit floating point types that have a
11102 different mode but share the same representation.
11104 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
11106 PR tree-optimization/85814
11107 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
11108 a null return from get_strinfo when unsharing the next
11109 strinfo in the chain.
11111 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11114 * varasm.c (weak_finish): Clean up weak_decls.
11116 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11118 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
11119 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
11120 UNSPEC_UADALP values.
11121 * config/aarch64/iterators.md (ABAL): New int iterator.
11124 (sur): Add mappings for the above.
11125 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
11127 (aarch64_<sur>abal<mode>_4): Likewise.
11128 (aarch64_<sur>adalp<mode>_3): Likewise.
11129 (<sur>sadv16qi): New define_expand.
11131 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
11133 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
11134 (*movdf_internal): Ditto.
11135 (*rcpsf2_sse): Ditto.
11136 (*rsqrtsf2_sse): Ditto.
11137 (*sqrt<mode>2_sse): Ditto.
11139 2018-05-21 Tamar Christina <tamar.christina@arm.com>
11141 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
11143 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
11144 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
11145 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
11146 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
11148 * config/aarch64/arm_neon.h: Likewise.
11149 * config/aarch64/iterators.md (VQ_I): New.
11151 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
11153 * config.gcc: Add arc/t-multilib-linux to tmake_file for
11155 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
11158 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11160 * config/nds32/constraints.md (S): New constraint.
11161 * config/nds32/nds32.md (call_internal): Use constraint S.
11162 (call_value_internal): Likewise.
11163 (sibcall_internal): Likewise.
11164 (sibcall_value_internal): Likewise.
11166 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11167 Chung-Ju Wu <jasonwucj@gmail.com>
11169 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
11170 into consideration.
11172 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
11173 Chung-Ju Wu <jasonwucj@gmail.com>
11175 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
11176 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
11177 (nds32_rtx_costs_impl): Simplify.
11178 (nds32_address_cost_impl): Simplify.
11179 (nds32_init_rtx_costs): New function.
11180 (nds32_rtx_costs_speed_prefer): Likewise.
11181 (nds32_rtx_costs_size_prefer): Likewise.
11182 (nds32_address_cost_speed_prefer): Likewise.
11183 (nds32_address_cost_speed_fwprop): Likewise.
11184 (nds32_address_cost_size_prefer): Likewise.
11185 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
11186 * config/nds32/nds32.c (nds32_option_override): Use
11187 nds32_init_rtx_costs function.
11189 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
11191 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
11192 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
11193 (TARGET_PIPELINE_N8): Likewise.
11194 (TARGET_PIPELINE_N10): Likewise.
11195 (TARGET_PIPELINE_N13): Likewise.
11196 (TARGET_PIPELINE_GRAYWOLF): Likewise.
11198 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
11200 * config/nds32/nds32-fpu.md: Update copyright year.
11202 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11204 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
11206 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11208 * config/nds32/nds32.c
11209 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
11210 * config/nds32/nds32.opt (minline-asm-r15): New option.
11212 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11214 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
11216 * config/nds32/nds32.md (abssi2): New pattern.
11218 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
11220 * config/i386/i386.md (rex64namesuffix): New mode attribute.
11221 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
11222 Merge insn pattern from sse_cvtsi2ss<round_name> and
11223 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
11224 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
11225 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
11226 using SWI48 mode iterator.
11227 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
11228 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
11229 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
11230 pattern from sse_cvttss2si<round_saeonly_name>
11231 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
11232 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
11233 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
11234 using SWI48 mode iterator.
11235 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
11236 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
11237 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
11238 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
11239 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
11240 using SWI48 mode iterator.
11241 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
11242 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
11243 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
11244 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
11245 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
11246 SWI48 mode iterator.
11247 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
11248 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
11249 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
11250 pattern from sse_cvttsd2si<round_saeonly_name>
11251 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
11253 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11255 * config/nds32/nds32-md-auxiliary.c
11256 (nds32_valid_smw_lwm_base_p): Refine.
11257 (nds32_output_smw_single_word): Refine.
11258 (nds32_output_smw_double_word): New.
11259 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
11261 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11263 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
11264 (nds32_output_stack_pop): Refine.
11265 (nds32_expand_unaligned_load): Refine.
11266 (nds32_expand_unaligned_store): Refine.
11268 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11269 Chung-Ju Wu <jasonwucj@gmail.com>
11271 * config/nds32/constants.md: Add TP_REGNUM constant.
11272 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
11273 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
11275 * config/nds32/nds32-doubleword.md: Consider flag_pic.
11276 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
11277 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
11278 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
11279 and PIC code generation.
11280 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
11282 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
11284 * config/nds32/nds32.md: Support TLS and PIC.
11285 * config/nds32/nds32.c: Support TLS and PIC.
11286 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
11287 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
11290 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
11292 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
11293 mode with E_ prefix.
11295 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11296 Chung-Ju Wu <jasonwucj@gmail.com>
11298 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
11299 * config/nds32/nds32-md-auxiliary.c
11300 (symbolic_reference_mentioned_p): New.
11301 (nds32_legitimize_ict_address): New.
11302 (nds32_expand_ict_move): New.
11303 (nds32_indirect_call_referenced_p): New.
11304 (nds32_symbol_binds_local_p): Delete.
11305 (nds32_long_call_p): Modify.
11306 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
11307 * config/nds32/nds32-protos.h
11308 (symbolic_reference_mentioned_p): Declare.
11309 (nds32_legitimize_ict_address): Declare.
11310 (nds32_expand_ict_move): Declare.
11311 (nds32_indirect_call_referenced_p): Declare.
11312 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
11313 (nds32_relax_group): Use nds32_ict_const_p as condition.
11314 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
11315 (nds32_asm_file_start): Output ict_model directive in asm code.
11316 (nds32_legitimate_address_p): Consider indirect call.
11317 (nds32_print_operand): Consider indirect call.
11318 (nds32_print_operand_address): Consider indirect call.
11319 (nds32_insert_attributes): Handle "indirect_call" attribute.
11320 (TARGET_LEGITIMATE_ADDRESS_P): Define.
11321 (TARGET_LEGITIMATE_CONSTANT_P): Define.
11322 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
11323 (TARGET_DELEGITIMIZE_ADDRESS): Define.
11324 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
11325 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
11326 (TARGET_ICT_MODEL_SMALL): Define.
11327 (TARGET_ICT_MODEL_LARGE): Define.
11328 * config/nds32/nds32.md (movsi): Consider ict model.
11329 (call, call_value): Consider ict model.
11330 (sibcall, sibcall_value): Consider ict model.
11331 * config/nds32/nds32.opt (mict-model): New option.
11332 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
11335 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
11336 Monk Chiang <sh.chiang04@gmail.com>
11337 Jim Wilson <jimw@sifive.com>
11339 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
11340 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
11341 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
11342 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
11343 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
11344 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
11345 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
11346 compute save_libcall_adjustment properly.
11347 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
11348 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
11349 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
11350 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
11351 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
11352 (ABI_SPEC): Handle mabi=ilp32e.
11353 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
11354 (RVE): Add RVE mask.
11355 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
11356 <-march>: Add rv32e as an example.
11358 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
11361 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
11362 (intra_create_variable_infos): Handle C++ constructors.
11364 2018-05-18 Martin Liska <mliska@suse.cz>
11366 * passes.def: Remove a redundant pass.
11368 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
11371 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
11373 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11375 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
11377 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
11379 (armv2, armv2a, armv3, armv3m): Delete architectures.
11380 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
11381 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
11382 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
11384 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
11385 (*mulsidi3adddi): Likewise.
11386 (mulsidi3): Likewise.
11387 (*mulsidi3_nov6): Likewise.
11388 (umulsidi3): Likewise.
11389 (umulsidi3_nov6): Likewise.
11390 (umaddsidi4): Likewise.
11391 (*umulsidi3adddi): Likewise.
11392 (smulsi3_highpart): Likewise.
11393 (*smulsi3_highpart_nov6): Likewise.
11394 (umulsi3_highpart): Likewise.
11395 (*umulsi3_highpart_nov6): Likewise.
11396 * config/arm/arm.h (arm_arch3m): Delete.
11397 * config/arm/arm.c (arm_arch3m): Delete.
11398 (arm_option_override_internal): Update armv3-related comment.
11399 (arm_configure_build_target): Delete use of isa_bit_mode32.
11400 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
11401 (arm_rtx_costs_internal): Delete check of arm_arch3m.
11402 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
11403 (mulsa3): Likewise.
11404 (mulusa3): Likewise.
11405 * config/arm/arm-protos.h (arm_arch3m): Delete.
11406 * config/arm/arm-tables.opt: Regenerate.
11407 * config/arm/arm-tune.md: Likewise.
11408 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
11409 deleted architectures.
11411 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11413 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
11414 (armv5t, armv5te): New features.
11415 (ARMv5, ARMv5e): Delete fgroups.
11416 (ARMv5t, ARMv5te): Adjust for above changes.
11417 (ARMv6m): Likewise.
11418 (armv5, armv5e): Delete arches.
11419 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
11421 (*call_reg_arm): Likewise.
11422 (*call_value_reg_armv5): Likewise.
11423 (*call_value_reg_arm): Likewise.
11424 (*call_symbol): Likewise.
11425 (*call_value_symbol): Likewise.
11426 (*sibcall_insn): Likewise.
11427 (*sibcall_value_insn): Likewise.
11428 (clzsi2): Likewise.
11429 (prefetch): Likewise.
11430 (define_split and define_peephole2 dependent on arm_arch5):
11432 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
11434 (TARGET_ARM_QBIT): Likewise.
11435 (TARGET_DSP_MULTIPLY): Likewise.
11436 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
11437 (arm_arch5, arm_arch5e): Delete.
11438 (arm_arch5t, arm_arch5te): Declare.
11439 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
11440 (arm_arch5t): Declare.
11441 (arm_option_reconfigure_globals): Update for the above.
11442 (arm_options_perform_arch_sanity_checks): Update comment, replace
11443 use of arm_arch5 with arm_arch5t.
11444 (use_return_insn): Likewise.
11445 (arm_emit_call_insn): Likewise.
11446 (output_return_instruction): Likewise.
11447 (arm_final_prescan_insn): Likewise.
11448 (arm_coproc_builtin_available): Likewise.
11449 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
11450 arm_arch5e with arm_arch5t and arm_arch5te.
11451 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
11452 (arm_arch5t, arm_arch5te): Declare.
11453 * config/arm/arm-tables.opt: Regenerate.
11454 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
11455 * config/arm/t-multilib: Likewise.
11456 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
11457 instead of arm_arch5.
11458 (*call_reg_thumb1): Likewise.
11459 (*call_value_reg_thumb1_v5): Likewise.
11460 (*call_value_reg_thumb1): Likewise.
11461 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
11463 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
11465 2018-05-18 Martin Liska <mliska@suse.cz>
11467 PR gcov-profile/84846
11468 * doc/gcov.texi: Document -t option of gcov tool.
11470 2018-05-18 Martin Liska <mliska@suse.cz>
11472 PR gcov-profile/84846
11473 * gcov.c (print_usage): Add new -t option.
11474 (process_args): Handle the option.
11475 (generate_results): Use stdout as output when requested by
11478 2018-05-18 Martin Liska <mliska@suse.cz>
11480 PR gcov-profile/84846
11481 * coverage.c (coverage_init): Write PWD to .gcno file.
11482 * doc/gcov.texi: Document how working directory is printed.
11483 * gcov-dump.c (dump_gcov_file): Print PWD.
11484 * gcov.c (output_intermediate_file): Likewise.
11485 (read_graph_file): Read PWD string.
11486 (output_lines): Print PWD.
11488 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11490 PR middle-end/85817
11491 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
11492 for retval and return false if all args to phi are zero.
11494 2018-05-18 Richard Biener <rguenther@suse.de>
11496 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
11498 (evrp_dom_walker::before_dom_children): Call it.
11500 2018-05-18 Richard Biener <rguenther@suse.de>
11502 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
11503 results when processing array refs with variable index.
11505 2018-05-18 Toon Moene <toon@moene.org>
11507 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
11508 directly after that of -floop-interchange. Indicate that both
11509 options are enabled by default when specifying -O3.
11511 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11513 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
11514 iterator. Delete separate integer-mode vec_set<mode> expander.
11515 (aarch64_simd_vec_setv2di): Delete.
11516 (vec_setv2di): Delete.
11517 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
11518 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
11519 the "w, r" alternative.
11521 2018-05-18 Martin Liska <mliska@suse.cz>
11523 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
11524 * tree-pass.h (make_pass_lower_switch_O0): New function.
11525 * tree-switch-conversion.c (node_has_low_bound): Remove.
11526 (node_has_high_bound): Likewise.
11527 (node_is_bounded): Likewise.
11528 (class pass_lower_switch): Make it a template type and create
11530 (pass_lower_switch::execute): Add template argument.
11531 (make_pass_lower_switch): New function.
11532 (make_pass_lower_switch_O0): New function.
11533 (do_jump_if_equal): Remove.
11534 (emit_case_nodes): Simplify to just handle all 3 cases and leave
11535 all the hard work to tree optimization passes.
11537 2018-05-18 Martin Liska <mliska@suse.cz>
11539 * dbgcnt.c (limit_low): Renamed from limit.
11540 (limit_high): New variable.
11541 (dbg_cnt_is_enabled): Check for upper limit.
11542 (dbg_cnt): Adjust dumping.
11543 (dbg_cnt_set_limit_by_index): Add new argument for high
11545 (dbg_cnt_set_limit_by_name): Likewise.
11546 (dbg_cnt_process_single_pair): Parse new format.
11547 (dbg_cnt_process_opt): Use strtok.
11548 (dbg_cnt_list_all_counters): Remove 'value' and add
11550 * doc/invoke.texi: Document changes.
11552 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
11554 * doc/sourcebuild.texi (scalar_all_fma): Document.
11555 * tree.def (FMA_EXPR): Delete.
11556 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
11557 * internal-fn.c (ternary_direct): New macro.
11558 (expand_ternary_optab_fn): Likewise.
11559 (direct_ternary_optab_supported_p): Likewise.
11560 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
11561 * builtins.c (fold_builtin_fma): Delete.
11562 (fold_builtin_3): Don't call it.
11563 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
11564 * expr.c (expand_expr_real_2): Likewise.
11565 * fold-const.c (operand_equal_p): Likewise.
11566 (fold_ternary_loc): Likewise.
11567 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
11568 * gimple.c (DEFTREECODE): Likewise.
11569 * gimplify.c (gimplify_expr): Likewise.
11570 * optabs-tree.c (optab_for_tree_code): Likewise.
11571 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
11572 * tree-eh.c (operation_could_trap_p): Likewise.
11573 (stmt_could_throw_1_p): Likewise.
11574 * tree-inline.c (estimate_operator_cost): Likewise.
11575 * tree-pretty-print.c (dump_generic_node): Likewise.
11576 (op_code_prio): Likewise.
11577 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11578 * tree-ssa-operands.c (get_expr_operands): Likewise.
11579 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
11580 * fold-const-call.h (fold_fma): Delete.
11581 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
11582 CFN_FNMA and CFN_FNMS.
11583 (fold_fma): Delete.
11584 * genmatch.c (combined_fn): New enum.
11585 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
11586 (commutative_op): New function.
11587 (commutate): Use it. Handle more than 2 operands.
11588 (dt_operand::gen_gimple_expr): Use commutative_op.
11589 (parser::parse_expr): Allow :c to be used with non-binary
11590 operators if the commutative operand is known.
11591 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
11592 CFN_FMS, CFN_FNMA and CFN_FNMS.
11593 (backprop::process_assign_use): Remove FMA_EXPR handling.
11594 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
11595 (gen_hsa_fma): New function.
11596 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
11597 IFN_FNMA and IFN_FNMS.
11598 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
11599 * gimple-fold.h (follow_all_ssa_edges): Declare.
11600 * gimple-fold.c (follow_all_ssa_edges): New function.
11601 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
11602 gimple_build interface and use follow_all_ssa_edges to fold the result.
11603 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
11604 instead of checking for optabs directly.
11605 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
11606 rather than FMA_EXPRs.
11607 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
11608 call to IFN_FMA instead of an FMA_EXPR.
11610 2018-05-17 Jim Wilson <jimw@sifive.com>
11612 * expr.c (do_tablejump): When converting index to Pmode, if we have a
11613 sign extended promoted subreg, and the range does not have the sign bit
11614 set, then do a sign extend.
11616 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
11617 test, check for sign extended subreg and/or constant operands, and
11618 do a sign extend in that case.
11620 2018-05-17 Steve Ellcey <sellcey@cavium.com>
11622 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
11623 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
11625 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
11626 Change logics_shift_reg to logics_shift_imm.
11627 (thunderx2t99_fp_loadpair_basic): Delete.
11628 (thunderx2t99_fp_storepair_basic): Delete.
11629 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
11630 (thunderx2t99_asimd_polynomial): Delete.
11631 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
11632 and neon_fp_mul_d_scalar_q.
11633 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
11634 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
11635 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
11636 (thunderx2t99_asimd_lut): Add missing tbl types.
11637 (thunderx2t99_asimd_ext): Delete.
11638 (thunderx2t99_asimd_load1_1_mult): Delete.
11639 (thunderx2t99_asimd_load1_2_mult): Delete.
11640 (thunderx2t99_asimd_load1_ldp): New.
11641 (thunderx2t99_asimd_load1): New.
11642 (thunderx2t99_asimd_load2): Add missing *load2* types.
11643 (thunderx2t99_asimd_load3): New.
11644 (thunderx2t99_asimd_load4): New.
11645 (thunderx2t99_asimd_store1_1_mult): Delete.
11646 (thunderx2t99_asimd_store1_2_mult): Delete.
11647 (thunderx2t99_asimd_store2_mult): Delete.
11648 (thunderx2t99_asimd_store2_onelane): Delete.
11649 (thunderx2t99_asimd_store_stp): New.
11650 (thunderx2t99_asimd_store1): New.
11651 (thunderx2t99_asimd_store2): New.
11652 (thunderx2t99_asimd_store3): New.
11653 (thunderx2t99_asimd_store4): New.
11655 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
11657 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
11658 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
11660 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
11661 Segher Boessenkool <segher@kernel.crashing.org>
11664 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
11667 2018-05-17 Richard Biener <rguenther@suse.de>
11669 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
11670 for pruning loop and prune defs feeding only already visited PHIs.
11672 2018-05-17 Richard Biener <rguenther@suse.de>
11674 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
11676 2018-05-17 Bin Cheng <bin.cheng@arm.com>
11677 Richard Biener <rguenther@suse.de>
11679 PR tree-optimization/85793
11680 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
11681 for VMAT_ELEMENTWISE.
11683 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11685 * internal-fn.h (lookup_internal_fn): Declare
11686 * internal-fn.c (lookup_internal_fn): New function.
11687 * gimple.c (gimple_build_call_from_tree): Handle calls to
11688 internal functions.
11689 * gimple-pretty-print.c (dump_gimple_call): Print "." before
11690 internal function names.
11691 * tree-pretty-print.c (dump_generic_node): Likewise.
11692 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
11694 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
11696 * gimple-fold.h (gimple_build): Make the function forms take
11697 combined_fn rather than built_in_function.
11698 (gimple_simplify): Likewise.
11699 * gimple-match-head.c (gimple_simplify): Likewise.
11700 * gimple-fold.c (gimple_build): Likewise.
11701 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
11702 rather than gimple_build_call_internal.
11703 (get_initial_defs_for_reduction): Likewise.
11704 (vect_create_epilog_for_reduction): Likewise.
11705 (vectorizable_live_operation): Likewise.
11707 2018-05-17 Martin Liska <mliska@suse.cz>
11709 * gimple-ssa-sprintf.c (format_directive): Do not use
11710 space in between 'G_' and '('.
11712 2018-05-17 Jakub Jelinek <jakub@redhat.com>
11715 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
11716 even if the mask is not all ones.
11719 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
11721 (ix86_gimple_fold_builtin): Likewise.
11724 * config/i386/i386.c: Include tree-vector-builder.h.
11725 (ix86_vector_shift_count): New function.
11726 (ix86_fold_builtin): Fold shift builtins by scalar count.
11727 (ix86_gimple_fold_builtin): Likewise.
11729 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
11730 _mm512_setzero): New intrinsics.
11732 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
11733 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11735 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
11736 code generation for cases where splatting a value is not useful.
11737 * simplify-rtx.c (simplify_ternary_operation): Simplify
11738 vec_merge across a vec_duplicate and a paradoxical subreg forming
11739 a vector mode to a vec_concat.
11741 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
11743 * config.gcc: Support "goldmont-plus".
11744 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
11746 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11747 PROCESSOR_GOLDMONT_PLUS.
11748 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
11749 (processor_target_table): Add "goldmont-plus".
11750 (PTA_GOLDMONT_PLUS): Define.
11751 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
11752 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
11753 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
11754 (fold_builtin_cpu): Add "goldmont-plus".
11755 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
11756 (ix86_option_override_internal): Add "goldmont-plus".
11757 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
11758 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
11759 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
11760 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
11762 2018-05-17 Richard Biener <rguenther@suse.de>
11764 PR tree-optimization/85757
11765 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
11766 remove defs that only feed that PHI from further processing.
11768 2018-05-16 Jim Wilson <jimw@sifive.com>
11770 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
11772 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
11774 2018-05-16 Mark Wielaard <mark@klomp.org>
11776 * dwarf2out.c (count_index_strings): New function.
11777 (output_indirect_strings): Call count_index_strings and generate
11778 header for dwarf_version >= 5.
11780 2018-05-16 Mark Wielaard <mark@klomp.org>
11782 * dwarf2out.c (dwarf_FORM): New function.
11783 (set_indirect_string): Use dwarf_FORM.
11784 (reset_indirect_string): Likewise.
11785 (size_of_die): Likewise.
11786 (value_format): Likewise.
11787 (output_die): Likewise.
11788 (add_skeleton_AT_string): Likewise.
11789 (output_macinfo_op): Likewise.
11790 (index_string): Likewise.
11791 (output_index_string_offset): Likewise.
11792 (output_index_string): Likewise.
11793 (count_index_strings): Likewise.
11795 2018-05-16 Carl Love <cel@us.ibm.com>
11797 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11798 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
11800 2018-05-16 Martin Jambor <mjambor@suse.cz>
11802 * ipa-prop.c (ipa_free_all_edge_args): Remove.
11803 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
11805 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
11807 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
11808 (fnma<mode>4): Likewise.
11809 (fms<mode>4): Likewise.
11810 (fnms<mode>4): Likewise.
11811 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
11812 (aarch64_fnma<mode>4): Likewise.
11813 (aarch64_fms<mode>4): Likewise.
11814 (aarch64_fnms<mode>4): Likewise.
11815 (aarch64_fnmadd<mode>4): Likewise.
11817 2018-05-16 Jason Merrill <jason@redhat.com>
11819 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
11821 2018-05-16 Richard Biener <rguenther@suse.de>
11823 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
11824 (dump_stmt_cost): Declare.
11825 (add_stmt_cost): Dump cost we add.
11826 (add_stmt_costs): New function.
11827 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
11828 No longer exported.
11829 (vect_analyze_stmt): Adjust prototype.
11830 (vectorizable_condition): Likewise.
11831 (vectorizable_live_operation): Likewise.
11832 (vectorizable_reduction): Likewise.
11833 (vectorizable_induction): Likewise.
11834 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
11835 cost vector to pass to vectorizable_ and record afterwards.
11836 (vect_model_reduction_cost): Take cost vector argument and adjust.
11837 (vect_model_induction_cost): Likewise.
11838 (vectorizable_reduction): Likewise.
11839 (vectorizable_induction): Likewise.
11840 (vectorizable_live_operation): Likewise.
11841 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
11842 SLP_TREE_NUMBER_OF_VEC_STMTS.
11843 (vect_analyze_slp_cost_1): Remove.
11844 (vect_analyze_slp_cost): Likewise.
11845 (vect_slp_analyze_node_operations): Take visited args and
11846 a target cost vector. Avoid processing already visited stmt sets.
11847 (vect_slp_analyze_operations): Use a local cost vector to gather
11848 costs and register those of non-discarded instances.
11849 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
11850 (vect_schedule_slp_instance): Remove copying of
11851 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
11853 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
11854 adding cost. Record cost entry location.
11855 (vect_prologue_cost_for_slp_op): Function to compute cost of
11856 a constant or invariant generated for SLP vect in the prologue,
11857 split out from vect_analyze_slp_cost_1.
11858 (vect_model_simple_cost): Make static. Adjust for SLP costing.
11859 (vect_model_promotion_demotion_cost): Likewise.
11860 (vect_model_store_cost): Likewise, make static.
11861 (vect_model_load_cost): Likewise.
11862 (vectorizable_bswap): Add cost vector arg and adjust.
11863 (vectorizable_call): Likewise.
11864 (vectorizable_simd_clone_call): Likewise.
11865 (vectorizable_conversion): Likewise.
11866 (vectorizable_assignment): Likewise.
11867 (vectorizable_shift): Likewise.
11868 (vectorizable_operation): Likewise.
11869 (vectorizable_store): Likewise.
11870 (vectorizable_load): Likewise.
11871 (vectorizable_condition): Likewise.
11872 (vectorizable_comparison): Likewise.
11873 (can_vectorize_live_stmts): Likewise.
11874 (vect_analyze_stmt): Likewise.
11875 (vect_transform_stmt): Adjust calls to vectorizable_*.
11876 * tree-vectorizer.c: Include gimple-pretty-print.h.
11877 (dump_stmt_cost): New function.
11879 2018-05-16 Richard Biener <rguenther@suse.de>
11881 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
11882 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
11883 * tree-ssa-dse.c: Include tree-ssa-loop.h.
11884 (check_name): New callback.
11885 (dse_classify_store): Track cycles via a visited bitmap of PHI
11886 defs and simplify handling of in-loop and across loop dead stores
11887 and properly fail for loop-variant refs. Handle byte-tracking with
11888 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
11891 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
11893 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
11894 (vect_get_mask_type_for_stmt): Likewise.
11895 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
11897 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
11898 to determine the statement's vector type and the vector type that
11899 should be used for calculating nunits. Deal with cases in which
11900 the type has to be deferred.
11901 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
11902 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
11903 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
11904 (vect_determine_vf_for_stmt): New functions, split out from...
11905 (vect_determine_vectorization_factor): ...here.
11906 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
11907 (vect_get_mask_type_for_stmt): New functions, split out from
11908 vect_determine_vectorization_factor.
11910 2018-05-16 Richard Biener <rguenther@suse.de>
11912 * tree-cfg.c (verify_gimple_assign_ternary): Properly
11913 verify the [VEC_]COND_EXPR embedded comparison.
11915 2018-05-15 Martin Sebor <msebor@redhat.com>
11917 PR tree-optimization/85753
11918 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
11919 RECORD_TYPE in addition to ARRAY_TYPE.
11921 2018-05-15 Martin Sebor <msebor@redhat.com>
11923 PR middle-end/85643
11924 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
11926 2018-05-15 Richard Biener <rguenther@suse.de>
11928 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
11929 add by_clobber_p one. Change algorithm to collect all defs
11930 representing uses we need to walk and try reducing them to
11931 a single one before failing.
11932 (dse_dom_walker::dse_optimize_stmt): Adjust.
11934 2018-05-13 Mark Wielaard <mark@klomp.org>
11936 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
11937 (size_of_loc_descr): Likewise.
11938 (output_loc_operands): Likewise.
11939 (output_loc_operands_raw): Likewise.
11940 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
11941 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
11942 (hash_loc_operands): Likewise.
11943 (compare_loc_operands): Likewise.
11945 2018-05-14 Mark Wielaard <mark@klomp.org>
11947 * dwarf2out.c (count_index_addrs): New function.
11948 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
11950 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11952 PR tree-optimization/83648
11953 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
11954 return value as malloc candidate.
11956 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11959 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
11960 param as true in call to suggest_attribute.
11962 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
11964 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
11965 -mreadonly-in-sdata.
11967 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11969 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
11971 (aarch64_crypto_aesd_fused): Likewise.
11973 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
11975 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
11976 (movsi_aarch64): Likewise.
11977 (load_pairsi): Likewise.
11978 (load_pairdi): Likewise.
11979 (store_pairsi): Likewise.
11980 (store_pairdi): Likewise.
11981 (load_pairsf): Likewise.
11982 (load_pairdf): Likewise.
11983 (store_pairsf): Likewise.
11984 (store_pairdf): Likewise.
11985 (zero_extend): Likewise.
11986 (trunc): Swap alternatives.
11987 (fcvt_target): Add '?' to prefer w over r.
11989 2018-05-14 Jakub Jelinek <jakub@redhat.com>
11992 * config/i386/i386.md: Disallow non-commutative arithmetics in
11993 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
11994 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
11995 in the peephole2 before it.
11997 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
11999 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12000 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
12001 (ix86_handle_option): Handle -mcldemote.
12002 * config.gcc: New header.
12003 * config/i386/cldemoteintrin.h: New file.
12004 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
12005 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12007 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12008 OPTION_MASK_ISA_CLDEMOTE.
12009 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
12010 (ix86_valid_target_attribute_inner_p): Ditto.
12011 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
12012 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
12013 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
12014 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
12015 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
12017 * config/i386/i386.opt: Add -mcldemote.
12018 * config/i386/x86intrin.h: New header.
12019 * doc/invoke.texi: Add -mcldemote.
12021 2018-05-14 Richard Biener <rguenther@suse.de>
12023 * doc/match-and-simplify.texi: Adjust :s documentation.
12025 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
12027 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
12028 intended memcpy size.
12029 (REORDER_45): Likewise.
12031 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
12033 * sort.cc: New file.
12034 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
12035 * vec.c (qsort_chk): Use gcc_qsort.
12036 * Makefile.in (OBJS-libcommon): Add sort.o.
12037 (build/sort.o): New target. Use it...
12038 (BUILD_RTL): ... here, and...
12039 (build/gencfn-macros): ... here, and...
12040 (build/genmatch): ... here.
12042 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12043 Chung-Ju Wu <jasonwucj@gmail.com>
12045 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
12046 * config/nds32/nds32-graywolf.md: New file.
12047 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
12048 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
12050 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
12051 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
12052 * config/nds32/nds32.md (pipeline_model): Add graywolf.
12053 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
12054 * config/nds32/pipelines.md: Include n15 settings.
12056 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12057 Chung-Ju Wu <jasonwucj@gmail.com>
12059 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
12060 * config/nds32/nds32-n13.md: New file.
12061 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
12062 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
12064 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
12065 * config/nds32/nds32.md (pipeline_model): Add n13.
12066 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
12067 * config/nds32/pipelines.md: Include n13 settings.
12069 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
12070 Chung-Ju Wu <jasonwucj@gmail.com>
12072 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
12073 * config/nds32/nds32-n10.md: New file.
12074 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
12075 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
12077 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
12078 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
12079 * config/nds32/nds32.md (pipeline_model): Add n10.
12080 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
12081 * config/nds32/pipelines.md: Include n10 settings.
12083 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
12084 Kito Cheng <kito.cheng@gmail.com>
12085 Chung-Ju Wu <jasonwucj@gmail.com>
12087 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
12088 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
12089 Add enum values for DSP extension instructions.
12090 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
12092 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
12093 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
12094 New code iterators.
12095 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
12096 * config/nds32/nds32-dspext.md: New file for DSP implementation.
12097 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
12098 * config/nds32/nds32-intrinsic.md: Likewise.
12099 * config/nds32/nds32_intrinsic.h: Likewise.
12100 * config/nds32/nds32-md-auxiliary.c: Likewise.
12101 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
12102 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
12103 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
12104 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
12105 * config/nds32/nds32-protos.h: New declarations for DSP extension.
12106 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
12107 TYPE_DMAC in switch statement.
12108 * config/nds32/nds32.c: New checking and implementation for DSP
12109 extension instructions.
12110 * config/nds32/nds32.h: Likewise.
12111 * config/nds32/nds32.md: Likewise.
12112 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
12113 * config/nds32/predicates.md: Implement new predicates for DSP
12116 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12118 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
12119 Reformat alternatives and attributes so it is easier to identify
12120 which constraints/attributes go with which instruction.
12121 (mov<mode>_hardfloat32, FMOVE64): Likewise.
12122 (mov<mode>_softfloat32, FMOVE64): Likewise.
12123 (mov<mode>_hardfloat64, FMOVE64): Likewise.
12124 (mov<mode>_softfloat64, FMOVE64): Likewise.
12126 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
12128 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12130 (Basic PowerPC Built-in Functions): The new name of the
12131 subsection previously known as "PowerPC Built-in Functions".
12132 (Basic PowerPC Built-in Functions Available on all Configurations):
12134 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
12135 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
12136 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
12137 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
12139 2018-05-11 Martin Jambor <mjambor@suse.cz>
12142 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
12145 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12148 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
12150 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
12152 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
12153 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
12154 (ix86_handle_option): Handle -mwaitpkg.
12155 * config.gcc: New header.
12156 * config/i386/cpuid.h (bit_WAITPKG): New bit.
12157 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
12158 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
12160 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12161 OPTION_MASK_ISA_WAITPKG.
12162 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
12163 (ix86_option_override_internal): Add PTA_WAITPKG.
12164 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
12165 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
12166 IX86_BUILTIN_TPAUSE.
12167 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
12168 __builtin_ia32_umwait and __builtin_ia32_tpause.
12169 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
12170 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
12171 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
12172 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
12173 UNSPECV_TPAUSE): New.
12174 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
12175 * config/i386/i386.opt: Add -mwaitpkg.
12176 * config/i386/waitpkgintrin.h: New file.
12177 * config/i386/x86intrin.h: New header.
12178 * doc/invoke.texi: Add -mwaitpkg.
12180 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
12183 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
12185 (cortex-m0): Use armv6s-m isa.
12186 (cortex-m0plus): Likewise.
12187 (cortex-m1): Likewise.
12188 (cortex-m0.small-multiply): Likewise.
12189 (cortex-m0plus.small-multiply): Likewise.
12190 (cortex-m1.small-multiply): Likewise.
12192 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
12193 Jakub Jelinek <jakub@redhat.com>
12195 PR tree-optimization/85692
12196 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
12197 source permute as well.
12199 2018-05-11 Martin Liska <mliska@suse.cz>
12202 * doc/extend.texi: Document LLVM style format for no_sanitize
12205 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12207 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
12208 mode_supports_vsx_dform_quad to mode_supports_dq_form.
12209 (mode_supports_vsx_dform_quad): Likewise.
12210 (mode_supports_vmx_dform): Move these functions to be next to the
12211 other mode_supports functions.
12212 (mode_supports_dq_form): Likewise.
12213 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
12214 mode_supports_dq_form.
12215 (reg_offset_addressing_ok_p): Likewise.
12216 (offsettable_ok_by_alignment): Likewise.
12217 (rs6000_legitimate_offset_address_p): Likewise.
12218 (legitimate_lo_sum_address_p): Likewise.
12219 (rs6000_legitimize_address): Likewise.
12220 (rs6000_legitimize_reload_address): Likewise.
12221 (rs6000_secondary_reload_inner): Likewise.
12222 (rs6000_preferred_reload_class): Likewise.
12223 (rs6000_output_move_128bit): Likewise.
12225 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12227 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
12228 Generate SImode target register for null target.
12229 <case IX86_BUILTIN_XGETBV>: Ditto.
12230 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
12231 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
12233 2018-05-10 Carl Love <cel@us.ibm.com>
12235 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12236 dcbtt and dcbtstt if operands[2] is 0.
12238 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
12241 * config/i386/sse.md (usadv64qi): New expander.
12243 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
12245 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
12246 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
12247 -maltivec=be support.
12248 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
12249 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
12250 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
12251 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
12252 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
12253 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
12254 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
12255 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
12256 altivec_vsumsws): Adjust.
12257 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
12258 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
12259 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
12261 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
12262 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
12263 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
12264 (altivec_lve<VI_char>x): Delete expand.
12265 (*altivec_lve<VI_char>x_internal): Rename to...
12266 (altivec_lve<VI_char>x): ... this.
12267 (altivec_lvxl_<mode>): Delete expand.
12268 (*altivec_lvxl_<mode>_internal): Rename to ...
12269 (altivec_lvxl_<mode>): ... this.
12270 (altivec_stvxl_<mode>): Delete expand.
12271 (*altivec_stvxl_<mode>_internal): Rename to ...
12272 (altivec_stvxl_<mode>): ... this.
12273 (altivec_stve<VI_char>x): Delete expand.
12274 (*altivec_stve<VI_char>x_internal): Rename to ...
12275 (altivec_stve<VI_char>x): ... this.
12276 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
12277 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
12278 reduc_plus_scal_<mode>): Adjust.
12279 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
12281 (rs6000_cpu_cpp_builtins): Adjust.
12282 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
12283 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
12284 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
12285 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
12286 -maltivec=be support.
12287 (rs6000_split_vec_extract_var): Adjust.
12288 (rs6000_split_v4si_init): Adjust.
12289 (swap_selector_for_mode): Delete.
12290 (altivec_expand_lvx_be, altivec_expand_stvx_be,
12291 altivec_expand_stvex_be): Delete.
12292 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
12293 -maltivec=be support.
12294 (rs6000_gimple_fold_builtin): Ditto.
12295 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
12297 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
12298 (TARGET_DIRECT_MOVE_64BIT): Adjust.
12299 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
12300 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
12301 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
12302 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
12303 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
12304 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
12305 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
12306 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
12307 anonymous split): Adjust.
12308 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
12309 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
12311 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
12313 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
12314 when --with-gxx-include-dir is also specified.
12315 * configure: Regenerate.
12317 2018-05-09 Jim Wilson <jimw@sifive.com>
12320 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
12321 * config/riscv/t-withmultilib: New.
12322 * config/riscv/withmultilib.h: New.
12323 * doc/install.texi: Document RISC-V --with-multilib-list support.
12325 2018-05-09 Richard Biener <rguenther@suse.de>
12327 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
12329 (vect_bb_vectorization_profitable_p): Adjust. Compute
12330 actual scalar cost using the cost vector and the add_stmt_cost
12333 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12335 PR rtl-optimization/85645
12336 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
12337 in the REG_CFA_REGISTER note for LR, don't leave it empty.
12339 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12341 PR rtl-optimization/85645
12342 * shrink-wrap.c (spread_components): Return a boolean saying if
12343 anything was changed.
12344 (try_shrink_wrapping_separate): Iterate spread_components until
12345 nothing changes anymore.
12347 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12349 PR rtl-optimization/85645
12350 * regrename.c (build_def_use): Also kill the chains that include the
12351 destination of a REG_CFA_REGISTER note.
12353 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
12355 PR rtl-optimization/85645
12356 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
12357 insn that has a REG_CFA_REGISTER note.
12359 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
12361 * cfgexpand.c (expand_clobber): New function.
12362 (expand_gimple_stmt_1): Use it.
12363 * tree-vect-stmts.c (vect_clobber_variable): New function,
12365 (vectorizable_simd_clone_call): ...here.
12366 (vectorizable_store): Emit a clobber either side of an
12367 IFN_STORE_LANES sequence.
12368 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
12370 2018-05-09 Tom de Vries <tom@codesourcery.com>
12373 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
12374 (define_insn "trap_if_false"): Add exit after trap.
12376 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
12378 PR rtl-optimization/85638
12379 * bb-reorder.c: Include common/common-target.h.
12380 (create_forwarder_block): New function extracted from...
12381 (fix_up_crossing_landing_pad): ...here. Rename into...
12382 (dw2_fix_up_crossing_landing_pad): ...this.
12383 (sjlj_fix_up_crossing_landing_pad): New function.
12384 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
12385 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
12386 from both partitions and exit the loop after one iteration.
12388 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12391 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12393 (Basic PowerPC Built-in Functions): The new name of the
12394 subsection previously known as "PowerPC Built-in Functions".
12395 (Basic PowerPC Built-in Functions Available on all Configurations):
12397 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12399 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12401 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12403 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12406 2018-05-08 Jim Wilson <jimw@sifive.com>
12408 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
12409 (LD_EMUL_SUFFIX): New.
12410 (LINK_SPEC): Use it.
12412 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
12414 * doc/extend.texi (PowerPC Built-in Functions): Rename this
12416 (Basic PowerPC Built-in Functions): The new name of the
12417 subsection previously known as "PowerPC Built-in Functions".
12418 (Basic PowerPC Built-in Functions Available on all Configurations):
12420 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
12422 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
12424 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
12426 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
12429 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12432 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
12433 after cmpelim optimization.
12435 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
12437 * config.gcc: Support "goldmont".
12438 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
12439 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12440 PROCESSOR_GOLDMONT.
12441 * config/i386/i386.c (m_GOLDMONT): Define.
12442 (processor_target_table): Add "goldmont".
12443 (PTA_GOLDMONT): Define.
12444 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
12445 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
12446 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
12447 (fold_builtin_cpu): Add "goldmont".
12448 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
12449 (ix86_option_override_internal): Add "goldmont".
12450 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
12451 (processor_type): Add PROCESSOR_GOLDMONT.
12452 * config/i386/i386.md: Add CPU "glm".
12453 * config/i386/glm.md: New file.
12454 * config/i386/x86-tune.def: Add m_GOLDMONT.
12455 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
12457 2018-05-08 Jakub Jelinek <jakub@redhat.com>
12460 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
12462 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
12463 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
12464 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
12467 * config/i386/i386.c (ix86_fold_builtin): Handle
12468 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
12471 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
12472 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
12474 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
12477 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
12478 (check_arch): Likewise.
12479 (check_fpu): Return the result rather than printing it.
12480 (end arch): Fix operator precedence.
12481 (end cpu): Likewise.
12482 (END): Print the result from check_fpu.
12484 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12485 Alan Hayward <alan.hayward@arm.com>
12486 David Sherwood <david.sherwood@arm.com>
12488 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
12489 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
12490 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
12491 (*fcmuo<mode>_and): New patterns.
12493 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
12495 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
12496 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
12497 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
12498 (cmp_op, sve_imm_con): New code attributes.
12499 (SVE_COND_INT_CMP, imm_con): Delete.
12500 (cmp_op): Remove above unspecs from int attribute.
12501 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
12503 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
12504 comparison-specific unspecs.
12505 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
12506 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
12507 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
12508 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
12509 (*vec_fcm<cmp_op><mode>): Rename to...
12510 (*fcm<cmp_op><mode>): ...this and adjust likewise.
12511 (*vec_fcmuo<mode>): Rename to...
12512 (*fcmuo<mode>): ...this and adjust likewise.
12513 (*pred_fcm<cmp_op><mode>): New pattern.
12514 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
12515 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
12517 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
12519 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
12520 (aarch64_emit_sve_predicated_cond): New function.
12521 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
12522 (aarch64_emit_unspec_cond_or): Replace with...
12523 (aarch64_emit_sve_or_conds): ...this new function. Use
12524 aarch64_emit_sve_ptrue_op for the individual comparisons and
12525 aarch64_emit_binop to OR them together.
12526 (aarch64_emit_inverted_unspec_cond): Replace with...
12527 (aarch64_emit_sve_inverted_cond): ...this new function. Use
12528 aarch64_emit_sve_ptrue_op for the comparison and
12529 aarch64_emit_unop to invert the result.
12530 (aarch64_expand_sve_vec_cmp_float): Update after the above
12531 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
12533 2018-05-07 Nathan Sidwell <nathan@acm.org>
12535 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
12536 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
12537 (Backwards Compatibility): Likewise.
12539 2018-05-07 Luis Machado <luis.machado@linaro.org>
12543 2018-05-07 Luis Machado <luis.machado@linaro.org>
12545 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12546 <prefetch_dynamic_strides>: New const bool field.
12547 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12548 prefetch_dynamic_strides.
12549 (exynosm1_prefetch_tune): Likewise.
12550 (thunderxt88_prefetch_tune): Likewise.
12551 (thunderx_prefetch_tune): Likewise.
12552 (thunderx2t99_prefetch_tune): Likewise.
12553 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12555 (aarch64_override_options_internal): Update to set
12556 PARAM_PREFETCH_DYNAMIC_STRIDES.
12557 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12558 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12559 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12560 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12561 prefetch-dynamic-strides setting.
12563 2018-05-07 Luis Machado <luis.machado@linaro.org>
12565 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12566 <minimum_stride>: New const int field.
12567 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12568 minimum_stride field.
12569 (exynosm1_prefetch_tune): Likewise.
12570 (thunderxt88_prefetch_tune): Likewise.
12571 (thunderx_prefetch_tune): Likewise.
12572 (thunderx2t99_prefetch_tune): Likewise.
12573 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12574 (aarch64_override_options_internal): Update to set
12575 PARAM_PREFETCH_MINIMUM_STRIDE.
12576 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12577 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12578 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12579 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12580 stride is constant and is below the minimum stride threshold.
12582 2018-05-07 Luis Machado <luis.machado@linaro.org>
12584 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
12587 2018-05-07 Luis Machado <luis.machado@linaro.org>
12589 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12590 <prefetch_dynamic_strides>: New const bool field.
12591 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12592 prefetch_dynamic_strides.
12593 (exynosm1_prefetch_tune): Likewise.
12594 (thunderxt88_prefetch_tune): Likewise.
12595 (thunderx_prefetch_tune): Likewise.
12596 (thunderx2t99_prefetch_tune): Likewise.
12597 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
12599 (aarch64_override_options_internal): Update to set
12600 PARAM_PREFETCH_DYNAMIC_STRIDES.
12601 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
12602 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
12603 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
12604 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
12605 prefetch-dynamic-strides setting.
12607 2018-05-07 Luis Machado <luis.machado@linaro.org>
12609 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
12610 <minimum_stride>: New const int field.
12611 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
12612 minimum_stride field.
12613 (exynosm1_prefetch_tune): Likewise.
12614 (thunderxt88_prefetch_tune): Likewise.
12615 (thunderx_prefetch_tune): Likewise.
12616 (thunderx2t99_prefetch_tune): Likewise.
12617 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
12618 (aarch64_override_options_internal): Update to set
12619 PARAM_PREFETCH_MINIMUM_STRIDE.
12620 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
12621 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
12622 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
12623 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
12624 stride is constant and is below the minimum stride threshold.
12626 2018-05-06 Jakub Jelinek <jakub@redhat.com>
12629 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
12630 the type is addressable. Don't force op into register if it has
12633 2018-05-05 Roland McGrath <mcgrathr@google.com>
12636 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
12637 any section for which we don't know a specific type it should have,
12638 regardless of name. Previously this was done only for the exact
12639 names ".init_array", ".fini_array", and ".preinit_array".
12640 (default_elf_asm_named_section): Add comment about
12641 relationship with default_section_type_flags and SECTION_NOTYPE.
12642 (get_section): Don't consider it a type conflict if one side has
12643 SECTION_NOTYPE and the other doesn't, as long as neither has the
12644 SECTION_BSS et al used in the default_section_type_flags logic.
12646 2018-05-05 Tom de Vries <tom@codesourcery.com>
12649 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
12650 (workaround_barsyncs): New function.
12651 (nvptx_reorg): Use workaround_barsyncs.
12652 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
12653 (define_expand "nvptx_membar_cta"): New define_expand.
12654 (define_insn "*nvptx_membar_cta"): New insn.
12656 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
12658 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
12659 To improve optimization opportunities.
12660 * builtin-types.def: The new needed builtin types for the above.
12662 2018-05-04 Richard Biener <rguenther@suse.de>
12664 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
12665 * gimple-ssa-store-merging.c
12666 (imm_store_chain_info::output_merged_store): Remove redundant create,
12667 release split_store vector contents on failure.
12668 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
12669 scalar stmt vector on cache hit.
12671 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
12673 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
12675 * config.gcc (powerpc-xilinx-eabi*): Remove.
12676 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
12678 (fusion_addis_mem_combo_load): Ditto.
12679 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
12681 (rs6000_cpu_cpp_builtins): Ditto.
12682 * config/rs6000/rs6000-linux.c
12683 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
12684 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
12685 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
12687 (rs6000_setup_reg_addr_masks): Ditto.
12688 (rs6000_init_hard_regno_mode_ok): Ditto.
12689 (rs6000_option_override_internal): Ditto.
12690 (legitimate_lo_sum_address_p): Ditto.
12691 (rs6000_legitimize_address): Ditto.
12692 (rs6000_legitimize_reload_address): Ditto.
12693 (rs6000_legitimate_address_p): Ditto.
12694 (abi_v4_pass_in_fpr): Ditto.
12695 (setup_incoming_varargs): Ditto.
12696 (rs6000_gimplify_va_arg): Ditto.
12697 (rs6000_split_multireg_move): Ditto.
12698 (rs6000_savres_strategy): Ditto.
12699 (rs6000_emit_prologue_components): Ditto.
12700 (rs6000_emit_epilogue_components): Ditto.
12701 (rs6000_emit_prologue): Ditto.
12702 (rs6000_emit_epilogue): Ditto.
12703 (rs6000_elf_file_end): Ditto.
12704 (rs6000_function_value): Ditto.
12705 (rs6000_libcall_value): Ditto.
12706 * config/rs6000/rs6000.h: Ditto.
12707 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
12708 (TARGET_MINMAX): ... this. New.
12709 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
12710 * config/rs6000/rs6000.md: Remove Xilinx FP support.
12711 (*movsi_internal1_single): Delete.
12712 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
12713 mfpu=, mxilinx-fpu): Delete.
12714 * config/rs6000/singlefp.h: Delete.
12715 * config/rs6000/sysv4.h: Remove Xilinx FP support.
12716 * config/rs6000/t-rs6000: Ditto.
12717 * config/rs6000/t-xilinx: Delete.
12718 * config/rs6000/titan.md: Adjust for fp_type removal.
12719 * config/rs6000/vsx.md: Remove Xilinx FP support.
12720 (VStype_simple): Delete.
12721 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
12722 * config/rs6000/xfpu.h: Delete.
12723 * config/rs6000/xfpu.md: Delete.
12724 * config/rs6000/xilinx.h: Delete.
12725 * config/rs6000/xilinx.opt: Delete.
12726 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
12727 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
12729 2018-05-04 Tom de Vries <tom@codesourcery.com>
12732 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
12735 2018-05-04 Richard Biener <rguenther@suse.de>
12737 PR middle-end/85627
12738 * tree-complex.c (update_complex_assignment): We are always in SSA form.
12739 (expand_complex_div_wide): Likewise.
12740 (expand_complex_operations_1): Likewise.
12741 (expand_complex_libcall): Preserve EH info of the original stmt.
12742 (tree_lower_complex): Handle removed blocks.
12743 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
12744 on complex multiplication and division libcall builtins.
12746 2018-05-04 Richard Biener <rguenther@suse.de>
12748 PR middle-end/85574
12749 * fold-const.c (negate_expr_p): Restrict negation of operand
12750 zero of a division to when we know that can happen without
12752 (fold_negate_expr_1): Likewise.
12754 2018-05-04 Jakub Jelinek <jakub@redhat.com>
12757 * real.h (real_nextafter): Declare.
12758 * real.c (real_nextafter): New function.
12759 * fold-const-call.c (fold_const_nextafter): New function.
12760 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
12761 CASE_CFN_NEXTTOWARD.
12762 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
12763 even when arg1_mode is different from arg0_mode.
12765 2018-05-03 Nathan Sidwell <nathan@acm.org>
12767 * doc/extend.texi (Deprecated Features): Remove
12768 -ffriend-injection.
12769 (Backwards Compatibility): Likewise.
12770 * doc/invoke.texi (C++ Language Options): Likewise.
12771 (C++ Dialect Options): Likewise.
12773 2018-05-03 Jakub Jelinek <jakub@redhat.com>
12776 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
12777 _mm512_mask_mullox_epi64): New intrinsics.
12779 2018-05-03 Tom de Vries <tom@codesourcery.com>
12782 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12783 dump files): Add offload-tree.
12785 2018-05-03 Richard Biener <rguenther@suse.de>
12787 PR tree-optimization/85615
12788 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
12789 to loops not nested in BBs loop father to avoid creating multi-entry
12792 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12794 PR tree-optimization/70291
12795 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
12796 arguments. Change return type to tree. Emit libcall as a new
12797 statement rather than replacing existing one when inplace_p is true.
12798 (expand_complex_multiplication_components): New function.
12799 (expand_complex_multiplication): Expand floating-point complex
12800 multiplication using the above.
12801 (expand_complex_division): Rename inner_type parameter to type.
12802 Update expand_complex_libcall call-site.
12803 (expand_complex_operations_1): Update expand_complex_multiplication
12804 and expand_complex_division call-sites.
12806 2018-05-02 Jakub Jelinek <jakub@redhat.com>
12809 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
12810 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
12811 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
12812 the highest significant bit of the shift count mask is clear. In
12813 check whether and[sq]i3 is needed verify that all significant bits
12814 of the shift count other than the highest are set.
12816 2018-05-02 Tom de Vries <tom@codesourcery.com>
12819 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
12820 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
12821 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
12822 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
12823 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
12824 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
12825 __builtin_goacc_parlevel_size.
12827 2018-05-02 Richard Biener <rguenther@suse.de>
12829 PR tree-optimization/85597
12830 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
12831 do not use split vect_get_vec_defs call but call vect_get_slp_defs
12834 2018-05-02 Tom de Vries <tom@codesourcery.com>
12837 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12838 dump files): Add ltrans-tree.
12840 2018-05-02 Tom de Vries <tom@codesourcery.com>
12843 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
12844 dump files): Add wpa-ipa.
12846 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
12848 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
12849 powerpc*-*-linux*paired* target.
12850 * config/rs6000/750cl.h: Delete.
12851 * config/rs6000/paired.h: Delete.
12852 * config/rs6000/paired.md: Delete.
12853 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
12855 * config/rs6000/rs6000-builtin.def: Remove paired float support.
12856 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
12857 comment. Remove paired float support.
12858 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
12859 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
12861 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
12862 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
12864 * config/rs6000/rs6000.c: Remove paired float support.
12865 (paired_expand_vector_init, paired_expand_vector_move,
12866 paired_emit_vector_compare, paired_emit_vector_cond_expr,
12867 (paired_expand_lv_builtin, paired_expand_stv_builtin,
12868 paired_expand_builtin, paired_expand_predicate_builtin,
12869 paired_init_builtins): Delete.
12870 * config/rs6000/rs6000.h: Remove paired float support.
12871 * config/rs6000/rs6000.md: Remove paired float support.
12872 (move_from_CR_ov_bit): Delete.
12873 * config/rs6000/rs6000.opt (mpaired): Delete.
12874 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
12875 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
12877 2018-05-02 Richard Biener <rguenther@suse.de>
12879 PR middle-end/85567
12880 * gimplify.c (gimplify_save_expr): When in SSA form allow
12881 SAVE_EXPRs to compute to SSA vars.
12883 2018-05-02 Jakub Jelinek <jakub@redhat.com>
12886 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
12887 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
12888 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
12889 clobber operands[2], instead use a new pseudo. Formatting fixes.
12891 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
12893 PR tree-optimization/85586
12894 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
12895 exit early for statements in the same group if the accesses are
12898 2018-05-02 Tom de Vries <tom@codesourcery.com>
12901 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
12904 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
12906 PR tree-optimization/85143
12907 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
12909 2018-05-01 Tom de Vries <tom@codesourcery.com>
12912 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
12913 not found" error message.
12915 2018-05-01 Tom de Vries <tom@codesourcery.com>
12918 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
12919 * vec.c (test_ordered_remove_if): New function.
12920 (vec_c_tests): Call test_ordered_remove_if.
12921 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
12922 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
12923 * tree-vect-patterns.c (vect_pattern_recog_1): Use
12924 VEC_ORDERED_REMOVE_IF.
12926 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12928 PR tree-optimization/82665
12929 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
12930 pointer subtraction where arguments come from a memchr call.
12932 2018-05-01 Jakub Jelinek <jakub@redhat.com>
12934 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
12935 --push-state --as-needed and --pop-state instead of --as-needed and
12936 --no-as-needed if ld supports it.
12937 * configure: Regenerated.
12940 * doc/install.texi2html: Replace _002d with - and _002a with * in
12941 generated html files using sed.
12943 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12946 * gcc-rich-location.c (blank_line_before_p): New function.
12947 (use_new_line): New function.
12948 (gcc_rich_location::add_fixit_insert_formatted): New function.
12949 * gcc-rich-location.h
12950 (gcc_rich_location::add_fixit_insert_formatted): New function.
12952 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12954 * selftest.c (assert_streq): Rename "expected" and "actual" to
12955 "val1" and "val2". Extend NULL-handling to cover both inputs
12956 symmetrically, while still requiring both to be non-NULL for a pass.
12957 * selftest.h (assert_streq): Rename "expected" and "actual" to
12959 (ASSERT_EQ): Likewise.
12960 (ASSERT_EQ_AT): Likewise.
12961 (ASSERT_KNOWN_EQ): Likewise.
12962 (ASSERT_KNOWN_EQ_AT): Likewise.
12963 (ASSERT_NE): Likewise.
12964 (ASSERT_MAYBE_NE): Likewise.
12965 (ASSERT_MAYBE_NE_AT): Likewise.
12966 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
12967 the assertion to pass.
12968 (ASSERT_STREQ_AT): Likewise.
12970 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
12972 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
12973 interaction with -pie.
12975 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12977 * selftest.h: Fix alphabetization of per-source-file selftest
12980 2018-04-30 Jason Merrill <jason@redhat.com>
12982 PR c++/61982 - dead stores to destroyed objects.
12983 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
12986 2018-04-30 Jason Merrill <jason@redhat.com>
12988 * tree.c (build_clobber): New.
12989 * tree.h: Declare it.
12990 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
12992 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12994 * diagnostic-show-locus.c (layout::layout): Update for
12995 location_get_source_line returning a char_span.
12996 (struct char_span): Move to input.h.
12997 (struct correction): Update for fields in char_span becoming
12999 (struct source_line): Update for location_get_source_line
13000 returning a char_span.
13001 (layout::print_line): Likewise.
13002 * edit-context.c (edited_file::print_content): Likewise.
13003 (edited_file::print_diff_hunk): Likewise.
13004 (edited_file::print_run_of_changed_lines): Likewise.
13005 (edited_file::get_num_lines): Likewise.
13006 (edited_line::edited_line): Likewise.
13007 * final.c (asm_show_source): Likewise.
13008 * input.c (location_get_source_line): Convert return type
13009 from const char * to char_span, losing the final "line_len"
13011 (dump_location_info): Update for the above.
13012 (get_substring_ranges_for_loc): Likewise. Use a char_span
13013 when handling the literal within the line.
13014 (test_reading_source_line): Update for location_get_source_line
13015 returning a char_span.
13016 * input.h (class char_span): Move here from
13017 diagnostic-show-locus.c, converting from a struct to a class.
13018 Make data members private.
13019 (char_span::operator bool): New.
13020 (char_span::length): New.
13021 (char_span::get_buffer): New.
13022 (char_span::operator[]): New.
13023 (char_span::subspan): Make const.
13024 (char_span::xstrdup): New.
13025 (location_get_source_line): Convert return type from const char *
13026 to char_span, losing the final "line_size" param.
13028 2018-04-30 Jan Hubicka <jh@suse.cz>
13030 * lto-wrapper.c (ltrans_priorities): New static var.
13031 (cmp_priority): New.
13032 (run_gcc): Read priorities and if doing parallel build order
13033 the Makefile by them.
13035 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13037 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
13039 2018-04-30 Richard Biener <rguenther@suse.de>
13041 * tree-cfg.c (verify_address): Remove base argument, add
13042 flag whether to check TREE_ADDRESSABLE and do that.
13043 (verify_expr): Remove.
13044 (verify_types_in_gimple_reference): Add pieces from verify_expr.
13045 (verify_gimple_assign_single): Likewise.
13046 (verify_gimple_switch): Likewise.
13047 (verify_expr_location_1): Dereference tp once. Add (disabled)
13048 piece from verify_expr.
13049 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13051 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13053 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
13055 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13057 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
13058 (small_data_pattern): Likewise.
13059 (arc_rewrite_small_data): Likewise.
13060 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
13061 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
13062 (get_symbol_alignment): New function.
13063 (legitimate_small_data_address_p): Likewise.
13064 (legitimate_scaled_address): Update, call
13065 legitimate_small_data_address_p.
13066 (output_sdata): New static variable.
13067 (arc_print_operand): Update how we handle small data operands.
13068 (arc_print_operand_address): Likewise.
13069 (arc_legitimate_address_p): Update, use
13070 legitimate_small_data_address_p.
13071 (arc_rewrite_small_data_p): Remove.
13072 (arc_rewrite_small_data_1): Likewise.
13073 (arc_rewrite_small_data): Likewise.
13074 (small_data_pattern): Likewise.
13075 (compact_sda_memory_operand): Update to use
13076 legitimate_small_data_address_p and get_symbol_alignment.
13077 (prepare_move_operands): Don't rewite sdata pattern.
13078 (prepare_extend_operands): Remove.
13079 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
13081 (zero_extendqisi2): Likewise.
13082 (zero_extendhisi2): Likewise.
13083 (extendqihi2): Likewise.
13084 (extendqisi2): Likewise.
13085 (extendhisi2): Likewise.
13086 (addsi3): Likewise.
13087 (subsi3): Likewise.
13088 (andsi3): Likewise.
13089 * config/arc/constraints.md (Usd): Change it to memory constraint.
13091 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
13093 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
13094 as source of std instructions.
13095 * config/arc/arc.md (movsi_insn): Update pattern predicate to
13096 allow 6-bit constants as source for store instructions.
13097 (movdi_insn): Update instruction pattern to allow 6-bit constants
13098 as source for store instructions.
13100 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13102 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
13104 2018-04-30 Nathan Sidwell <nathan@acm.org>
13105 Sandra Loosemore <sandra@codesourcery.com>
13107 * dumpfile.c (dump_open): Allow '-' for stdout.
13108 * doc/invoke.texi (Developer Options): Document dump filename
13109 determination early. Document stdin/stdout selection.
13111 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
13113 Microblaze Target: PIC data text relative
13115 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
13116 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
13118 * config/microblaze/microblaze.h (microblaze_constant_address_p):
13119 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
13120 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
13121 New addressing mode for data-text relative position indepenedent code.
13122 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
13123 'ADDRESS_SYMBOLIC_TXT_REL'.
13124 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
13125 (microblaze_legitimate_pic_operand): Exclude function calls from
13126 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
13127 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
13129 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13130 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
13131 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
13132 for 'address + offset'.
13133 (microblaze_expand_prologue): Add new function prologue call for
13135 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
13136 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
13137 table in case of TARGET_PIC_DATA_TEXT_REL.
13138 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
13139 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
13140 Add new macros 'UNSPEC_TEXT',
13141 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
13142 + exclude function calls from 'UNSPEC_PLT' in case of data text
13144 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13145 new target hook for generating address diff vector tables in case of
13147 * doc/tm.texi : Regenerate.
13148 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
13149 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
13150 of addr diff vector generation.
13151 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
13152 target hook definition.
13153 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
13154 Add default function for generate_pic_addr_diff_vec -> flag_pic.
13155 * doc/invoke.texi (Add new pic option): Add new microblaze pic
13156 option for data text relative.
13158 2018-04-30 Richard Biener <rguenther@suse.de>
13160 * tree-chrec.h (evolution_function_is_constant_p): Remove
13162 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
13164 2018-04-30 Richard Biener <rguenther@suse.de>
13167 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
13169 2018-04-30 Richard Biener <rguenther@suse.de>
13171 PR tree-optimization/28364
13172 PR tree-optimization/85275
13173 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
13174 copying first exit test.
13176 2018-04-28 Mark Wielaard <mark@klomp.org>
13178 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
13179 dwarf_version >= 5.
13180 (dwarf_AT): Handle DW_AT_addr_base.
13181 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
13183 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
13186 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
13187 (*ashl<dwi>3_doubleword_mask_1): Ditto.
13188 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
13189 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
13191 2018-04-28 Richard Biener <rguenther@suse.de>
13193 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
13194 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
13195 to reflect use. Only add interesting stmts.
13197 2018-04-27 Martin Jambor <mjambor@suse.cz>
13200 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
13201 the jump function allows for passing through aggregate values.
13203 2018-04-27 David Malcolm <dmalcolm@redhat.com>
13205 * input.h (in_system_header_at): Convert from macro to inline
13207 (from_macro_expansion_at): Likewise.
13208 (from_macro_definition_at): Likewise.
13210 2018-04-27 Jeff Law <law@redhat.com>
13212 * config.gcc: Mark tile* targets as deprecated/obsolete.
13214 2018-04-27 Richard Biener <rguenther@suse.de>
13216 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
13219 2018-04-27 Richard Biener <rguenther@suse.de>
13221 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
13223 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
13225 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
13226 with Yd constraint. Set "preferred_for_speed" attribute from
13227 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
13228 with Yd constraint.
13229 (*movdi_internal): Ditto.
13230 (movti_interunit splitters): Remove
13231 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
13232 (movdi_interunit splitters): Ditto.
13233 * config/i386/constraints.md (Ye): Remove.
13234 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
13236 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13239 * config/aarch64/constraints.md (Usg): Limit to 31.
13240 (Usj): Limit to 63.
13242 2018-04-27 Jakub Jelinek <jakub@redhat.com>
13244 PR tree-optimization/85529
13245 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
13246 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
13247 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
13248 zero extension or masking of the MSB bit.
13249 (optimize_range_tests): Add FIRST_BB argument, pass it through
13250 to optimize_range_tests_var_bound.
13251 (maybe_optimize_range_tests, reassociate_bb): Adjust
13252 optimize_range_tests callers.
13254 2018-04-26 Richard Biener <rguenther@suse.de>
13255 Jakub Jelinek <jakub@redhat.com>
13257 * cgraph.h (symbol_table): Just declare debug method here.
13258 * symtab.c (symbol_table::debug): Define.
13260 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
13262 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
13264 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
13266 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
13267 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
13268 (*movdi_internal): Substitute Yi and Yj constraint with x
13269 and Ym and Yn constraint with y constraint. Update "isa"
13270 attribute and set "preferred_for_speed" attribute from
13271 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
13272 (*movsi_internal): Ditto.
13273 (*movdf_internal): Ditto.
13274 (*movsf_internal): Ditto.
13275 (*zero_extendsidi2): Ditto.
13276 * config/i386/sse.md (vec_set<mode>_0): Ditto.
13277 (sse2_loadld): Ditto.
13278 (*vec_extract<ssevecmodelower>_0): Ditto.
13279 (*vec_extractv4si_0_zext_sse4): Ditto.
13280 (vec_concatv2di): Ditto.
13281 (*vec_dup<mode>): Ditto.
13282 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
13283 * config/i386/constraints.md (Yi): Remove.
13288 2018-04-26 Nathan Sidwell <nathan@acm.org>
13290 * dumpfile.c (dump_open): New.
13291 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
13292 (dump_finish): Detect stdio/stderr by value not name.
13294 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
13296 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
13298 2018-04-26 Tom de Vries <tom@codesourcery.com>
13301 * config/nvptx/nvptx.c (verify_neutering_jumps)
13302 (verify_neutering_labels): New function
13303 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
13305 2018-04-26 Tom de Vries <tom@codesourcery.com>
13308 * config/nvptx/nvptx.c (needs_neutering_p): New function.
13309 (nvptx_single): Use needs_neutering_p to skip over insns that do not
13312 2018-04-26 Richard Biener <rguenther@suse.de>
13313 Tom de Vries <tom@codesourcery.com>
13316 * lto-streamer-out.c (output_function): Fixup loops if required to match
13317 discovery done in the reader.
13319 2018-04-26 Richard Biener <rguenther@suse.de>
13321 PR tree-optimization/85116
13322 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
13323 have a loop exit from the single latch predecessor. Remove
13324 case of header with just condition.
13325 (ch_base::copy_headers): Exclude infinite loops from any
13327 (pass_ch::execute): Record exits.
13329 2018-04-26 Richard Biener <rguenther@suse.de>
13331 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
13332 prologue cost vector and pass it to vect_get_load_cost.
13333 (vect_get_peeling_costs_all_drs): Likewise.
13334 (vect_peeling_hash_get_lowest_cost): Likewise.
13335 (vect_enhance_data_refs_alignment): Likewise.
13337 2018-04-26 Richard Biener <rguenther@suse.de>
13339 PR middle-end/85450
13340 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
13341 checking of integer<->pointer conversions.
13342 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
13343 sign-/zero-extending pointer types.
13344 (expand_omp_for_static_chunk): Likewise.
13346 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
13347 Jean Lee <xiaoyur347@gmail.com>
13349 * config/mips/mips.c (mips_asan_shadow_offset): New function.
13350 (TARGET_ASAN_SHADOW_OFFSET): Define.
13351 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
13352 true for -fsanitize=address.
13354 2018-04-25 Mark Wielaard <mark@klomp.org>
13356 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
13359 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13361 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
13362 than "alu", remove explicit "memory" and "imm_disp" attributes.
13363 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
13365 PR middle-end/85414
13366 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
13367 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
13368 gen_lowpart_no_emit.
13370 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
13373 * config/i386/i386.c (ix86_expand_builtin): Change memory
13374 operand to XI, extend p0 to Pmode.
13375 * config/i386/i386.md: Change unspec volatile and operand
13376 1 mode to XI, change operand 0 mode to P.
13378 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13380 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
13381 GET_MODE_MASK before any checking.
13382 (nds32_can_use_bset_p): Likewise.
13383 (nds32_can_use_btgl_p): Likewise.
13385 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13387 * config/nds32/nds32-doubleword.md: New define_split pattern for
13388 illegal register number.
13390 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13392 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
13394 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
13396 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
13398 2018-04-25 Richard Biener <rguenther@suse.de>
13400 * lto-streamer.h (LTO_major_version): Bump to 8.
13402 2018-04-25 Jakub Jelinek <jakub@redhat.com>
13404 * BASE-VER: Set to 9.0.0.
13406 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
13408 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
13409 in __abskf2 and __powikf2.
13411 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13414 * config/aarch64/constraints.md (Usg, Usj): New constraints.
13415 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
13416 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
13417 Use the above on operand 2. Reindent.
13418 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
13420 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13423 * common/config/i386/i386-common.c (ix86_handle_option): Don't
13425 * config/i386/i386.opt (mcet): Removed.
13426 * doc/install.texi: Remove -mcet documentation.
13427 * doc/invoke.texi: Likewise.
13429 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
13432 * doc/install.texi: Remove -mcet from bootstrap-cet.
13434 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13437 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
13438 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
13442 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
13443 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
13444 containing a CONST_VECTOR.
13446 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
13448 * doc/install.texi: Update newlib dependency for nvptx.
13450 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13453 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
13454 instead of INTVAL when shifting x left.
13456 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
13458 PR tree-optimization/85478
13459 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
13460 vect_grouped_store_supported for single element vectors.
13462 2018-04-24 Richard Biener <rguenther@suse.de>
13465 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
13466 load cost increase to the case of non-constant step.
13468 2018-04-24 Jakub Jelinek <jakub@redhat.com>
13471 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
13472 destination if any_malformed_asm.
13474 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
13476 PR middle-end/85496
13477 * expr.c (store_field): In the bitfield case, if the value comes from
13478 a function call and is returned in registers by means of a PARALLEL,
13479 do not change the mode of the temporary unless BLKmode and VOIDmode.
13481 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
13483 PR rtl-optimization/85423
13484 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
13485 dependencies to debug insns when the previous insn is non-debug.
13487 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13489 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
13490 enums into a single definition.
13491 (fls): Fix predicates and printing.
13494 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
13496 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
13497 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
13498 and short u6 immediate.
13499 (check_if_valid_sleep_operand): Remove.
13500 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
13502 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13504 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
13505 flag_always_save_lp condition.
13506 * config/nds32/nds32.opt (malways-save-lp): New option.
13508 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13510 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
13511 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
13512 * config/nds32/nds32.h
13513 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
13514 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
13516 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13518 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
13519 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
13521 2018-04-22 Shiva Chen <shiva0217@gmail.com>
13522 Chung-Ju Wu <jasonwucj@gmail.com>
13524 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
13526 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
13527 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
13529 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13531 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
13533 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13535 * config/nds32/nds32-protos.h (nds32_data_alignment,
13536 nds32_local_alignment): Declare.
13537 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
13538 nds32_local_alignment): New functions.
13539 (TARGET_CONSTANT_ALIGNMENT): Define.
13540 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
13542 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13544 * config/nds32/nds32.c
13545 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
13546 (TARGET_MODES_TIEABLE_P): Likewise.
13548 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
13550 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
13551 level Ofast and Og.
13553 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
13554 Chung-Ju Wu <jasonwucj@gmail.com>
13556 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13557 for unaligned access.
13558 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13560 * config/nds32/nds32-intrinsic.md: Likewise.
13561 * config/nds32/nds32_intrinsic.h: Likewise.
13562 * config/nds32/nds32.h (nds32_builtins): Likewise.
13563 * config/nds32/nds32.opt (munaligned-access): New option.
13564 * config/nds32/nds32.c (nds32_asm_file_start): Display
13565 flag_unaligned_access status.
13567 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13569 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
13570 -mno-relax is present.
13571 * config/riscv/linux.h (LINK_SPEC): Ditto.
13573 2018-04-20 Martin Sebor <msebor@redhat.com>
13576 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
13578 (gimple_fold_builtin_stxcpy_chk): Same.
13579 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
13581 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
13584 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
13585 __powikf2 when long double is IEEE 128-bit.
13587 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
13589 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
13590 step to make sure stack always aligned.
13592 2018-04-20 Carl Love <cel@us.ibm.com>
13595 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
13596 size check for arg0.
13598 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
13599 Tom de Vries <tom@codesourcery.com>
13602 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
13603 Emit insns for calls too.
13604 (nvptx_find_par): Always look for worker-level predecessor insn.
13605 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
13607 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
13608 (nvptx_process_pars): Propagate frames for calls.
13610 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
13613 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
13615 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
13616 (ix86_handle_option): Don't handle OPT_mibt.
13617 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
13619 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
13621 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
13622 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
13623 (ix86_target_macros): Define __CET__ with flag_cf_protection
13624 for -fcf-protection.
13625 * config/i386/i386.c (isa2_opts): Remove -mibt.
13626 * config/i386/i386.h (TARGET_IBT): Removed.
13627 (TARGET_IBT_P): Likewise.
13628 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
13629 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
13630 * config/i386/i386.opt (mcet): Update help message.
13631 (mshstk): Likewise.
13633 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
13634 -mcet as an alias for -mshstk.
13636 2018-04-20 Richard Biener <rguenther@suse.de>
13638 PR middle-end/85475
13639 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
13640 complexity by forcing a single use of the multiply operand.
13642 2018-04-20 Martin Jambor <mjambor@suse.cz>
13645 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
13646 recursion dependency to only apply to non-clones.
13648 2018-04-20 Martin Jambor <mjambor@suse.cz>
13651 * ipa-cp.c (create_specialized_node): Check that clones of
13652 self-recursive edges exist during IPA-CP.
13654 2018-04-19 Toon Moene <toon@moene.org>
13656 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
13659 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13661 PR tree-optimization/85467
13662 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
13663 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
13664 VECTOR_CST element to type.
13666 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13669 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
13670 * config/i386/i386.md (builtin_setjmp_setup): Removed.
13671 (builtin_longjmp): Likewise.
13672 (save_stack_nonlocal): New pattern.
13673 (restore_stack_nonlocal): Likewise.
13675 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13678 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13679 Replace ASM_OUTPUT_LABEL with fprintf.
13681 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
13684 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
13685 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
13686 * config/i386/i386-c.c (ix86_target_macros_internal): Also
13687 define __IBT__ and __SHSTK__ for -fcf-protection.
13688 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
13690 (ix86_trampoline_init): Likewise.
13691 (x86_output_mi_thunk): Likewise.
13692 (ix86_notrack_prefixed_insn_p): Likewise.
13693 (ix86_option_override_internal): Don't disallow -fcf-protection.
13694 * config/i386/i386.md (rdssp<mode>): Also enable for
13696 (incssp<mode>): Likewise.
13697 (nop_endbr): Likewise.
13698 * config/i386/i386.opt (mcet): Change help message to built-in
13701 (mshstk): Likewise.
13702 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
13703 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
13704 enable CET built-in functions.
13706 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
13708 * common/config/i386/i386-common.c
13709 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
13710 OPTION_MASK_ISA_MOVDIRI_UNSET,
13711 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
13712 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
13713 * config.gcc (movdirintrin.h): New header.
13714 * config/i386/cpuid.h (bit_MOVDIRI,
13715 bit_MOVDIR64B): New bits.
13716 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
13718 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
13719 (VOID, PVOID, PCVOID)): New function types.
13720 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
13721 __builtin_ia32_directstoreu_u64,
13722 __builtin_ia32_movdir64b): New builtins.
13723 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
13724 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
13726 (ix86_valid_target_attribute_inner_p): Ditto.
13727 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
13728 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
13729 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
13730 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
13731 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
13732 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
13733 (movdiri<mode>, movdir64b_<mode>): New.
13734 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
13735 * config/i386/immintrin.h: Include movdirintrin.h.
13736 * config/i386/movdirintrin.h: New file.
13737 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
13739 2018-04-19 Richard Biener <rguenther@suse.de>
13741 PR middle-end/85455
13742 * cfg.c (clear_bb_flags): When loop state says we have
13743 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
13745 2018-04-19 Richard Biener <rguenther@suse.de>
13747 PR tree-optimization/84737
13748 * tree-vect-data-refs.c (vect_copy_ref_info): New function
13749 copying restrict info.
13750 (vect_setup_realignment): Use it.
13751 * tree-vectorizer.h (vect_copy_ref_info): Declare.
13752 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
13753 the first DR to all generated stores.
13754 (vectorizable_load): Likewise for loads.
13756 2018-04-19 Jakub Jelinek <jakub@redhat.com>
13758 PR tree-optimization/85446
13759 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
13760 the integral and pointer types to have the same precision.
13762 * doc/install.texi: Document --disable-cet being the default and
13765 2018-04-18 Martin Liska <mliska@suse.cz>
13767 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
13770 2018-04-18 Martin Liska <mliska@suse.cz>
13773 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13776 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13777 arguments if they are comparable.
13779 2018-04-18 Martin Liska <mliska@suse.cz>
13782 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13785 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13788 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
13791 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
13792 ENDBR after calling __morestack.
13794 2018-04-18 David Malcolm <dmalcolm@redhat.com>
13797 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
13798 by using gcc_base_ver to generate a gcc_driver_version, and use
13799 it when generating GCC_DRIVER_NAME.
13800 * configure: Regenerate.
13802 2018-04-18 Jakub Jelinek <jakub@redhat.com>
13805 * config.gcc: Obsolete powerpc*-*-*spe*.
13807 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13810 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
13811 (stabstr_D): Change type of unum from unsigned int to
13812 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
13815 2018-04-17 Jim Wilson <jimw@sifive.com>
13818 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
13819 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
13820 Set arg_pointer_offset after using pretend_args_size.
13822 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13824 PR rtl-optimization/85431
13825 * dse.c (record_store): Ignore zero width stores.
13828 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
13829 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
13830 __builtin_stack_restore rather than after it.
13831 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
13832 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
13833 argument instead of virtual_dynamic_stack_rtx.
13835 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
13837 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
13839 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
13840 Add note to error message to explain internal mapping of overloaded
13841 built-in function name to non-overloaded built-in function name.
13842 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
13845 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
13848 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
13849 where the inputs overlap with the output.
13851 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13853 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
13854 (=v, v) alternative and explicit "memory" attribute.
13855 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
13856 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13858 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
13859 "sselog1" type instead of "sselog".
13860 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
13861 "sselog". Remove explicit "memory" attribute.
13862 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
13863 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13865 (vec_extract_hi_v32hi): Merge all alternatives into one, use
13866 "sselog1" type instead of "sselog". Remove explicit "memory"
13868 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
13869 use "sselog1" type instead of "sselog". Remove explicit "memory"
13871 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
13872 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
13874 (vec_extract_hi_v64qi): Merge all alternatives into one, use
13875 "sselog1" type instead of "sselog". Remove explicit "memory"
13877 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
13878 use "sselog1" type instead of "sselog". Remove explicit "memory"
13882 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
13884 PR middle-end/85414
13885 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
13888 2018-04-17 Martin Jambor <mjambor@suse.cz>
13891 * ipa-cp.c (create_specialized_node): Call
13892 expand_all_artificial_thunks if necessary.
13894 2018-04-17 Martin Liska <mliska@suse.cz>
13897 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
13898 in message, remote space in between '_G' and '('.
13900 2018-04-17 Jakub Jelinek <jakub@redhat.com>
13903 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
13904 avx512f_vmcmp<mode>3<round_saeonly_name>,
13905 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
13906 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
13907 avx512f_rndscale<mode><round_saeonly_name>,
13908 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
13909 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
13910 Use %<iptr>2 instead of %2 for -masm=intel.
13911 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
13912 avx512f_vcvttss2usi<round_saeonly_name>,
13913 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
13915 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
13916 avx512f_vcvttsd2usi<round_saeonly_name>,
13917 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
13918 Use %q1 instead of %1 for -masm=intel.
13919 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
13920 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
13921 of %3 for -masm=intel.
13922 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
13924 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
13926 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
13928 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
13929 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
13931 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
13933 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
13934 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
13935 %g1 and one with %0 and %1.
13936 (avx512er_vmrcp28<mode><round_saeonly_name>,
13937 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
13938 %1 for -masm=intel.
13939 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
13940 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
13941 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
13942 of %0 and %{%4%} for -masm=intel.
13943 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
13944 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
13945 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
13946 order of %0 and %{%5%}%{z%} for -masm=intel.
13948 2018-04-17 Jan Hubicka <jh@suse.cz>
13951 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
13953 2018-04-17 Martin Liska <mliska@suse.cz>
13956 * multiple_target.c (create_dispatcher_calls): Set apostrophes
13957 for target_clone error message. Make default implementation
13958 clone to be a local declaration.
13959 (separate_attrs): Add new argument and check for an empty
13961 (expand_target_clones): Handle it.
13962 (ipa_target_clone): Make redirection just for target_clones
13965 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
13966 Tom de Vries <tom@codesourcery.com>
13968 PR middle-end/84955
13969 * omp-expand.c (expand_oacc_for): Add dummy false branch for
13970 tiled basic blocks without omp continue statements.
13972 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
13975 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
13976 vec_extract expression as having side effects to make sure it gets
13979 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
13982 * config/i386/i386.c (get_builtin_code_for_version): Check
13985 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
13988 * config.gcc: Support "skylake".
13989 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13991 * config/i386/i386.c (m_SKYLAKE): Define.
13992 (processor_target_table): Add "skylake".
13993 (ix86_option_override_internal): Add "skylake".
13994 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
13995 PROCESSOR_CANNONLAKE.
13996 (get_builtin_code_for_version): Fix priority for
13997 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
13998 PROCESSOR_SKYLAKE-AVX512.
13999 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14000 (processor_type): Add PROCESSOR_SKYLAKE.
14002 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
14003 Jason Merrill <jason@redhat.com>
14006 * convert.c (convert_to_integer_1): Use direct recursion for
14007 enumeral types and types with a precision less than the number
14008 of bits in their mode.
14010 2018-04-16 Julia Koval <julia.koval@intel.com>
14013 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
14014 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
14016 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
14019 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
14020 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
14021 and -mno-direct-move.
14023 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
14026 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
14027 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
14028 Ensure negative shifts result in {0}.
14030 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
14032 PR rtl-optimization/79916
14033 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
14034 regs (if any) to define how to gnerate SD moves when LRA is in
14037 2018-04-13 Jakub Jelinek <jakub@redhat.com>
14039 PR rtl-optimization/85393
14040 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
14041 * except.c (expand_dw2_landing_pad_for_region): Make static.
14042 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
14043 a label and unconditional jump to old_bb, rather than
14044 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
14047 PR rtl-optimization/85376
14048 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
14049 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
14050 instead of a specific value.
14052 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14053 Bin Cheng <bin.cheng@arm.com>
14055 PR tree-optimization/82965
14056 PR tree-optimization/83991
14057 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
14058 by_profile_only parameter.
14059 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
14060 information if the loop was predicted to iterate too many times.
14061 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
14063 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
14066 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
14069 2018-04-13 Martin Liska <mliska@suse.cz>
14070 Jakub Jelinek <jakub@redhat.com>
14072 PR middle-end/81657
14073 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
14074 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
14075 * builtins.c (expand_builtin_memory_copy_args): Use
14076 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
14077 handle dest_addr == pc_rtx.
14079 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
14082 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
14083 asked to not generate direct moves.
14084 (fix_trunc<mode>si2_stfiwx): Similar.
14085 (fix_trunc<mode>si2_internal): Similar.
14087 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14090 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
14091 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
14092 lookup if dest in some wider mode is known to be const0_rtx and
14093 if so, record permanent equivalence for it to be ZERO_EXTEND of
14094 the narrower mode destination.
14096 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14098 * lto-streamer-out.c (output_function): Revert 259346.
14099 * omp-expand.c (expand_oacc_for): Likewise.
14101 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
14103 PR rtl-optimization/85354
14104 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
14105 * sel-sched.c (sel_global_init): ... here.
14107 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
14110 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
14111 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
14112 mode for PE-COFF targets.
14113 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
14114 (i386_pe_asm_lto_end): Likewise.
14115 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
14116 (TARGET_ASM_LTO_END): Likewise.
14117 * config/i386/winnt.c (saved_debug_info_level): New static variable.
14118 (i386_pe_asm_lto_start): New function.
14119 (i386_pe_asm_lto_end): Likewise.
14121 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
14122 Richard Biener <rguenther@suse.de>
14124 PR middle-end/84955
14125 * lto-streamer-out.c (output_function): Fix CFG loop state before
14127 * omp-expand.c (expand_oacc_for): Handle calls to internal
14128 functions like regular functions.
14130 2018-04-12 Richard Biener <rguenther@suse.de>
14133 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
14134 for the early LTO debug to properly generate references to it
14135 during DIE emission. Do not re-use that for the skeleton for
14137 (dwarf2out_early_finish): Likewise.
14139 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14142 * config/i386/sse.md
14143 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
14144 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
14145 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
14146 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
14147 and output is a reg, avoid creating invalid lowpart subreg, but
14148 instead split into a 512-bit move. Don't split if not AVX512VL,
14149 input is xmm16+ reg and output is a mem.
14150 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
14151 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14152 xmm16+ reg and output is a mem.
14154 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14156 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
14157 also for flag_dwarf2_cfi_asm.
14159 2018-04-12 Jakub Jelinek <jakub@redhat.com>
14161 PR rtl-optimization/85342
14162 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
14163 a bool scalar var inside of the loop instead. Don't try to update
14164 recog_data.operand after failed apply_change_group.
14166 2018-04-12 Tom de Vries <tom@codesourcery.com>
14169 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
14170 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
14171 array with flexible array member as array without given dimension.
14172 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
14173 argument for undefined param to true.
14175 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
14178 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14179 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
14180 from PowerPC section.
14181 * config/rs6000/sysv4.opt (mcall-): Improve help text.
14182 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
14183 help text that is too long.
14184 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
14185 help text that is too long.
14186 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
14187 help text that is too long.
14189 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
14191 * config/alpha/alpha.md (stack_probe_internal): Rename
14192 from "probe_stack". Update all callers.
14194 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14196 PR rtl-optimization/84566
14197 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
14198 sched_macro_fuse_insns.
14200 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14203 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
14204 (compute_block_dependences): ... from here.
14206 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14208 PR tree-optimization/85331
14209 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
14210 from int to HOST_WIDE_INT.
14212 2018-04-11 Martin Jambor <mjambor@suse.cz>
14215 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
14216 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
14217 not the same as the source val.
14218 (cgraph_edge_brings_value_p): New parameter.
14219 (gather_edges_for_value): Pass destination value to
14220 cgraph_edge_brings_value_p.
14221 (perhaps_add_new_callers): Likewise.
14222 (get_info_about_necessary_edges): Likewise and exclude values brought
14223 only by self-recursive edges.
14224 (create_specialized_node): Redirect only clones of self-calling edges.
14225 (+self_recursive_pass_through_p): New function.
14226 (find_more_scalar_values_for_callers_subset): Use it.
14227 (find_aggregate_values_for_callers_subset): Likewise.
14228 (known_aggs_to_agg_replacement_list): Removed.
14229 (decide_whether_version_node): Re-calculate known constants for all
14230 remaining context clones.
14232 2018-04-11 Richard Biener <rguenther@suse.de>
14235 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
14236 from early DWARF output.
14237 (dwarf2out_early_finish): Output line info unconditionally into
14238 early DWARF and add reference to it.
14240 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14243 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
14244 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
14245 other than V2DFmode using iptr mode attribute.
14246 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
14248 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
14250 PR rtl-optimization/84659
14251 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
14253 2018-04-11 Jakub Jelinek <jakub@redhat.com>
14256 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
14258 (output_loc_list): Pass address of a dummy size variable even in the
14259 locview handling loop.
14260 (index_location_lists): Add comment on why skip_loc_list_entry can't
14263 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
14266 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
14269 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
14272 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
14273 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
14274 and -mstring-compare-inline-limit.
14276 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14279 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
14280 for stack clash protection in a register whenever we need it to be in
14283 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14285 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
14286 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
14288 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
14291 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
14293 (mlong-double-): Ditto.
14294 * config/rs6000/sysv4.opt (msdata=): Ditto.
14295 (mtls-size=): Ditto.
14297 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
14299 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
14300 erroneous entries for
14301 "vector int vec_ldl (int, long int *)", and
14302 "vector unsigned int vec_ldl (int, unsigned long int *)".
14303 Add comments and entries for
14304 "vector bool char vec_ldl (int, bool char *)",
14305 "vector bool short vec_ldl (int, bool short *)",
14306 "vector bool int vec_ldl (int, bool int *)",
14307 "vector bool long long vec_ldl (int, bool long long *)",
14308 "vector pixel vec_ldl (int, pixel *)",
14309 "vector long long vec_ldl (int, long long *)",
14310 "vector unsigned long long vec_ldl (int, unsigned long long *)".
14311 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
14312 type tree bool_long_long_type_node and correct definition of
14313 bool_V2DI_type_node to make reference to this new type tree.
14314 (rs6000_mangle_type): Replace erroneous reference to
14315 bool_long_type_node with bool_long_long_type_node.
14316 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
14317 comments to emphasize sign distinctions for char and int types and
14318 replace RS6000_BTI_bool_long constant with
14319 RS6000_BTI_bool_long_long constant. Also add comment to restrict
14320 use of RS6000_BTI_pixel.
14321 (bool_long_type_node): Remove this macro definition.
14322 (bool_long_long_type_node): New macro definition
14324 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14326 PR rtl-optimization/85300
14327 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
14328 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
14329 simplify_unary_operation fails.
14331 2018-04-10 Martin Liska <mliska@suse.cz>
14333 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
14334 cgraph_edge and ipa_ref.
14336 2018-04-10 Jakub Jelinek <jakub@redhat.com>
14340 * config/i386/sse.md
14341 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
14342 computation of the VEC_MERGE selector from mask.
14343 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
14344 Fix decoding of the VEC_MERGE selector into mask.
14346 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14348 PR tree-optimization/85286
14349 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
14351 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
14353 * final.c (final_1): Set insn_last_address as well as
14354 insn_current_address.
14356 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14359 * explow.c (emit_stack_probe): Call validize_mem on memory location
14360 before passing it to gen_probe_stack. Create address operand and
14361 legitimize it for the probe_stack_address case.
14363 2018-04-09 Jan Hubicka <jh@suse.cz>
14366 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
14367 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
14368 * tree.c (free_lang_data_in_type): Fix handling of binfos;
14370 (free_lang_data): Rebuild type inheritance graph.
14372 2018-04-09 Martin Sebor <msebor@redhat.com>
14374 * invoke.texi (-finline-small-functions): Mention other optimization
14376 (-findirect-inlining, -fpartial-inlining): Same.
14377 (-finline-functions-called-once): Same.
14378 (-freorder-blocks-and-partition): Same.
14380 2018-04-09 Jan Hubicka <jh@suse.cz>
14383 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
14384 jumps; choose last target that matches the criteria (i.e.
14385 no partition changes for non-crossing jumps).
14386 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
14387 support for redirecting crossing jumps to non-crossing.
14389 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
14391 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
14392 also for naked functions.
14394 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
14396 * config/arc/arc.md (add_shift): New pattern.
14397 (add_shift2): Likewise.
14398 (sub_shift): Likewise.
14399 (sub_shift_cmp0_noout): Likewise.
14400 (compare_si_ashiftsi): Likewise.
14401 (xbfu_cmp0_noout): New combine pattern.
14402 (xbfu_cmp0"): Likewise.
14403 (movsi_set_cc_insn): Place the predicable variant first.
14404 (commutative_binary_cmp0_noout): Remove clobber.
14405 (commutative_binary_cmp0): New pattern.
14406 (noncommutative_binary_cmp0): Likewise.
14407 (noncommutative_binary_cmp0_noout): Likewise.
14408 (noncommutative_binary_comparison_result_used): Removed.
14409 (rsub_cmp0): New pattern.
14410 (rsub_cmp0_noout): Likewise.
14411 (extzvsi): Changed, keep only meaningful variants.
14412 (SQH, SEZ): New iterators.
14413 (SQH_postfix): New mode attribute.
14414 (SEZ_prefix): New code attribute.
14415 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
14416 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
14417 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
14418 of numerical value.
14419 (noncommutative_operator): Check the availability of barrel
14422 2018-04-09 Richard Biener <rguenther@suse.de>
14424 PR tree-optimization/85284
14425 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
14426 Only use the niter constraining form of simple_iv when the exit
14427 is always executed.
14429 2018-04-09 Tom de Vries <tom@codesourcery.com>
14432 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14433 (define_expand "*memory_barrier"): New define_expand.
14434 (define_insn "memory_barrier"): New insn.
14436 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14438 PR rtl-optimization/80463
14439 PR rtl-optimization/83972
14440 PR rtl-optimization/83480
14442 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
14443 correct producer for the insn.
14444 (tidy_control_flow): Fixup seqnos in case of debug insns.
14446 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14448 PR rtl-optimization/83913
14450 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
14451 different sched-times when merging exprs.
14453 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14455 PR rtl-optimization/83962
14457 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
14458 tidy_fallthru_edge and tidy_control_flow.
14460 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
14462 PR rtl-optimization/83530
14464 * sel-sched.c (force_next_insn): New global variable.
14465 (remove_insn_for_debug): When force_next_insn is true, also leave only
14466 next insn in the ready list.
14467 (sel_sched_region): When the region wasn't scheduled, make another pass
14468 over it with force_next_insn set to 1.
14470 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14472 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
14474 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14475 for interrupt control.
14476 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14477 functions for interrupt control.
14478 * config/nds32/nds32-intrinsic.md: Likewise.
14479 * config/nds32/nds32_intrinsic.h: Likewise.
14480 * config/nds32/nds32.h (nds32_builtins): Likewise.
14482 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14484 * config/nds32/nds32.c (nds32_init_machine_status,
14485 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
14486 strict_aligned_p field.
14487 (nds32_expand_to_rtl_hook): New function.
14488 (TARGET_EXPAND_TO_RTL_HOOK): Define.
14489 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
14491 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14492 Chung-Ju Wu <jasonwucj@gmail.com>
14494 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
14495 * config/nds32/nds32-n7.md: New file.
14496 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
14497 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
14499 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
14500 * config/nds32/nds32.md (pipeline_model): Add n7.
14501 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
14502 * config/nds32/pipelines.md: Include n7 settings.
14504 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14505 Chung-Ju Wu <jasonwucj@gmail.com>
14507 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
14508 * config/nds32/nds32-e8.md: New file.
14509 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
14510 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
14512 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
14513 * config/nds32/nds32.md (pipeline_model): Add e8.
14514 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
14515 * config/nds32/pipelines.md: Include e8 settings.
14517 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14518 Chung-Ju Wu <jasonwucj@gmail.com>
14520 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
14521 * config/nds32/nds32-n8.md: New file.
14522 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
14523 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
14525 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
14526 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
14527 * config/nds32/nds32.md (pipeline_model): Add n8.
14528 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
14529 * config/nds32/pipelines.md: Include n8 settings.
14531 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
14532 Chung-Ju Wu <jasonwucj@gmail.com>
14534 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
14535 * config/nds32/nds32-n9-2r1w.md: New file.
14536 * config/nds32/nds32-n9-3r2w.md: New file.
14537 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
14538 nds32_register_ports): New or modify for cpu n9.
14539 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
14541 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
14542 * config/nds32/nds32-utils.c: New file.
14543 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
14544 TARGET_MUL_SLOW): Define.
14545 * config/nds32/nds32.md (pipeline_model): New attribute.
14546 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
14547 New options that support cpu n9.
14548 * config/nds32/pipelines.md: Include n9 settings.
14549 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
14551 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
14553 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
14554 information if necessary.
14555 (output_cond_branch_compare_zero): Likewise.
14556 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
14557 (nds32_target_alignment): Refine for alignment.
14558 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
14559 (FUNCTION_BOUNDARY): Modify.
14560 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
14562 * config/nds32/nds32.opt (malways-align, malign-functions): New.
14564 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
14566 * config/nds32/constants.md (unspec_volatile_element): Add values for
14567 TLB operation and data prefetch.
14568 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
14569 functions for TLB operation and data prefetch.
14570 * config/nds32/nds32-intrinsic.md: Likewise.
14571 * config/nds32/nds32_intrinsic.h: Likewise.
14572 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
14573 (nds32_print_operand): Likewise.
14574 * config/nds32/nds32.h (nds32_builtins): Likewise.
14576 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
14577 Andrew Pinski <pinsika@gcc.gnu.org>
14579 PR middle-end/82976
14580 * match.pd: Use constant_boolean_node of correct type instead of
14581 boolean_true_node or boolean_false_node for simplifying
14582 pointer comparisons to zero.
14584 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14586 PR tree-optimization/80021
14587 * tree.c (verify_type_variant): Make error call in verify_variant_match
14588 translatable and remove final full stop.
14590 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14592 * config/nds32/constants.md (unspec_volatile_element): Add
14593 UNSPEC_VOLATILE_EH_RETURN.
14594 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
14595 nds32_output_stack_pop): Support dwarf exception handling process.
14596 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
14597 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
14598 exception handling process.
14599 (nds32_compute_stack_frame): Likewise.
14600 (nds32_return_addr_rtx): Likewise.
14601 (nds32_initial_elimination_offset): Likewise.
14602 (nds32_expand_prologue): Likewise.
14603 (nds32_expand_epilogue): Likewise.
14604 (nds32_dynamic_chain_address): New function.
14605 * config/nds32/nds32.h (machine_function): Add fields for dwarf
14606 exception handling.
14607 (DYNAMIC_CHAIN_ADDRESS): Define.
14608 (EH_RETURN_DATA_REGNO): Define.
14609 (EH_RETURN_STACKADJ_RTX): Define.
14610 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
14611 patterns for dwarf exception handling.
14613 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14615 * config/nds32/nds32.h: Clean up obsolete macros.
14617 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14619 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14620 Add enum values for particular instructions.
14621 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14622 particular intrinsic functions.
14623 * config/nds32/nds32-intrinsic.md: Likewise.
14624 * config/nds32/nds32_intrinsic.h: Likewise.
14625 * config/nds32/nds32.h (nds32_builtins): Likewise.
14626 * config/nds32/nds32.md (type): Add pbsad and pbsada.
14627 (btst, ave): New patterns for particular instructions.
14629 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14631 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14632 Add enum values for atomic load/store and memory sync.
14633 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
14635 * config/nds32/nds32-intrinsic.md: Likewise.
14636 * config/nds32/nds32_intrinsic.h: Likewise.
14637 * config/nds32/nds32.h (nds32_builtins): Likewise.
14639 2018-04-07 Jakub Jelinek <jakub@redhat.com>
14641 PR tree-optimization/85257
14642 * fold-const.c (native_encode_vector): If not all elts could fit
14643 and off is -1, return 0 rather than offset.
14644 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
14645 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
14646 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
14647 adjust buffer in native_interpret_expr call.
14649 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14651 * config/nds32/constants.md (unspec_volatile_element): Add cache
14652 control enum values.
14653 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
14654 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
14655 * config/nds32/nds32.c (nds32_cctl_names): New.
14656 (nds32_print_operand): Handle cache control register names.
14657 * config/nds32/nds32.h (nds32_builtins): New enum values.
14658 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
14660 * config/nds32/nds32.md (type): Add mmu.
14661 * config/nds32/pipelines.md (simple_insn): Add mmu.
14663 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14665 * config/nds32/nds32.md (type): Remove call.
14666 * config/nds32/pipelines.md (simple_insn): Likewise.
14668 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14670 * config/nds32/constants.md (unspec_volatile_element): Add
14671 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
14672 UNSPEC_VOLATILE_FMFCFG.
14673 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
14674 description for fmfcfg and fmfcsr.
14675 (bdesc_1arg): Add fmtcsr.
14676 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
14677 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
14678 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
14679 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
14680 unspec_fmfcfg): New patterns.
14681 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
14682 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
14683 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
14684 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
14685 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
14686 __nds32__fmfcfg): Define.
14688 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
14690 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
14691 intrinsic register names.
14692 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
14693 intrinsic register enum values and macros.
14695 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
14697 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
14698 for load/store addressing form.
14699 (nds32_print_operand_address): Likewise.
14701 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
14704 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
14705 based on LABEL_REF. Remove useless assertion.
14706 (pic_address_needs_scratch): Fix formatting.
14707 (sparc_legitimize_pic_address): Minor tweaks.
14708 (sparc_delegitimize_address): Adjust assertion accordingly.
14709 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
14710 into symbolic_operand.
14711 (movsi_high_pic_label_ref): Likewise.
14712 (movsi_lo_sum_pic_label_ref): Likewise.
14713 (movdi_pic_label_ref): Likewise.
14714 (movdi_high_pic_label_ref): Likewise.
14715 (movdi_lo_sum_pic_label_ref): Likewise.
14717 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
14719 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
14720 custom LIB_SPEC setup.
14722 2018-04-06 Ruslan Bukin <br@bsdpad.com>
14723 Kito Cheng <kito.cheng@gmail.com>
14725 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
14726 * config/riscv/freebsd.h: New.
14728 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14730 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
14731 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
14734 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14735 Kito Cheng <kito.cheng@gmail.com>
14737 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
14738 nds32_output_call, nds32_symbol_binds_local_p): New functions.
14739 * config/nds32/nds32-protos.h (nds32_output_call,
14740 nds32_output_return): Declare.
14741 * config/nds32/nds32.md: Refine all the call and return patterns.
14743 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14746 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
14747 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
14749 PR rtl-optimization/84872
14750 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
14751 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
14752 EDGE_CROSSING edge.
14754 2018-04-06 Tamar Christina <tamar.christina@arm.com>
14756 * expr.c (copy_blkmode_to_reg): Revert 254862.
14757 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
14759 2018-04-06 Richard Biener <rguenther@suse.de>
14761 PR middle-end/85244
14762 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
14763 after seeing a component reference with an adjacent field. Treat
14764 refs to arrays at struct end of external decls similar to
14765 refs to unconstrained commons.
14767 2018-04-06 Jakub Jelinek <jakub@redhat.com>
14770 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
14771 look through SAVE_EXPRs with non-side-effects argument. Adjust
14773 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
14776 2018-04-06 Richard Biener <rguenther@suse.de>
14778 PR middle-end/85180
14779 * alias.c (find_base_term): New wrapper around find_base_term
14780 unwinding CSELIB_VAL_PTR changes.
14781 (find_base_term): Do not restore CSELIB_VAL_PTR during the
14784 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14786 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
14788 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
14789 constant definitions.
14790 ("nop"): lr 0,0 -> nopr r0
14791 ("nop_lr0", "nop_lr1"): New insn definitions.
14793 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
14795 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
14796 NDS32_V3PUSH_AVAILABLE_P macro.
14798 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
14799 Chung-Ju Wu <jasonwucj@gmail.com>
14801 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
14802 (nds32*-*-*): Add float and fpu_config into supported_defaults.
14803 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
14804 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
14805 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
14806 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
14807 * config/nds32/constraints.md: New constraints and checking for hard
14808 float configuration.
14809 * config/nds32/iterators.md: New mode iterator and attribute for hard
14810 float configuration.
14811 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
14813 * config/nds32/nds32-fpu.md: New file.
14814 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
14815 deal with hard float code generation.
14816 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
14818 (abi_type, float_reg_number): New enum type.
14819 * config/nds32/nds32-predicates.c: New predicates for hard float.
14820 * config/nds32/nds32-protos.h: Declare functions for hard float.
14821 * config/nds32/nds32.c: Implementation for hard float configuration.
14822 * config/nds32/nds32.h: Definitions for hard float configuration.
14823 * config/nds32/nds32.md: Include hard float machine description and
14824 modify patterns for hard float configuration.
14825 * config/nds32/nds32.opt: New options for hard float configuration.
14826 * config/nds32/predicates.md: New predicates for hard float
14829 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14831 * common/config/nds32/nds32-common.c
14832 (nds32_option_optimization_table): Enable -mreleax-hint by default.
14834 2018-04-05 Jakub Jelinek <jakub@redhat.com>
14836 PR middle-end/85195
14837 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
14838 CONSTRUCTOR_ELT (ctor, ...)->value.
14840 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
14843 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
14845 2018-04-05 Tom de Vries <tom@codesourcery.com>
14848 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
14851 2018-04-05 Shiva Chen <shiva0217@gmail.com>
14852 Kito Cheng <kito.cheng@gmail.com>
14854 * config/nds32/constraints.md (U33): Fine-tune checking condition.
14855 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
14856 * config/nds32/nds32.h (nds32_16bit_address_type): Add
14857 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
14859 2018-04-05 Shiva Chen <shiva0217@gmail.com>
14860 Kito Cheng <kito.cheng@gmail.com>
14862 * config/nds32/constraints.md (Ufe): New memory constraint.
14863 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
14864 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
14865 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
14867 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
14868 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
14870 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14872 * config/nds32/nds32.md: Use optimize_size in the condition for
14873 alu-shift instructions.
14875 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14877 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
14879 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14881 * config/nds32/nds32.md (negsi2): Refine pattern.
14883 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
14884 Chung-Ju Wu <jasonwucj@gmail.com>
14886 * config/nds32/iterators.md (shift_rotate): New code iterator.
14887 (shift): New code attribute.
14888 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
14889 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
14890 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
14891 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
14892 bit-wise operations.
14893 (andsi3, *andsi3): Ditto.
14894 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
14895 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
14896 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
14897 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
14898 nds32_ior_operand, nds32_xor_operand): New predicates.
14900 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14902 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
14903 (addsi3, subsi3): ... this.
14905 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14907 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
14909 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14911 * config/nds32/nds32.md: Adjust indention.
14913 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
14915 * config/nds32/nds32.md (feature): New attribute.
14917 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
14919 * config/nds32/nds32.md (subtype): New attribute.
14921 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
14924 * config/arm/arm-builtins.c (arm_expand_builtin): Change
14925 expansion to perform a bitwise AND of the argument followed by a
14926 boolean negation of the result.
14928 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
14930 PR rtl-optimization/84878
14931 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
14932 the basic block. Assert the use reference is not artificial and that
14933 it has an associated insn.
14935 2018-04-04 Michael Matz <matz@suse.de>
14937 * builtins.c (compute_objsize): Pass correct operand
14938 to array_at_struct_end_p.
14940 2018-04-04 Richard Biener <rguenther@suse.de>
14943 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
14944 from contexts for DINFO_LEVEL_TERSE and below.
14946 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14948 * config/nds32/nds32-doubleword.md (move_<mode>): Require
14949 resiter_operand condition.
14950 * config/nds32/nds32.md (*move<mode>): Ditto.
14952 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14953 Monk Chiang <sh.chiang04@gmail.com>
14955 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
14957 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14959 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
14961 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14962 Kito Cheng <kito.cheng@gmail.com>
14964 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
14965 nds32_cond_code_str, output_cond_branch,
14966 output_cond_branch_compare_zero, nds32_expand_cbranch,
14967 nds32_expand_cstore, nds32_expand_movcc,
14968 nds32_output_cbranchsi4_equality_zero,
14969 nds32_output_cbranchsi4_equality_reg,
14970 nds32_output_cbranchsi4_equality_reg_or_const_int,
14971 nds32_output_cbranchsi4_greater_less_zero: New functions.
14972 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
14973 nds32_expand_cstore, nds32_expand_movcc,
14974 nds32_output_cbranchsi4_equality_zero,
14975 nds32_output_cbranchsi4_equality_reg,
14976 nds32_output_cbranchsi4_equality_reg_or_const_int,
14977 nds32_output_cbranchsi4_greater_less_zero): Declare.
14978 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
14979 nds32_rimm11s_operand): New predicates.
14980 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
14981 * config/nds32/nds32.md: Rewrite all the branch and conditional move
14984 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
14986 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
14987 * config/nds32/nds32.md: Ditto.
14988 * config/nds32/pipelines.md: Ditto.
14990 2018-04-04 Richard Biener <rguenther@suse.de>
14992 PR tree-optimization/85168
14993 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
14994 propagating abnormals.
14996 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
14998 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15000 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15001 Kito Cheng <kito.cheng@gmail.com>
15003 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
15004 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
15005 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
15006 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
15007 * config/nds32/nds32.md (sibcall_internal): New.
15008 (sibcall_register): Remove.
15009 (sibcall_immediate): Remove.
15010 (sibcall_value_internal): New.
15011 (sibcall_value_register): Remove.
15012 (sibcall_value_immediate): Remove.
15013 * config/nds32/predicates.md (nds32_general_register_operand): New.
15014 (nds32_call_address_operand): New.
15016 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15018 PR rtl-optimization/85167
15019 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
15020 bb_defs if *split_p, instead preinitialize it to NULL.
15022 PR tree-optimization/85156
15023 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
15024 evaluating the argument multiple times.
15026 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
15028 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
15030 (_mm_cvtpd_ps): Likewise.
15031 (_mm_cvttpd_epi32): Likewise.
15032 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
15033 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
15034 vector, pixel, and bool following altivec.h include.
15036 2018-04-03 Martin Sebor <msebor@redhat.com>
15038 * doc/extend.texi (Common Function Attributes): Clarify.
15039 (const attribute): Likewise.
15040 (pure attribute): Likewise.
15042 2018-04-03 Jakub Jelinek <jakub@redhat.com>
15045 * config/i386/i386.c (ix86_expand_vector_set): Use
15046 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
15048 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
15050 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
15051 instructions when changing rounding bits to preserve precision bits
15052 in the x87 control word.
15054 2018-04-03 Martin Liska <mliska@suse.cz>
15056 PR tree-optimization/82491
15057 * rtl.h (strip_offset_and_add): Replace += suboffset with
15058 poly_uint64 () + suboffset.
15060 2018-03-29 Martin Liska <mliska@suse.cz>
15061 Martin Jambor <mjambor@suse.cz>
15064 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
15065 param_type is not an integral or pointer type.
15067 2018-04-03 Richard Biener <rguenther@suse.de>
15069 * sese.h (recompute_all_dominators): Remove.
15071 2018-04-02 Martin Sebor <msebor@redhat.com>
15073 * doc/invoke.texi (-Wrestrict): Fix typos.
15075 2018-04-02 Jim Wilson <jimw@sifive.com>
15077 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
15078 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
15079 (<optab>di3, <optab>si3_extend): Likewise.
15080 (<optab>si3_mask, <optab>si3_mask_1): New.
15081 (<optab>di3_mask, <optab>di3_mask_1): New.
15082 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
15083 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
15084 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
15086 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
15088 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
15091 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
15093 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
15094 (nds32_canonicalize_comparison): New function.
15096 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15097 Kito Cheng <kito.cheng@gmail.com>
15098 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15100 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
15101 * config/nds32/constants.md (unspec_volatile_element): Add
15102 UNSPEC_VOLATILE_RELAX_GROUP.
15103 * config/nds32/nds32-relax-opt.c: New file.
15104 * config/nds32/nds32-predicates.c
15105 (nds32_symbol_load_store_p): New function.
15106 * config/nds32/nds32-protos.h
15107 (nds32_symbol_load_store_p): Declare function.
15108 (make_pass_nds32_relax_opt): Declare new rtl pass function.
15109 * config/nds32/nds32.c
15110 (nds32_register_pass): New function to register pass.
15111 (nds32_register_passes): New function to register passes.
15112 * config/nds32/nds32.md (relax_group): New pattern.
15113 * config/nds32/nds32.opt (mrelax-hint): New option.
15114 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
15116 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
15118 * config/nds32/t-nds32: Modify files dependency.
15120 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15122 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
15123 (PROFILE_HOOK): Define its implementation.
15125 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
15127 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
15128 type and 32-bit size.
15130 2018-04-01 Jakub Jelinek <jakub@redhat.com>
15132 PR middle-end/85090
15133 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
15134 (V_128_256): New mode iterator.
15135 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
15136 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
15137 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
15139 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
15140 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
15142 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
15145 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
15146 NaN inputs correctly.
15148 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
15151 * config/rs6000/vsx.md (??r): New mode attribute.
15152 (*vsx_mov<mode>_64bit): Use it.
15153 (*vsx_mov<mode>_32bit): Likewise.
15155 2018-03-30 Martin Sebor <msebor@redhat.com>
15157 PR tree-optimization/84818
15158 * builtins.c (check_access): Use warning_n.
15160 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15163 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
15165 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
15168 2018-03-30 Julia Koval <julia.koval@intel.com>
15171 * x86-tune.def (movx, partial_reg_dependency): Enable for
15174 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
15176 PR inline-asm/84985
15177 * lra-constraints.c (process_alt_operands): Move setting
15178 this_alternative_matches below.
15180 2018-03-29 Martin Liska <mliska@suse.cz>
15183 * doc/invoke.texi: Document how LTO works with debug info.
15184 Describe auto-load support of binutils. Mention 'x86-64'
15185 as valid option value of -march option.
15187 2018-03-29 Jakub Jelinek <jakub@redhat.com>
15189 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
15192 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
15193 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
15194 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
15197 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15200 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
15201 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
15202 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
15203 for RS6000_BTM_POWERPC64.
15204 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
15205 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
15206 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
15209 (DIVDEU): Likewise.
15211 2018-03-28 Carl Love <cel@us.ibm.com>
15214 2017-09-27 Carl Love <cel@us.ibm.com>
15216 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
15217 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
15218 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
15221 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15223 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
15224 instead of __vector bool.
15225 (_mm_max_pu8): Likewise.
15226 (_mm_min_pi16): Likewise.
15228 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
15231 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
15232 (DIVWEUO): Likewise.
15233 (DIVDEO): Likewise.
15234 (DIVDEUO): Likewise.
15235 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
15236 DIVWEUO and DIVDEUO.
15237 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
15238 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
15239 (div_extend): Likewise.
15240 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
15242 (__builtin_divweuo): Likewise.
15243 (__builtin_divdeo): Likewise.
15244 (__builtin_divdeuo): Likewise.
15246 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15249 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
15250 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
15252 PR tree-optimization/82004
15253 * gimple-match-head.c (optimize_pow_to_exp): New function.
15254 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
15255 Don't fold to exp if optimize_pow_to_exp is false.
15257 2018-03-28 Martin Liska <mliska@suse.cz>
15260 * calls.c (initialize_argument_information): Fix trailing space.
15261 * common.opt: Fix typo and provide better explanation for
15262 -fsanitize-coverage option.
15263 * config/i386/i386.opt: Fix typo.
15265 2018-03-28 Jakub Jelinek <jakub@redhat.com>
15266 Martin Liska <mliska@suse.cz>
15269 * gimplify.c (asan_poison_variable): Don't do the check for
15270 gimplify_omp_ctxp here.
15271 (gimplify_decl_expr): Do it here.
15272 (gimplify_target_expr): Likewise.
15274 2018-03-28 Martin Liska <mliska@suse.cz>
15277 * config/i386/i386.c (ix86_function_arg_advance): Do not call
15278 chkp_type_bounds_count if MPX is not enabled.
15280 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
15282 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
15284 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
15287 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
15288 function to create the function decl for complex long double
15289 multiply and divide for -mabi=ieeelongdouble.
15290 (init_float128_ieee): Call it.
15292 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
15295 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
15296 -fcf-protection=branch -mibt.
15297 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
15299 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15302 * config/arm/arm.c (arm_valid_symbolic_address): Handle
15303 arm_word_relocations.
15305 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
15308 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
15309 extern array declarations.
15311 2018-03-27 Richard Biener <rguenther@suse.de>
15313 PR middle-end/84067
15314 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
15315 explicit single_use checks.
15317 2018-03-27 Richard Biener <rguenther@suse.de>
15319 PR tree-optimization/85082
15320 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
15323 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
15325 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
15326 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
15327 Turn on fasynchronous-unwind-tables and funwind-tables.
15329 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
15332 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
15333 (*bmi_blsr_<mode>_ccz): Ditto.
15335 2018-03-26 Tom de Vries <tom@codesourcery.com>
15337 PR tree-optimization/85063
15338 * omp-general.c (offloading_function_p): New function. Factor out
15340 * omp-offload.c (pass_omp_target_link::gate): ... here.
15341 * omp-general.h (offloading_function_p): Declare.
15342 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
15343 with attribute omp declare target for offloading functions.
15345 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
15347 PR tree-optimization/84005
15348 * tree-data-ref.h (get_base_for_alignment): Declare.
15349 * tree-data-ref.c (get_base_for_alignment_1): New function.
15350 (get_base_for_alignment): Likewise.
15351 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
15352 get_base_for_alignment to find a suitable base object, instead
15353 of always using drb->base_address.
15355 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15357 PR inline-asm/85022
15358 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
15359 known size by default.
15361 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
15363 PR inline-asm/85030
15364 * lra-constraints.c (process_alt_operands): Don't match BLKmode
15365 and non BLKmode operands.
15367 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15370 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
15371 Clean up attributes.
15373 2018-03-23 Richard Biener <rguenther@suse.de>
15376 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
15377 we are going to emit early debug for LTO.
15379 2018-03-23 Jakub Jelinek <jakub@redhat.com>
15381 PR inline-asm/85034
15382 * function.c (match_asm_constraints_1): Don't optimize if input
15383 doesn't satisfy general_operand predicate for output's mode.
15385 PR inline-asm/85022
15386 * alias.c (write_dependence_p): Don't require for x_canonicalized
15387 non-VOIDmode if x has VOIDmode.
15390 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
15391 just don't try to optimize it rather than assert it never happens.
15393 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15395 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
15396 macro expansions for definition of ST_INTERNAL_<mode> and
15397 LD_INTERNAL_<mode> builtins.
15398 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
15400 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
15402 (altivec_expand_st_builtin): Likewise.
15403 (altivec_expand_builtin): Remove calls to deleted functions.
15404 (rs6000_address_for_altivec): Delete this function.
15405 * config/rs6000/vector.md: Remove expands for
15406 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
15408 2018-03-22 Sudakshina Das <sudi.das@arm.com>
15411 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
15412 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
15413 re-computing once computed.
15414 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
15415 (arm_init_machine_status): Initialize
15416 machine->static_chain_stack_bytes.
15418 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15421 * doc/extend.texi: Add four new prototypes for vec_ld.
15422 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
15423 definitions for more logical presentation.
15424 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
15425 entries for V1TI variants of __builtin_altivec_ld builtin.
15426 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
15427 handling of V1TI variant of LVX icode pattern.
15428 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
15429 (rs6000_gimple_fold_builtin): Likewise.
15430 (altivec_init_builtins): Add code to define
15431 __builtin_altivec_lvx_v1ti function.
15433 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15435 PR inline-asm/84941
15436 * function.c (match_asm_constraints_1): Don't do the optimization
15437 if input isn't a REG, SUBREG, MEM or constant.
15439 2018-03-22 Tom de Vries <tom@codesourcery.com>
15441 PR tree-optimization/84956
15442 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
15443 bb_has_abnormal_pred.
15445 2018-03-22 Jakub Jelinek <jakub@redhat.com>
15448 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
15449 DECL_INITIAL (decl) to decl at the end.
15450 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
15451 adjust the comment.
15453 2018-03-21 Joseph Myers <joseph@codesourcery.com>
15455 * doc/extend.texi (__builtin_tgmath): Document when complex
15456 integer types are treated as _Complex _Float64.
15458 2018-03-21 Tom de Vries <tom@codesourcery.com>
15460 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
15462 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15464 PR tree-optimization/84960
15465 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
15466 if it is ENTRY block, move them into single succ of ENTRY in that case.
15468 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
15470 PR tree-optimization/84811
15471 * poly-int.h (poly_span_traits): Remove the T3 parameter and
15472 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
15473 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
15474 (known_subrange_p): Update accordingly. Cast each value involved
15475 in the size comparison, rather than casting the result of the
15478 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15480 PR tree-optimization/84982
15481 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
15482 by flipping the least significant bit rather than all bits from
15483 bitpos to bitpos + bitsize - 1.
15485 2018-03-21 Nathan Sidwell <nathan@acm.org>
15487 * doc/extend.texi (Deprecated Features): Remove mention of
15488 long-deleted deprecations.
15490 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15493 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
15494 * configure: Regenerate.
15496 2018-03-21 Tom de Vries <tom@codesourcery.com>
15498 PR tree-optimization/83126
15499 * tree-parloops.c (num_phis): New function.
15500 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
15502 2018-03-21 Nathan Sidwell <nathan@acm.org>
15504 * doc/extend.texi (Deprecated Features): Update deprecated flags,
15505 mention anon-struct/union members and trailing attributes.
15507 2018-03-21 Bin Cheng <bin.cheng@arm.com>
15509 PR tree-optimization/84969
15510 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
15511 builtin memset partitions if they set different rhs values.
15513 2018-03-21 Jakub Jelinek <jakub@redhat.com>
15515 PR rtl-optimization/84989
15516 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
15517 VEC_DUPLICATE with scalar result mode.
15519 2018-03-21 Martin Liska <mliska@suse.cz>
15522 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
15523 not intended return statement.
15525 2018-03-21 Martin Liska <mliska@suse.cz>
15528 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
15529 (chkp_find_bound_slots_1): Limit number of iterations.
15531 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
15534 * Minor grammar fixes for x86 options.
15536 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15539 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
15540 holding REG_CFA_RESTORE notes, instead turn them into a USE.
15542 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
15545 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
15546 (altivec_lvx_<mode>_1op): Likewise.
15547 (altivec_stvx_<mode>_2op): Likewise.
15548 (altivec_stvx_<mode>_1op): Likewise.
15549 (altivec_lvx_<VM2:mode>): New define_expand.
15550 (altivec_stvx_<VM2:mode>): Likewise.
15551 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
15552 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15553 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
15554 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
15555 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
15556 (rs6000_gen_lvx): Likewise.
15557 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
15558 (altivec_expand_stv_builtin): Likewise.
15559 (altivec_expand_builtin): Likewise.
15560 * config/rs6000/vector.md: Likewise.
15562 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15565 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
15568 2018-03-20 Richard Biener <rguenther@suse.de>
15571 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
15572 sign-conversions as zero, fall back to standard scalar_stmt
15575 2018-03-20 Martin Liska <mliska@suse.cz>
15578 * predict.c (rebuild_frequencies): Handle case when we have
15579 PROFILE_ABSENT, but flag_guess_branch_prob is false.
15581 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15584 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
15585 flag_section_anchors.
15586 * varasm.c (use_blocks_for_decl_p): Remove hack for
15587 dw2_force_const_mem.
15590 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
15592 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
15593 be created, use lowpart_subreg of operands[0] rather than operands[0]
15595 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
15596 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
15597 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
15598 and n constraint instead of aarch64_shift_imm_di and Usd.
15599 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
15600 (*aarch64_<optab>_reg_minus<mode>3): ... this.
15602 2018-03-20 Sudakshina Das <sudi.das@arm.com>
15605 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
15606 to favor GPR over NEON registers.
15607 (<shift>di3_neon): Likewise.
15609 2018-03-20 Tom de Vries <tom@codesourcery.com>
15612 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
15613 (nvptx_process_pars): Emit bar.sync asap and alap.
15615 2018-03-20 Tom de Vries <tom@codesourcery.com>
15618 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
15619 seen_label if seen_label is already set.
15621 2018-03-20 Jakub Jelinek <jakub@redhat.com>
15624 * config/i386/i386.c (fold_builtin_cpu): For features above 31
15625 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
15626 Use 1U instead of 1. Formatting fixes.
15629 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
15630 instead of TREE_TYPE (s1) for the return value.
15632 2018-03-19 Jakub Jelinek <jakub@redhat.com>
15634 PR tree-optimization/84946
15635 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
15636 bitsize + bitsize in poly_uint64 rather than poly_int64.
15639 * dwarf2asm.c: Include fold-const.c.
15640 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
15641 of decl rather than decl itself.
15643 PR rtl-optimization/84643
15644 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
15646 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
15649 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
15650 calling assemble_variable.
15652 2018-03-19 Sudakshina Das <sudi.das@arm.com>
15655 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
15656 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
15658 2018-03-19 Jim Wilson <jimw@sifive.com>
15661 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
15662 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
15663 (riscv_first_stack_step): Likewise.
15664 (riscv_option_override): Use STACK_BOUNDARY instead of
15665 MIN_STACK_BOUNDARY.
15666 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
15667 MIN_STACK_BOUNDARY.
15668 (BIGGEST_ALIGNMENT): Set to 128.
15669 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
15670 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
15673 2018-03-19 Richard Biener <rguenther@suse.de>
15675 PR tree-optimization/84933
15676 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
15677 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
15679 2018-03-19 Richard Biener <rguenther@suse.de>
15681 PR tree-optimization/84859
15682 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
15683 (cond_if_else_store_replacement): Perform sinking operation on
15684 single-store BBs regardless of MAX_STORES_TO_SINK setting.
15685 Generalize what a BB with a single eligible store is.
15687 2018-03-19 Richard Biener <rguenther@suse.de>
15689 PR tree-optimization/84929
15690 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
15691 chrec_is_positive against non-chrec arg.
15693 2018-03-19 Tamar Christina <tamar.christina@arm.com>
15696 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
15698 2018-03-18 Martin Liska <mliska@suse.cz>
15700 PR rtl-optimization/84635
15701 * regrename.c (build_def_use): Use matches_mode only when
15704 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
15706 PR tree-optimization/84913
15707 * tree-vect-loop.c (vectorizable_reduction): Don't try to
15708 vectorize chains of COND_EXPRs.
15710 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15712 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
15714 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15716 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
15718 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
15720 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
15722 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15723 Kito Cheng <kito.cheng@gmail.com>
15725 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
15726 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
15727 (nds32_adjust_reg_alloc_order): New function.
15728 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
15730 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
15732 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
15733 nds32_print_operand, nds32_print_operand_address): Use
15734 HOST_WIDE_INT_PRINT_DEC instead.
15736 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
15738 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
15740 2018-03-17 Jakub Jelinek <jakub@redhat.com>
15743 * config/i386/i386.c (initial_ix86_tune_features,
15744 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
15745 unsigned long long.
15746 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
15747 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
15748 rather than 1u << ix86_tune. Formatting fix.
15749 (ix86_option_override_internal): Change ix86_arch_mask from
15750 unsigned int to unsigned HOST_WIDE_INT, initialize to
15751 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
15752 (ix86_function_specific_restore): Likewise.
15754 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15757 * postreload.c (reload_combine_recognize_pattern): Perform
15758 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
15759 truncate_int_for_mode the result for the destination's mode.
15762 * hsa-gen.c (mem_type_for_type): Fix comment typo.
15763 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
15765 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15768 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
15771 * lra-assigns.c (lra_split_hard_reg_for): Don't use
15772 regno_allocno_class_array and sorted_pseudos.
15773 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
15774 insns where regno is used.
15776 2018-03-16 Martin Liska <mliska@suse.cz>
15779 * multiple_target.c (create_dispatcher_calls): Redirect
15780 reference in the symbol table.
15782 2018-03-16 Martin Liska <mliska@suse.cz>
15785 * multiple_target.c (create_dispatcher_calls): Redirect also
15788 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15792 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
15793 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
15794 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
15796 2018-03-16 Julia Koval <julia.koval@intel.com>
15798 * doc/invoke.texi (Skylake Server): Add CLWB.
15799 Cannonlake): Remove CLWB.
15801 2018-03-16 Jakub Jelinek <jakub@redhat.com>
15803 PR tree-optimization/84841
15804 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
15806 (FLOAT_ONE_CONST_TYPE): Define.
15807 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
15808 (sort_by_operand_rank): Put entries with higher constant_type last
15809 rather than first to match comments.
15811 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
15813 * config/nios2/nios2.md (movsi_internal): Fix thinko in
15816 2018-03-15 Jakub Jelinek <jakub@redhat.com>
15819 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
15820 check and use address of target always.
15822 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
15825 * config/i386/i386.c (indirect_thunk_needed): Update comments.
15826 (indirect_thunk_bnd_needed): Likewise.
15827 (indirect_thunks_used): Likewise.
15828 (indirect_thunks_bnd_used): Likewise.
15829 (indirect_return_needed): New.
15830 (indirect_return_bnd_needed): Likewise.
15831 (output_indirect_thunk_function): Add a bool argument for
15833 (output_indirect_thunk_function): Don't generate alias for
15834 function return thunk.
15835 (ix86_code_end): Call output_indirect_thunk_function to generate
15836 function return thunks.
15837 (ix86_output_function_return): Set indirect_return_bnd_needed
15838 and indirect_return_needed instead of indirect_thunk_bnd_needed
15839 and indirect_thunk_needed.
15841 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
15843 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
15844 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
15845 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
15847 2018-03-15 David Malcolm <dmalcolm@redhat.com>
15848 Paul Hua <paul.hua.gm@gmail.com>
15851 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
15853 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
15855 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
15856 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
15857 resp. SFmode cases.
15859 2018-03-15 Tamar Christina <tamar.christina@arm.com>
15862 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
15863 instead of GET_MODE_SIZE when comparing Units.
15865 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
15868 * varasm.c (hash_section): Return an unchangeble hash value
15869 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
15870 Return !aarch64_can_use_per_function_literal_pools_p ().
15872 2018-03-15 Jakub Jelinek <jakub@redhat.com>
15875 * optabs.c (emit_conditional_move): Pass address of cmode's copy
15876 rather than address of cmode as last argument to prepare_cmp_insn.
15878 2018-03-15 Julia Koval <julia.koval@intel.com>
15880 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
15881 F_AVX512VNNI, F_AVX512BITALG): New.
15883 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
15886 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
15887 insn for floating-point loads and stores.
15889 2018-03-14 Carl Love <cel@us.ibm.com>
15891 * config/rs6000/rs6000-c.c: Add macro definitions for
15892 ALTIVEC_BUILTIN_VEC_PERMXOR.
15893 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
15894 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
15895 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
15896 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
15898 * config/doc/extend.texi: Add prototypes for vec_permxor.
15900 2018-03-14 David Malcolm <dmalcolm@redhat.com>
15903 * diagnostic-show-locus.c (class layout_point): Convert m_line
15904 from int to linenum_type.
15905 (line_span::comparator): Use linenum "compare" function when
15906 comparing line numbers.
15907 (test_line_span): New function.
15908 (layout_range::contains_point): Convert param "row" from int to
15910 (layout_range::intersects_line_p): Likewise.
15911 (layout::will_show_line_p): Likewise.
15912 (layout::print_source_line): Likewise.
15913 (layout::should_print_annotation_line_p): Likewise.
15914 (layout::print_annotation_line): Likewise.
15915 (layout::print_leading_fixits): Likewise.
15916 (layout::annotation_line_showed_range_p): Likewise.
15917 (struct line_corrections): Likewise for field m_row.
15918 (line_corrections::line_corrections): Likewise for param "row".
15919 (layout::print_trailing_fixits): Likewise.
15920 (layout::get_state_at_point): Likewise.
15921 (layout::get_x_bound_for_row): Likewise.
15922 (layout::print_line): Likewise.
15923 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
15924 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
15925 * input.c (selftest::test_linenum_comparisons): New function.
15926 (selftest::input_c_tests): Call it.
15927 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
15928 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
15929 * selftest.h (ASSERT_GT): New macro.
15930 (ASSERT_GT_AT): New macro.
15931 (ASSERT_LT): New macro.
15932 (ASSERT_LT_AT): New macro.
15934 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
15936 PR rtl-optimization/84780
15937 * combine.c (distribute_links): Don't make a link based on pc_rtx.
15939 2018-03-14 Martin Liska <mliska@suse.cz>
15941 * tree.c (record_node_allocation_statistics): Use
15942 get_stats_node_kind.
15943 (get_stats_node_kind): New function extracted from
15944 record_node_allocation_statistics.
15945 (free_node): Use get_stats_node_kind.
15947 2018-03-14 Richard Biener <rguenther@suse.de>
15949 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
15950 that the value-set of ANTIC_IN doesn't grow.
15953 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15955 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15956 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15958 2018-03-14 Julia Koval <julia.koval@intel.com>
15960 * config.gcc (icelake-client, icelake-server): New.
15962 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
15963 (initial_ix86_arch_features): Ditto.
15964 (PTA_SKYLAKE): Add SGX.
15965 (PTA_ICELAKE): Remove.
15966 (PTA_ICELAKE_CLIENT): New.
15967 (PTA_ICELAKE_SERVER): New.
15968 (ix86_option_override_internal): Split up icelake on icelake client and
15970 (get_builtin_code_for_version): Ditto.
15971 (fold_builtin_cpu): Ditto.
15972 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
15973 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
15974 * config/i386/i386.h (processor_type): Ditto.
15975 * doc/invoke.texi: Ditto.
15977 2018-03-14 Jakub Jelinek <jakub@redhat.com>
15980 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
15981 INTEGER_CST offset, add it together with bitpos / 8 and
15982 sign extend based on POINTER_SIZE.
15986 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
15989 * config/i386/constraints.md (Yc): New register constraint.
15990 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
15991 Use Yc constraint for alternative 2 of operand 0. Remove
15992 preferred_for_speed attribute.
15994 2018-03-14 Richard Biener <rguenther@suse.de>
15996 PR tree-optimization/84830
15997 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
15998 with the old one to avoid oscillations.
16000 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
16003 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
16005 (assign_by_spills): Return a flag of reload assignment failure.
16006 Do not process the reload assignment failures. Do not spill other
16007 reload pseudos if they has the same reg class. Update n if
16009 (lra_assign): Add a return arg. Set up from the result of
16010 assign_by_spills call.
16011 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16012 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16013 usage_insns if it is not NULL.
16014 (spill_hard_reg_in_range): New function.
16015 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16016 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16017 function prototypes.
16018 (lra_assign): Change prototype.
16019 * lra.c (lra): Add code to deal with fails by splitting hard reg
16022 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
16024 * config/riscv/riscv.opt (mrelax): New option.
16025 * config/riscv/riscv.c (riscv_file_start): Emit ".option
16026 "norelax" when riscv_mrelax is disabled.
16027 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
16029 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16032 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
16033 reassociation for int modes.
16035 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16037 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
16038 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
16040 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
16041 * config/aarch64/aarch64-sve.md
16042 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
16043 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
16044 (*extend<mode><Vwide>2): Rename to...
16045 (aarch64_sve_extend<mode><Vwide>2): ...this.
16046 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
16047 renaming the old pattern to...
16048 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
16050 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
16051 define_expand, renaming the old pattern to...
16052 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
16053 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
16054 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
16055 account when deciding which SVE instruction the optab should use.
16056 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
16058 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16060 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
16061 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
16062 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
16063 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
16064 (tlsdesc_small_<mode>): Turn a define_expand and use
16065 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
16066 (tlsdesc_small_advsimd_<mode>): ...this.
16067 (tlsdesc_small_sve_<mode>): New pattern.
16069 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
16071 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
16072 (UNSPEC_UMUL_HIGHPART): New constants.
16073 (MUL_HIGHPART): New int iteraor.
16074 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
16075 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
16077 (*<su>mul<mode>3_highpart): New define_insn.
16079 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16082 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16085 2018-03-13 Martin Liska <mliska@suse.cz>
16088 * (sem_item_optimizer::sem_item_optimizer): Initialize new
16090 (sem_item_optimizer::~sem_item_optimizer): Release it.
16091 (sem_item_optimizer::merge_classes): Register variable aliases.
16092 (sem_item_optimizer::fixup_pt_set): New function.
16093 (sem_item_optimizer::fixup_points_to_sets): Likewise.
16094 * ipa-icf.h: Declare new variables and functions.
16096 2018-03-13 Jakub Jelinek <jakub@redhat.com>
16098 PR middle-end/84834
16099 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
16100 integer_pow2p@2 and test integer_pow2p in condition.
16101 (A < 0 ? C : 0): Similarly for @1.
16103 PR middle-end/84831
16104 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
16105 characters starting at p contain '\0' character, don't look beyond
16109 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
16110 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
16113 * reg-stack.c (change_stack): Change update_end var from int to
16114 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
16115 also call set_block_for_insn on the newly added insns and rescan.
16118 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
16119 on the last operand.
16122 * tree.c (stabilize_reference_1): Return save_expr (e) for
16123 STATEMENT_LIST even if it doesn't have side-effects.
16125 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
16127 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
16129 2018-03-12 Renlin Li <renlin.li@arm.com>
16131 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
16132 aarch64_output_scalar_simd_mov_immediate.
16134 2018-03-12 Martin Sebor <msebor@redhat.com>
16136 PR tree-optimization/83456
16137 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
16138 for perfectly overlapping calls to memcpy.
16139 (gimple_fold_builtin_memory_chk): Same.
16140 (gimple_fold_builtin_strcpy): Handle no-warning.
16141 (gimple_fold_builtin_stxcpy_chk): Same.
16142 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
16144 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16146 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
16147 parameter. Use it for SFmode.
16148 (rs6000_function_arg_advance_1): Adjust.
16149 (rs6000_function_arg): Adjust.
16150 (rs6000_gimplify_va_arg): Pass false for that new parameter.
16152 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
16154 PR rtl-optimization/84169
16155 PR rtl-optimization/84780
16156 * combine.c (can_combine_p): Check for a 2-insn combination whether
16157 the destination register is used between the two insns, too.
16159 2018-03-12 Richard Biener <rguenther@suse.de>
16161 PR tree-optimization/84803
16162 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
16163 for refs DR analysis didn't process.
16165 2018-03-12 Richard Biener <rguenther@suse.de>
16167 PR tree-optimization/84777
16168 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
16169 force-vectorize loops ignore whether we are optimizing for size.
16171 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
16173 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
16174 (TARGET_MD_ASM_ADJUST): Define.
16176 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
16177 Kito Cheng <kito.cheng@gmail.com>
16178 Chung-Ju Wu <jasonwucj@gmail.com>
16180 * config/nds32/nds32.c (nds32_compute_stack_frame,
16181 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
16182 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
16183 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
16184 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
16185 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
16186 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
16187 * config/nds32/nds32.md (prologue, epilogue): Use macro
16188 NDS32_V3PUSH_AVAILABLE_P to do checking.
16190 2018-03-11 Jakub Jelinek <jakub@redhat.com>
16193 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
16194 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
16195 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
16196 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
16197 addition of most attributes on !orig_type_die or the attribute not
16198 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
16200 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16201 Chung-Ju Wu <jasonwucj@gmail.com>
16203 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
16204 __NDS32_VH__ macro.
16205 * config/nds32/nds32.opt (mvh): New option.
16207 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16208 Chung-Ju Wu <jasonwucj@gmail.com>
16210 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
16212 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
16213 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
16216 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
16217 Chung-Ju Wu <jasonwucj@gmail.com>
16219 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
16221 * config/nds32/nds32-multiple.md (strlensi): New pattern.
16222 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
16224 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
16225 Kito Cheng <kito.cheng@gmail.com>
16226 Chung-Ju Wu <jasonwucj@gmail.com>
16228 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
16229 UNSPEC_FFMISM and UNSPEC_FLMISM.
16230 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
16231 for ffb, ffmism and flmism.
16232 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
16233 (unspec_ffmism): Ditto.
16234 (unspec_flmism): Ditto.
16235 (nds32_expand_builtin_impl): Check if string extension is available.
16236 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
16237 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
16239 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
16242 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16245 * lra-assigns.c (assign_by_spills): Return a flag of reload
16246 assignment failure. Do not process the reload assignment
16247 failures. Do not spill other reload pseudos if they has the same
16249 (lra_assign): Add a return arg. Set up from the result of
16250 assign_by_spills call.
16251 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16252 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16253 usage_insns if it is not NULL.
16254 (spill_hard_reg_in_range): New function.
16255 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16256 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16257 function prototypes.
16258 (lra_assign): Change prototype.
16259 * lra.c (lra): Add code to deal with fails by splitting hard reg
16262 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
16265 * config/i386/i386.opt: Replace Enforcment with Enforcement.
16267 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
16270 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
16271 (dw_val_node): Add val_symbolic_view.
16272 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
16273 (symview_upper_bound): New.
16274 (new_line_info_table): Initialize symviews_since_reset.
16275 (dwarf2out_source_line): Count symviews_since_reset and set
16276 symview_upper_bound.
16277 (dw_val_equal_p): Handle symview.
16278 (add_AT_symview): New.
16279 (print_dw_val): Handle symview.
16280 (attr_checksum, attr_checksum_ordered): Likewise.
16281 (same_dw_val_p, size_of_die): Likewise.
16282 (value_format, output_die): Likewise.
16283 (add_high_low_attributes): Use add_AT_symview for entry_view.
16284 (dwarf2out_finish): Reset symview_upper_bound, clear
16287 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
16290 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
16291 Add strict argument and use it.
16292 (rs6000_split_multireg_move): Update for new strict argument.
16293 (mem_operand_gpr): Disallow all non-offsettable addresses.
16294 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
16296 2018-03-09 Jakub Jelinek <jakub@redhat.com>
16299 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
16300 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
16301 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
16304 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
16305 decl, use remap_type if we want to use the type.
16307 2018-03-09 Martin Sebor <msebor@redhat.com>
16309 PR tree-optimization/84526
16310 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16312 (builtin_access::generic_overlap): Be prepared to handle non-array
16315 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
16317 PR rtl-optimization/84682
16318 * lra-constraints.c (process_address_1): Check is_address flag
16319 for address constraints.
16320 (process_alt_operands): Likewise.
16321 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
16322 preprocess_constraints.
16323 * recog.h (preprocess_constraints): Add oploc parameter.
16325 * recog.c (preprocess_constraints): Test address_operand for
16326 CT_ADDRESS constraints.
16328 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
16331 * lra-assigns.c (assign_by_spills): Return a flag of reload
16332 assignment failure. Do not process the reload assignment
16333 failures. Do not spill other reload pseudos if they has the same
16335 (lra_assign): Add a return arg. Set up from the result of
16336 assign_by_spills call.
16337 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
16338 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
16339 usage_insns if it is not NULL.
16340 (spill_hard_reg_in_range): New function.
16341 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
16342 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
16343 function prototypes.
16344 (lra_assign): Change prototype.
16345 * lra.c (lra): Add code to deal with fails by splitting hard reg
16348 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16351 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
16352 Accept complain bool parameter. Only emit errors if it is true.
16353 (arm_parse_cpu_option_name): Likewise.
16354 (arm_target_thumb_only): Adjust callers of the above.
16355 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
16356 prototype to take a default true bool parameter.
16357 (arm_parse_arch_option_name): Likewise.
16359 2018-03-09 David Malcolm <dmalcolm@redhat.com>
16360 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
16364 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
16365 * configure: Regenerate.
16366 * configure.ac ("linker --version-script option"): New.
16367 ("linker soname option"): New.
16369 2018-03-09 Richard Biener <rguenther@suse.de>
16371 PR tree-optimization/84775
16372 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
16373 immediate uses of predicate stmts and mark them modified.
16376 PR tree-optimization/84178
16377 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16379 (version_loop_for_if_conversion): Delay update_ssa call.
16380 (tree_if_conversion): Delay update_ssa until after predicate
16383 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
16386 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
16387 when the function accesses prior frames.
16389 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16392 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
16393 gen_llsym, otherwise call maybe_gen_llsym.
16395 PR inline-asm/84742
16396 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
16397 has ',' character inside of it.
16399 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16402 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
16403 as clobbering CC_REGNUM.
16405 2018-03-08 Richard Biener <rguenther@suse.de>
16407 PR middle-end/84552
16408 * tree-scalar-evolution.c: Include tree-into-ssa.h.
16409 (follow_copies_to_constant): Do not follow SSA names registered
16412 2018-03-08 Richard Biener <rguenther@suse.de>
16414 PR tree-optimization/84178
16415 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
16417 (version_loop_for_if_conversion): Delay update_ssa call.
16418 (tree_if_conversion): Delay update_ssa until after predicate
16421 2018-03-08 David Malcolm <dmalcolm@redhat.com>
16423 PR tree-optimization/84178
16424 * tree-if-conv.c (release_bb_predicate): Remove the
16425 the assertion that the stmts have NULL use_ops.
16426 Discard the statements, asserting that they haven't
16427 yet been added to a BB.
16429 2018-03-08 Richard Biener <rguenther@suse.de>
16431 PR tree-optimization/84746
16432 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
16433 (phi_translate): Pass in destination ANTIC_OUT set.
16434 (phi_translate_1): Likewise. For a simplified result lookup
16435 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
16436 (phi_translate_set): Adjust.
16437 (do_pre_regular_insertion): Likewise.
16438 (do_pre_partial_partial_insertion): Likewise.
16440 2018-03-08 Martin Liska <mliska@suse.cz>
16442 PR gcov-profile/84735
16443 * doc/gcov.texi: Document usage of profile files.
16444 * gcov-io.h: Document changes in the format.
16446 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
16450 * dwarf2out.c (struct dw_line_info_table): Update comments for
16452 (FORCE_RESET_NEXT_VIEW): New.
16453 (FORCE_RESETTING_VIEW_P): New.
16454 (RESETTING_VIEW_P): Check for -1 too.
16455 (ZERO_VIEW_P): Likewise.
16456 (new_line_info_table): Force-reset next view.
16457 (dwarf2out_begin_function): Likewise.
16458 (dwarf2out_source_line): Simplify zero_view_p initialization.
16459 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
16460 view directly. Omit view when omitting .loc at line 0.
16462 2018-03-08 Jakub Jelinek <jakub@redhat.com>
16464 PR tree-optimization/84740
16465 * tree-switch-conversion.c (process_switch): Call build_constructors
16466 only if info.phi_count is non-zero.
16468 PR tree-optimization/84739
16469 * tree-tailcall.c (find_tail_calls): Check call arguments against
16470 DECL_ARGUMENTS (current_function_decl) rather than
16471 DECL_ARGUMENTS (func) when checking for tail recursion.
16473 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16475 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
16476 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16477 Volker Reichelt's entry and add entries for people that perform
16478 GCC fuzzy testing and report numerous bugs.
16480 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
16483 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
16484 readonly data in sdata, if that is disabled.
16485 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
16486 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
16487 -mreadonly-in-sdata option.
16489 2018-03-07 Martin Sebor <msebor@redhat.com>
16491 PR tree-optimization/84468
16492 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
16493 basic block when looking for nul assignment.
16495 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
16498 * except.h (output_function_exception_table): Adjust prototype.
16499 * except.c (output_function_exception_table): Remove FNNAME parameter
16500 and add SECTION parameter. Ouput one part of the table at a time.
16501 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
16502 the first part of the exception table and emit unwind directives.
16503 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
16504 (i386_pe_seh_cold_init): Likewise.
16505 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
16506 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
16507 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
16508 (ix86_output_call_insn): Emit a nop in one more case for SEH.
16509 * config/i386/winnt.c: Include except.h.
16510 (struct seh_frame_state): Add reg_offset, after_prologue and
16511 in_cold_section fields.
16512 (i386_pe_seh_end_prologue): Set seh->after_prologue.
16513 (i386_pe_seh_cold_init): New function.
16514 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
16515 to seh->in_cold_section.
16516 (seh_emit_push): Record the offset of the push.
16517 (seh_emit_save): Record the offet of the save.
16518 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
16519 Test seh->after_prologue to disregard the epilogue.
16520 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
16521 (i386_pe_end_cold_function): New function.
16523 2018-03-07 Jakub Jelinek <jakub@redhat.com>
16526 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
16527 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
16530 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
16532 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
16533 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
16535 PR middle-end/84723
16536 * multiple_target.c: Include tree-inline.h and intl.h.
16537 (expand_target_clones): Diagnose and fail if node->definition and
16538 !tree_versionable_function_p (node->decl).
16540 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16542 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
16544 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
16545 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
16546 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
16548 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16551 * combine.c (try_combine): Use reg_or_subregno instead of handling
16552 just paradoxical SUBREGs and REGs.
16554 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
16556 * config/arc/arc.c (arc_finalize_pic): Remove function.
16557 (arc_must_save_register): We use single base PIC register, remove
16558 checks to save/restore the PIC register.
16559 (arc_expand_prologue): Likewise.
16560 * config/arc/arc-protos.h (arc_set_default_type_attributes):
16562 (arc_verify_short): Likewise.
16563 (arc_attr_type): Likewise.
16564 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
16565 (walk_stores): Likewise.
16566 (arc_address_cost): Make it static.
16567 (arc_verify_short): Likewise.
16568 (branch_dest): Likewise.
16569 (arc_attr_type): Likewise.
16570 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
16571 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
16572 (arc_final_prescan_insn): Remove inserting the nops due to
16573 hardware hazards. It is done in reorg step.
16574 (insn_length_variant_t): Remove.
16575 (insn_length_parameters_t): Likewise.
16576 (arc_insn_length_parameters): Likewise.
16577 (arc_get_insn_variants): Likewise.
16578 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
16580 2018-03-06 Jakub Jelinek <jakub@redhat.com>
16582 PR inline-asm/84683
16583 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
16586 PR tree-optimization/84687
16587 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
16589 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
16591 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16593 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
16594 Rename to ppc_speculation_barrier.
16595 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
16596 __builtin_ppc_speculation_barrier.
16598 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16601 * combine.c (combine_simplify_rtx): Don't try to simplify if
16602 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
16605 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
16607 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
16608 to 32 bytes when compiling for POWER9.
16610 2018-03-05 Jakub Jelinek <jakub@redhat.com>
16613 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
16614 regparm >= 3 with no arg reg available also for calls with
16615 flag_force_indirect_call. Pass decl to ix86_function_regparm.
16618 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
16620 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
16622 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
16625 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
16627 2018-03-05 Richard Biener <rguenther@suse.de>
16629 PR tree-optimization/84486
16630 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
16631 When inserting a __builtin_assume_aligned call set the LHS
16632 SSA name alignment info accordingly.
16634 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
16636 PR tree-optimization/84114
16637 * config/aarch64/aarch64.c (aarch64_reassociation_width)
16638 Avoid reassociation of FLOAT_MODE addition.
16640 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
16642 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
16643 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
16644 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
16645 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
16646 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
16647 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
16648 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
16650 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
16651 __builtin_ia32_wbinvd): New builtins.
16652 (SPECIAL_ARGS2): New.
16653 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
16654 (SPECIAL_ARGS2): New.
16655 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
16656 (ix86_valid_target_attribute_inner_p): Ditto.
16657 (ix86_init_mmx_sse_builtins): Add special_args2.
16658 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
16659 TARGET_WBNOINVD_P): New.
16660 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
16661 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
16662 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
16663 * config/i386/immintrin.h (_wbinvd): New intrinsic.
16664 * config/i386/pconfigintrin.h: New file.
16665 * config/i386/wbnoinvdintrin.h: Ditto.
16666 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
16668 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
16670 2018-03-05 Richard Biener <rguenther@suse.de>
16672 PR tree-optimization/84670
16673 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16675 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16676 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16677 (compute_antic_aux): Only assert the number of values in ANTIC_IN
16678 doesn't grow if all successors (recursively) were visited at least
16681 2018-03-05 Richard Biener <rguenther@suse.de>
16683 PR tree-optimization/84650
16684 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
16685 if executed in the loop pipeline.
16687 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
16689 * doc/configfiles.texi (Configuration Files): Move info about
16690 conditionalizing $target-protos.h to...
16691 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
16692 differs from $target-protos.h.
16694 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
16695 Chung-Ju Wu <jasonwucj@gmail.com>
16697 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
16698 * config/nds32/nds32-multiple.md (setmemsi): Define.
16699 * config/nds32/nds32-memory-manipulation.c
16700 (nds32_gen_dup_4_byte_to_word_value): New.
16701 (emit_setmem_word_loop): New.
16702 (emit_setmem_byte_loop): New.
16703 (nds32_expand_setmem_loop): New.
16704 (nds32_expand_setmem_loop_v3m): New.
16705 (nds32_expand_setmem_unroll): New.
16706 (nds32_expand_setmem): New.
16708 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16709 Chung-Ju Wu <jasonwucj@gmail.com>
16711 * config/nds32/nds32-memory-manipulation.c
16712 (nds32_emit_load_store): New.
16713 (nds32_emit_post_inc_load_store): New.
16714 (nds32_emit_mem_move): New.
16715 (nds32_emit_mem_move_block): New.
16716 (nds32_expand_movmemsi_loop_unknown_size): New.
16717 (nds32_expand_movmemsi_loop_known_size): New.
16718 (nds32_expand_movmemsi_loop): New.
16719 (nds32_expand_movmemsi_unroll): New.
16720 (nds32_expand_movmemqi): Rename ...
16721 (nds32_expand_movmemsi): ... to this.
16722 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
16723 (movmemsi): ... to this.
16724 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
16725 (nds32_expand_movmemsi): ... to this.
16727 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16728 Monk Chiang <sh.chiang04@gmail.com>
16729 Chung-Ju Wu <jasonwucj@gmail.com>
16731 * config/nds32/nds32-protos.h
16732 (nds32_expand_load_multiple): New arguments.
16733 (nds32_expand_store_multiple): Ditto.
16734 (nds32_valid_multiple_load_store): Rename ...
16735 (nds32_valid_multiple_load_store_p): ... to this.
16736 * config/nds32/nds32-memory-manipulation.c
16737 (nds32_expand_load_multiple): Refine implementation.
16738 (nds32_expand_store_multiple): Ditto.
16739 * config/nds32/nds32-multiple.md
16740 (load_multiple): Update nds32_expand_load_multiple interface.
16741 (store_multiple): Update nds32_expand_store_multiple interface.
16742 * config/nds32/nds32-predicates.c
16743 (nds32_valid_multiple_load_store): Rename ...
16744 (nds32_valid_multiple_load_store_p): ... to this and refine
16746 * config/nds32/predicates.md
16747 (nds32_load_multiple_and_update_address_operation): New predicate.
16748 (nds32_store_multiple_and_update_address_operation): New predicate.
16750 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
16751 Chung-Ju Wu <jasonwucj@gmail.com>
16753 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
16754 (combo): New attribute.
16755 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
16757 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
16759 * config/nds32/nds32.opt: Change -mcmodel= default value.
16761 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
16762 Monk Chiang <sh.chiang04@gmail.com>
16763 Chung-Ju Wu <jasonwucj@gmail.com>
16765 * config/nds32/constants.md (unspec_element): New enum.
16766 * config/nds32/constraints.md (Umw): New constraint.
16767 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
16768 * config/nds32/nds32-intrinsic.md: Likewise.
16769 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
16770 (nds32_valid_smw_lwm_base_p): New.
16771 (nds32_output_smw_single_word): New.
16772 (nds32_output_lmw_single_word): New.
16773 (nds32_expand_unaligned_load): New.
16774 (nds32_expand_unaligned_store): New.
16775 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
16776 (nds32_output_smw_single_word): Declare.
16777 (nds32_output_lmw_single_word): Declare.
16778 (nds32_expand_unaligned_load): Declare.
16779 (nds32_expand_unaligned_store): Declare.
16780 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
16781 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
16782 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
16783 NDS32_BUILTIN_UASTORE_DW.
16784 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
16787 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
16788 Kito Cheng <kito.cheng@gmail.com>
16789 Chung-Ju Wu <jasonwucj@gmail.com>
16791 * config/nds32/nds32-intrinsic.c
16792 (nds32_expand_builtin_null_ftype_reg): Delete.
16793 (nds32_expand_builtin_reg_ftype_imm): Ditto.
16794 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
16795 (nds32_read_argument): New.
16796 (nds32_legitimize_target): Ditto.
16797 (nds32_legitimize_argument): Ditto.
16798 (nds32_check_constant_argument): Ditto.
16799 (nds32_expand_unop_builtin): Ditto.
16800 (nds32_expand_unopimm_builtin): Ditto.
16801 (nds32_expand_binop_builtin): Ditto.
16802 (nds32_builtin_decl_impl): Ditto.
16803 (builtin_description): Ditto.
16804 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
16805 (nds32_init_builtins_impl): Ditto.
16806 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
16807 (nds32_builtin_decl): New.
16808 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
16809 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
16811 2018-03-02 Jeff Law <law@redhat.com>
16813 * reorg.c (stop_search_p): Handle DEBUG_INSN.
16814 (redundant_insn, fill_simple_delay_slots): Likewise.
16815 (fill_slots_from_thread): Likewise.
16816 * resource.c (mark_referenced_resources): Likewise.
16817 (mark_set_resources, find_dead_or_set_registers): Likewise.
16819 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16821 * substring-locations.h (format_warning_va): Formatting fix for
16822 ATTRIBUTE_GCC_DIAG.
16823 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
16825 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
16826 * substring-locations.c: Include intl.h.
16827 (format_warning_va): Turned into small wrapper around
16828 format_warning_n_va, renamed to ...
16829 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
16830 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
16832 (format_warning_at_substring_n): New function.
16833 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
16834 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
16835 format_warning_at_substring with just a shorter name instead of
16836 const function pointer.
16837 (fmtwarn_n): New function.
16838 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
16839 appropriate, get rid of all the fmtstr temporaries, move conditionals
16840 with G_() wrapped string literals directly into fmtwarn arguments,
16841 cast dir.len to (int), formatting fixes.
16843 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
16845 * doc/invoke.texi: Remove "Cilk Plus" references.
16847 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16848 Richard Biener <rguenther@suse.de>
16851 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
16852 for error or warning attributes if CALL_FROM_THUNK_P is set.
16855 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16858 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
16859 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
16862 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
16863 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
16865 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
16866 instead of -1U in last predictors element's probability member.
16868 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16871 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16872 arguments if they are comparable.
16874 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
16876 PR tree-optimization/84634
16877 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
16878 masks and masked_loop_p with a single loop_masks, making sure it's
16879 null for bb vectorization.
16881 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
16883 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
16884 (vect_analyze_data_ref_access): Use loop->safe_len rather than
16885 loop->force_vectorize to check whether there is no alias.
16887 2018-03-02 Jakub Jelinek <jakub@redhat.com>
16890 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
16892 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
16894 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
16895 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
16896 instead of a loop around prev_real_insn.
16897 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
16900 PR inline-asm/84625
16901 * config/i386/i386.c (ix86_print_operand): Use conditional
16902 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
16905 2018-03-02 Richard Biener <rguenther@suse.de>
16907 PR tree-optimization/84427
16908 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
16909 (bitmap_set_subtract_values): Rewrite to handle multiple
16912 (prune_clobbered_mems): Likewise.
16913 (phi_translate): Take edge instead of pred/phiblock.
16914 (phi_translate_1): Likewise.
16915 (phi_translate_set): Likewise. Insert all translated
16916 exprs for a value into the set, keeping possibly multiple
16917 expressions per value.
16918 (compute_antic_aux): Adjust for phi_translate changes.
16919 When intersecting union the expressions and prune those
16920 not in the final value set, keeping possibly multiple
16921 expressions per value. Do not use value-insertion
16922 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
16923 all expressions. Add verification that the value-sets
16924 only shrink during iteration.
16925 (compute_partial_antic_aux): Adjust for the phi_translate changes.
16926 (do_pre_regular_insertion): Likewise.
16927 (do_pre_partial_partial_insertion): Likewise.
16929 2018-03-02 Richard Biener <rguenther@suse.de>
16932 * config/darwin.c (saved_debug_info_level): New static global.
16933 (darwin_asm_lto_start): Disable debug info generation for LTO out.
16934 (darwin_asm_lto_end): Restore debug info generation settings.
16936 2018-03-01 Martin Liska <mliska@suse.cz>
16939 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
16940 volatile arguments.
16942 2018-03-01 Richard Biener <rguenther@suse.de>
16945 * dwarf2out.c (gen_variable_die): Properly handle late VLA
16946 type annotation with LTO when debug was disabled at compile-time.
16948 2018-03-01 Matthew Fortune <mfortune@gmail.com>
16950 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
16952 (mips_final_postscan_insn): Likewise.
16954 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
16956 PR rtl-optimization/84528
16957 * alias.c (init_alias_target): Add commentary.
16958 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
16959 a unique base value if the frame pointer is not eliminated
16960 to the stack pointer.
16962 2018-03-01 Tom de Vries <tom@codesourcery.com>
16964 PR rtl-optimization/83327
16965 * lra-int.h (hard_regs_spilled_into): Declare.
16966 * lra.c (hard_regs_spilled_into): Define.
16967 (init_reg_info): Init hard_regs_spilled_into.
16968 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
16969 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
16970 (process_bb_lives): Handle hard_regs_spilled_into.
16971 (lra_create_live_ranges_1): Before doing liveness propagation, clear
16972 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
16974 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
16976 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
16977 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
16978 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
16979 * config/rs6000/aix72.h: New file.
16981 2018-02-28 Jakub Jelinek <jakub@redhat.com>
16983 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
16984 instead of warning_at with conditional singular and plural messages
16988 * stor-layout.c (update_alignment_for_field): For
16989 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
16990 && !DECL_PACKED (field), do the alignment update, just use
16991 only desired_align instead of MAX (type_align, desired_align)
16993 (place_field): Don't do known_align < desired_align handling
16994 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
16995 is non-NULL, instead do it after rli->prev_field handling and
16996 only if not within a bitfield word. For DECL_PACKED (field)
16997 use type_align of BITS_PER_UNIT.
16999 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
17001 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
17002 superfluous parentheses and trailing spaces.
17004 2018-02-28 Richard Biener <rguenther@suse.de>
17006 PR tree-optimization/84584
17007 * graphite-scop-detection.c (scop_detection::add_scop): Discard
17008 SCoPs with fake exit edge.
17010 2018-02-28 Martin Liska <mliska@suse.cz>
17013 * timevar.c (timer::print): Fix format to properly print 100%
17016 2018-02-28 Richard Biener <rguenther@suse.de>
17018 PR middle-end/84607
17019 * genmatch.c (capture_info::walk_match): Do not mark
17020 captured expressions without operands as expr_p given
17021 they act more like predicates and should be subject to
17022 "lost tail" side-effect preserving.
17024 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
17026 PR rtl-optimization/81611
17027 * auto-inc-dec.c (attempt_change): Move dead note from
17028 mem_insn if it's the next use of regno
17029 (find_address): Take address use of reg holding
17030 non-incremented value. Add parm to limit search to the named
17032 (merge_in_block): Attempt to use a mem insn that is the next
17033 use of the original regno.
17035 2018-02-27 Martin Sebor <msebor@redhat.com>
17038 * doc/invoke.texi (-Wmissing-attributes): New option.
17039 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
17041 2018-02-27 Martin Sebor <msebor@redhat.com>
17043 PR translation/84207
17044 * diagnostic-core.h (warning_n, error_n, inform_n): Change
17045 n argument to unsigned HOST_WIDE_INT.
17046 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
17047 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
17048 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
17049 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
17051 2018-02-27 Richard Biener <rguenther@suse.de>
17053 PR tree-optimization/84512
17054 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
17055 Do not use the estimate returned from record_stmt_cost for
17056 the scalar iteration cost but sum properly using add_stmt_cost.
17058 2018-02-27 Richard Biener <rguenther@suse.de>
17060 PR tree-optimization/84466
17061 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17062 Adjust last change to less strictly validate use operands.
17064 2018-02-27 Martin Liska <mliska@suse.cz>
17066 PR gcov-profile/84548
17067 * gcov.c (process_file): Allow partial overlap and consider it
17068 also as group functions.
17069 (output_lines): Properly calculate range of lines for a group.
17071 2018-02-27 Martin Liska <mliska@suse.cz>
17073 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
17074 'ggc' suffixes. Change first column width.
17075 (timer::print): Fix formatting of the column.
17077 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
17079 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
17080 preserve inline entry blocks for the sake of debug inline
17081 entry point markers alone.
17082 (remove_unused_locals): Suggest in comments a better place to
17083 force the preservation of inline entry blocks that are
17084 otherwise unused, but do not preserve them.
17086 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17088 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
17090 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17093 * config/i386/constraints.md (Bs): Replace
17094 ix86_indirect_branch_register with
17095 TARGET_INDIRECT_BRANCH_REGISTER.
17097 * config/i386/i386.md (indirect_jump): Likewise.
17098 (tablejump): Likewise.
17099 (*sibcall_memory): Likewise.
17100 (*sibcall_value_memory): Likewise.
17101 Peepholes of indirect call and jump via memory: Likewise.
17102 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
17103 (*sibcall_value_GOT_32): Likewise.
17104 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17105 (GOT_memory_operand): Likewise.
17106 (call_insn_operand): Likewise.
17107 (sibcall_insn_operand): Likewise.
17108 (GOT32_symbol_operand): Likewise.
17109 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
17111 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17113 PR rtl-optimization/83496
17114 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
17115 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
17116 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
17117 redundant insn, if any.
17118 (relax_delay_slots): Likewise.
17119 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
17121 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
17123 PR tree-optimization/83965
17124 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
17125 that grouped statements are part of a reduction chain. Return
17126 true if the statement is not marked as a reduction itself but
17127 is part of a group.
17128 (vect_recog_dot_prod_pattern): Don't check whether the statement
17129 is part of a group here.
17130 (vect_recog_sad_pattern): Likewise.
17131 (vect_recog_widen_sum_pattern): Likewise.
17133 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
17136 * final.c (rest_of_clean_state): Also look for calls inside sequences.
17138 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
17141 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
17143 (ix86_output_indirect_function_return): New prototype.
17144 (ix86_split_simple_return_pop_internal): Likewise.
17145 * config/i386/i386.c (indirect_return_via_cx): New.
17146 (indirect_return_via_cx_bnd): Likewise.
17147 (indirect_thunk_name): Handle return va CX_REG.
17148 (output_indirect_thunk_function): Create alias for
17149 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
17150 (ix86_output_indirect_jmp): Remove the bool argument.
17151 (ix86_output_indirect_function_return): New function.
17152 (ix86_split_simple_return_pop_internal): Likewise.
17153 * config/i386/i386.md (*indirect_jump): Don't pass false
17154 to ix86_output_indirect_jmp.
17155 (*tablejump_1): Likewise.
17156 (simple_return_pop_internal): Change it to define_insn_and_split.
17157 Call ix86_split_simple_return_pop_internal to split it for
17158 -mfunction-return=.
17159 (simple_return_indirect_internal): Call
17160 ix86_output_indirect_function_return instead of
17161 ix86_output_indirect_jmp.
17163 2018-02-26 Jakub Jelinek <jakub@redhat.com>
17166 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
17167 memset and value initialization afterwards.
17169 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
17171 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
17173 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17176 * common/config/aarch64/aarch64-common.c
17177 (aarch_option_optimization_table[]): Switch
17178 off fomit-frame-pointer
17180 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17181 Chung-Ju Wu <jasonwucj@gmail.com>
17183 * config/nds32/nds32-multiple.md (load_multiple): Disallow
17185 (store_multiple): Ditto.
17187 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
17189 * config.gcc: Add --with-cpu support for nds32 target.
17190 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
17191 * config/nds32/nds32.opt: Add -mcpu= option.
17193 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
17195 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
17196 isel=yes): Warn for these deprecated options.
17198 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
17200 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
17201 ISA_2_5_MASKS_EMBEDDED.
17203 2018-02-23 Jakub Jelinek <jakub@redhat.com>
17205 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
17206 p->max as pointers rather than using iterative_hash_expr.
17208 2018-02-23 Carl Love <cel@us.ibm.com>
17210 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
17211 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
17213 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
17214 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
17215 P8V_BUILTIN_VEC_VUNSIGNED2.
17217 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
17220 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
17221 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
17223 * lra-constraints.c (curr_insn_transform): Set up
17224 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
17226 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
17227 * lra-eliminations.c (spill_pseudos): Ditto.
17228 (process_insn_for_elimination): Ditto.
17229 * lra-lives.c (reg_early_clobber_p): Use the new macros.
17230 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
17231 LRA_NON_CLOBBERED_ALT.
17233 2018-02-22 Martin Sebor <msebor@redhat.com>
17235 PR tree-optimization/84480
17236 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
17237 to maybe_diag_stxncpy_trunc. Call it.
17238 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
17239 from gimple_fold_builtin_strcpy. Print inlining stack.
17240 (handle_builtin_stxncpy): Print inlining stack.
17241 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
17243 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
17246 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
17247 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
17248 and -fcheck-pointer-bounds are used together.
17249 (indirect_thunk_prefix): New enum.
17250 (indirect_thunk_need_prefix): New function.
17251 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
17252 "_nt" instead of "_bnd" for NOTRACK prefix.
17253 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
17254 (output_indirect_thunk_function): Likewise.
17256 (ix86_code_end): Update output_indirect_thunk_function calls.
17257 (ix86_output_indirect_branch_via_reg): Replace
17258 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
17259 (ix86_output_indirect_branch_via_push): Likewise.
17260 (ix86_output_function_return): Likewise.
17261 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
17262 incompatible with -fcf-protection=branch and
17263 -fcheck-pointer-bounds.
17265 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17268 * config/aarch64/aarch64.c (aarch64_print_address_internal):
17269 Change gcc_assert call to output_operand_lossage.
17271 2018-02-22 Steve Ellcey <sellcey@cavium.com>
17273 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
17275 2018-02-22 DJ Delorie <dj@redhat.com>
17276 Sebastian Perta <sebastian.perta@renesas.com>
17277 Oleg Endo <olegendo@gcc.gnu.org>
17279 * config/rx/rx.c (rx_rtx_costs): New function.
17280 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
17282 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
17284 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
17286 2018-02-22 Martin Liska <mliska@suse.cz>
17289 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
17290 Add "native" as a possible value.
17292 2018-02-22 Martin Liska <mliska@suse.cz>
17295 * config/i386/i386.c (ix86_option_override_internal):
17296 Add "native" as a possible value for -march and -mtune.
17298 2018-02-22 Jakub Jelinek <jakub@redhat.com>
17301 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
17302 to all type variants.
17304 PR tree-optimization/84503
17305 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
17306 width as info->bitpos + info->bitsize - start.
17307 (merged_store_group::merge_overlapping): Simplify width computation.
17308 (check_no_overlap): New function.
17309 (imm_store_chain_info::try_coalesce_bswap): Compute expected
17310 start + width and last_order of the group, fail if check_no_overlap
17312 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
17313 to group if check_no_overlap fails.
17315 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17317 * config/rs6000/altivec.md: Delete contraint arguments to
17318 define_expand, define_split, and define_peephole2, and in
17319 define_insn_and_split if always unused.
17320 * config/rs6000/darwin.md: Ditto.
17321 * config/rs6000/dfp.md: Ditto.
17322 * config/rs6000/rs6000.md: Ditto.
17323 * config/rs6000/sync.md: Ditto.
17324 * config/rs6000/vector.md: Ditto.
17325 * config/rs6000/vsx.md: Ditto.
17327 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
17329 * config/rs6000/altivec.md: Write output control strings as braced
17330 blocks instead of double-quoted strings.
17331 * config/rs6000/darwin.md: Ditto.
17332 * config/rs6000/rs6000.md: Ditto.
17333 * config/rs6000/vector.md: Ditto.
17334 * config/rs6000/vsx.md: Ditto.
17336 2018-02-21 Jason Merrill <jason@redhat.com>
17338 PR c++/84314 - ICE with templates and fastcall attribute.
17339 * attribs.c (build_type_attribute_qual_variant): Remove assert.
17341 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17343 * ipa-cp.c (determine_versionability): Fix comment typos.
17345 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
17348 * ipa-cp.c (determine_versionability): Do not version functions caling
17351 2018-02-21 Martin Liska <mliska@suse.cz>
17354 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
17355 Add "native" as a possible value.
17356 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
17357 the macro when native cpu detection is available.
17359 2018-02-21 Martin Liska <mliska@suse.cz>
17362 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
17363 Add "native" as a possible value.
17364 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
17365 when native cpu detection is available.
17367 2018-02-21 Jakub Jelinek <jakub@redhat.com>
17368 Martin Sebor <msebor@redhat.com>
17370 PR tree-optimization/84478
17371 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
17373 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
17374 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
17375 support which is conservatively correct, for 2 only stay conservative
17376 for maxlen. Formatting and comment capitalization fixes. Add STRICT
17377 argument to the 2 argument get_range_strlen, adjust 6 arg
17378 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
17380 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
17381 (gimple_fold_builtin_strlen): Pass true as last argument to
17384 2018-02-20 Martin Sebor <msebor@redhat.com>
17386 PR middle-end/84095
17387 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
17388 (builtin_memref::set_base_and_offset): Same. Handle inner references.
17389 (builtin_memref::builtin_memref): Factor out parts into
17390 set_base_and_offset and call it.
17392 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17394 PR middle-end/84406
17395 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
17396 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
17397 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
17398 search at the associated MODE_INT.
17400 2018-02-20 Jeff Law <law@redhat.com>
17402 PR middle-end/82123
17403 PR tree-optimization/81592
17404 PR middle-end/79257
17405 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
17406 for range data rather than using global data.
17407 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
17408 range data rather than using global data.
17409 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
17410 pass it to children as needed.
17411 (struct directive::fmtresult): Similarly.
17412 (struct directive::set_width): Similarly.
17413 (struct directive::set_precision): Similarly.
17414 (format_integer, format_directive, parse_directive): Similarly.
17415 (format_none): Accept unnamed vr_values parameter.
17416 (format_percent, format_floating, format_character): Similarly.
17417 (format_string, format_plain): Similarly.
17418 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
17419 the EVRP range analyzer for range data rather than using global data.
17420 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
17421 gimple-ssa-evrp-analyze.h
17422 (class sprintf_dom_walker): Add after_dom_children member function.
17423 Add evrp_range_analyzer member.
17424 (sprintf_dom_walker::before_dom_children): Call into the EVRP
17425 range analyzer as needed.
17426 (sprintf_dom_walker::after_dom_children): New member function.
17427 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
17429 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
17430 (evrp_range_analyzer::pop_to_marker): Likewise.
17432 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
17434 PR tree-optimization/84419
17435 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
17436 with the required type if its current type is compatible but
17439 2018-02-20 Jakub Jelinek <jakub@redhat.com>
17441 PR middle-end/82004
17442 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
17443 after vectorization.
17445 2018-02-20 Martin Liska <mliska@suse.cz>
17448 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
17449 possible values if we don't have a hint.
17451 2018-02-20 Martin Liska <mliska@suse.cz>
17455 * final.c (shorten_branches): Build align_tab array with one
17457 * opts.c (finish_options): Add alignment option limit check.
17458 (MAX_CODE_ALIGN): Likewise.
17459 (MAX_CODE_ALIGN_VALUE): Likewise.
17460 * doc/invoke.texi: Document maximum allowed option value for
17461 all -falign-* options.
17463 2018-02-19 Jakub Jelinek <jakub@redhat.com>
17466 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
17467 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
17468 * var-tracking.c (emit_note_insn_var_location): Remove all references
17469 to NOTE_INSN_CALL_ARG_LOCATION.
17470 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
17471 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
17472 Use copy_rtx_if_shared.
17473 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
17474 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
17475 (dwarf2out_var_location): Remove handling of
17476 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
17478 * final.c (final_scan_insn): Remove all references to
17479 NOTE_INSN_CALL_ARG_LOCATION.
17480 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
17481 before dumping final insns.
17482 * except.c (emit_note_eh_region_end): Remove all references to
17483 NOTE_INSN_CALL_ARG_LOCATION.
17484 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
17485 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
17486 * config/arc/arc.c (hwloop_optimize): Likewise.
17487 * config/arm/arm.c (create_fix_barrier): Likewise.
17488 * config/s390/s390.c (s390_chunkify_start): Likewise.
17489 * config/sh/sh.c (find_barrier): Likewise.
17490 * config/i386/i386.c (rest_of_insert_endbranch,
17491 ix86_seh_fixup_eh_fallthru): Likewise.
17492 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
17493 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
17494 * config/frv/frv.c (frv_function_prologue): Likewise.
17495 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
17497 (note_outside_basic_block_p): Remove all references to
17498 NOTE_INSN_CALL_ARG_LOCATION.
17499 * gengtype.c (adjust_field_rtx_def): Likewise.
17500 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
17502 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
17503 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
17506 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
17509 PR tree-optimization/84452
17510 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
17511 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
17514 2018-02-19 Martin Liska <mliska@suse.cz>
17517 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
17519 2018-02-19 Martin Liska <mliska@suse.cz>
17520 Richard Sandiford <richard.sandiford@linaro.org>
17522 PR tree-optimization/82491
17523 * gimple-fold.c (get_base_constructor): Make earlier bail out
17526 2018-02-19 Carl Love <cel@us.ibm.com>
17528 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
17529 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
17531 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
17532 P8V_BUILTIN_VEC_NEG.
17534 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
17536 * config/rl78/rl78.md (movdf): New define expand.
17538 2018-02-19 Martin Liska <mliska@suse.cz>
17541 * doc/invoke.texi: Fix typo.
17542 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
17544 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
17546 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
17547 handle rs6000_single_float and rs6000_double_float specially for
17550 2018-02-16 Jeff Law <law@redhat.com>
17552 * config/rx/rx.c (add_pop_cfi_notes): New function.;
17553 (pop_regs): Use it.
17555 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17558 * ipa-inline.c (inline_small_functions): Fix a typo.
17560 2018-02-16 Nathan Sidwell <nathan@acm.org>
17562 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
17564 2018-02-16 Carl Love <cel@us.ibm.com>
17566 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
17567 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
17568 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
17569 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
17570 expansion to P8V_BUILTIN_VEC_FLOAT2.
17572 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
17574 PR rtl-optimization/70023
17575 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
17576 src_regno into account.
17578 2018-02-16 Carl Love <cel@us.ibm.com>
17580 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
17581 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
17582 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
17583 * config/rs6000/rs6000.c: Remove case statements for
17584 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
17585 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
17586 and P9V_BUILTIN_VEC_VINSERT4B.
17587 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
17588 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
17589 * config/rs6000/vsx.md:
17590 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
17593 2018-02-16 Carl Love <cel@us.ibm.com>
17595 * config/rs6000/altivec.h: Add builtin names vec_extract4b
17597 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
17599 * config/rs6000/rs6000-c.c: Add the definitions for
17600 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
17601 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
17602 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
17603 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
17604 definition for insert4b and define insn *insert3b_internal.
17605 * doc/extend.texi: Add documentation for vec_extract4b.
17607 2018-02-16 Nathan Sidwell <nathan@acm.org>
17609 * doc/extend.texi (Backwards Compatibility): Mention friend
17610 injection. Note for-scope is deprecated.
17611 * doc/invoke.texi (-ffriend-injection): Deprecate.
17613 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
17615 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
17616 that moved to I2, also allow destinations that are a paradoxical
17617 subreg (instead of a normal reg).
17619 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
17622 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
17625 2018-02-16 Richard Biener <rguenther@suse.de>
17627 PR tree-optimization/84037
17628 PR tree-optimization/84016
17630 * config/i386/i386.c (ix86_builtin_vectorization_cost):
17631 Adjust vec_construct for the fact we need additional higher latency
17632 128bit inserts for AVX256 and AVX512 vector builds.
17633 (ix86_add_stmt_cost): Scale vector construction cost for
17636 2018-02-16 Richard Biener <rguenther@suse.de>
17638 PR tree-optimization/84417
17639 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
17640 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
17641 (non_rewritable_lvalue_p): Likewise, use poly-ints.
17643 2018-02-16 Martin Liska <mliska@suse.cz>
17646 * internal-fn.def (ASAN_CHECK): Set proper flags.
17647 (ASAN_MARK): Likewise.
17649 2018-02-16 Julia Koval <julia.koval@intel.com>
17651 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
17652 from PTA_CANNONLAKE.
17654 2018-02-16 Jakub Jelinek <jakub@redhat.com>
17657 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
17658 Use ++iter rather than iter++ for std::list iterators.
17659 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
17660 defer deleting them until all nodes in the forest are processed. Do
17661 free even leaf nodes. Change to_process into auto_vec.
17664 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
17665 * vec.h (vec_default_construct): Use memset instead of placement new
17666 if BROKEN_VALUE_INITIALIZATION is defined.
17667 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
17668 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
17671 PR rtl-optimization/83723
17672 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
17673 * lra.c (lra_substitute_pseudo): Likewise. If true, use
17674 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
17676 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
17678 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
17680 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
17682 PR rtl-optimization/81443
17683 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
17684 from inner REGs to paradoxical SUBREGs.
17686 2018-02-16 Richard Biener <rguenther@suse.de>
17688 PR tree-optimization/84399
17689 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
17690 For operands we can analyze at their definition make sure we can
17691 analyze them at each use as well.
17693 2018-02-16 Richard Biener <rguenther@suse.de>
17695 PR tree-optimization/84190
17696 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
17697 volatile accesses if the decl isn't volatile.
17699 2018-02-15 Jason Merrill <jason@redhat.com>
17701 PR c++/84314 - ICE with templates and fastcall attribute.
17702 * attribs.c (build_type_attribute_qual_variant): Don't clobber
17703 TYPE_CANONICAL on an existing type.
17705 2018-02-15 Jakub Jelinek <jakub@redhat.com>
17707 PR tree-optimization/84383
17708 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
17709 dstoff nor call operand_equal_p if dstbase is NULL.
17711 PR tree-optimization/84334
17712 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
17713 also a CONSTANT_CLASS_P, punt.
17715 2018-02-14 Jim Wilson <jimw@sifive.com>
17717 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
17718 first SMALL_OPERAND check. New local min_second_step. Move assert
17719 to where locals are set. Add TARGET_RVC support.
17720 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
17722 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
17724 * doc/invoke.texi: Correct -Wformat-overflow code sample.
17726 2018-02-14 Martin Sebor <msebor@redhat.com>
17728 PR tree-optimization/83698
17729 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
17730 arrays constrain the offset range to their bounds.
17731 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
17732 (builtin_access::overlap): Avoid setting the size of overlap if it's
17734 (maybe_diag_overlap): Also consider arrays when deciding what values
17735 of offsets to include in diagnostics.
17737 2018-02-14 Martin Sebor <msebor@redhat.com>
17740 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
17741 that correspond to the kind of a declaration.
17743 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
17746 * config/pa/pa.md: Load address of PIC label using the linkage table
17747 if the label is nonlocal.
17749 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
17751 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
17752 warning message if user requests -maltivec=be.
17753 * doc/invoke.texi: Document deprecation of -maltivec=be.
17755 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
17758 * config/rs6000/rs6000-c.c: Update definitions for
17759 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
17760 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
17762 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17765 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
17766 add _get_ssp intrinsics. Remove argument from
17767 __builtin_ia32_rdssp[d|q].
17768 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
17769 * config/i386/i386-builtin.def: Remove argument from
17770 __builtin_ia32_rdssp[d|q].
17771 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
17772 ix86_expand_special_args_builtin for _rdssp[d|q].
17773 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
17774 Clear register before usage.
17775 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
17776 Add documentation for new _get_ssp and _inc_ssp intrinsics.
17778 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
17780 PR tree-optimization/84357
17781 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
17782 operand 1 of an ARRAY_REF too.
17784 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
17787 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
17788 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17790 (set_of_reg): New struct.
17791 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
17792 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
17793 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
17795 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
17796 Split into bitclr, bitset, bitinvert patterns if appropriate.
17797 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
17798 use rx_fuse_in_memory_bitop.
17799 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
17800 to named insn, correct maximum insn length.
17802 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
17805 * machmode.def: Define a complex mode for PARTIAL_INT.
17806 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
17808 * doc/rtl.texi: Document CSPImode.
17809 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
17811 (msp430_hard_regno_nregs_with_padding): Likewise.
17813 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
17816 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
17818 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
17820 PR rtl-optimization/84169
17821 * combine.c (try_combine): New variable split_i2i3. Set it to true if
17822 we generated a parallel as new i3 and we split that to new i2 and i3
17823 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
17824 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
17825 those to i2, not i1. Partially rewrite this scan code.
17827 2018-02-13 Jakub Jelinek <jakub@redhat.com>
17830 * stor-layout.c (place_field): For variable length fields, adjust
17831 offset_align afterwards not just based on the field's alignment,
17832 but also on the size.
17834 PR middle-end/84309
17835 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
17836 of exps and logs in the use_exp2 case.
17838 2018-02-13 Jeff Law <law@redhat.com>
17840 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
17841 entry for "vector".
17843 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
17846 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
17850 * common.opt (gas-loc-support, gas-locview-support): New.
17851 (ginline-points, ginternal-reset-location-views): New.
17852 * doc/invoke.texi: Document them. Use @itemx where intended.
17853 (gvariable-location-views): Adjust.
17854 * target.def (reset_location_view): New.
17855 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
17856 (TARGET_RESET_LOCATION_VIEW): New.
17857 * doc/tm.texi: Rebuilt.
17858 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
17859 (dwarf2out_default_as_locview_support): New.
17860 (output_asm_line_debug_info): Use option variables.
17861 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
17862 (output_loc_list): Likewise.
17863 (add_high_low_attributes): Check option variables.
17864 Don't output entry view attribute in strict mode.
17865 (gen_inlined_subroutine_die): Check option variables.
17866 (dwarf2out_inline_entry): Likewise.
17867 (init_sections_and_labels): Likewise.
17868 (dwarf2out_early_finish): Likewise.
17869 (maybe_reset_location_view): New, from...
17870 (dwarf2out_var_location): ... here. Call it.
17871 * debug.h (dwarf2out_default_as_loc_support): Declare.
17872 (dwarf2out_default_as_locview_support): Declare.
17873 * hooks.c (hook_int_rtx_insn_0): New.
17874 * hooks.h (hook_int_rtx_insn_0): Declare.
17875 * toplev.c (process_options): Take -gas-loc-support and
17876 -gas-locview-support from dwarf2out. Enable
17877 -gvariable-location-views by default only with locview
17878 assembler support. Enable -ginternal-reset-location-views by
17879 default only if the target defines the corresponding hook.
17880 Enable -ginline-points by default if location views are
17881 enabled; force it disabled if statement frontiers are
17883 * tree-inline.c (expand_call_inline): Check option variables.
17884 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
17886 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
17888 PR tree-optimization/84321
17889 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
17890 handling. Also check whether the anti-range contains any values
17891 that satisfy the mask; switch to a VR_RANGE if not.
17893 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
17896 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
17898 2018-02-13 Martin Jambor <mjambor@suse.cz>
17901 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
17902 of call statements, also set location of a load to a temporary.
17904 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
17906 * config/rl78/rl78.c (add_vector_labels): New function.
17907 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
17908 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
17909 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
17910 which checks that no arguments are passed.
17911 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
17912 * doc/extend.texi: Documentation for the new attribute.
17914 2018-02-13 Andreas Schwab <schwab@suse.de>
17916 * config/riscv/linux.h (CPP_SPEC): Define.
17918 2018-02-13 Jakub Jelinek <jakub@redhat.com>
17921 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
17922 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
17923 OPTION_MASK_ISA_AES as first argument to def_builtin_const
17924 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
17925 instead of OPTION_MASK_ISA_PCLMUL as first argument to
17926 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
17927 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
17928 temporarily for AES and PCLMUL builtins.
17930 PR tree-optimization/84339
17931 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
17932 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
17935 PR middle-end/84309
17936 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
17937 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
17938 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
17940 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
17942 * omp-simd-clone.h: New file.
17943 * omp-simd-clone.c: Include omp-simd-clone.h.
17944 (expand_simd_clones): No longer static.
17945 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
17946 cgraph.h and omp-simd-clone.h.
17947 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
17948 (vect_recog_widen_shift_pattern): Formatting fix.
17949 (vect_pattern_recog_1): Don't check optab for calls.
17952 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
17953 operands[2] into a REG before using gen_lowpart on it.
17955 2018-02-12 Jeff Law <law@redhat.com>
17958 * config/sh/sh.c (find_barrier): Consider a sibling call
17961 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
17962 successfully back substituting a reg.
17964 2018-02-12 Richard Biener <rguenther@suse.de>
17966 PR tree-optimization/84037
17967 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
17968 parameter, move visited init to caller.
17969 (vect_slp_analyze_operations): Separate cost from validity
17970 check, initialize visited once for all instances.
17971 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
17973 * tree-vect-stmts.c (vect_model_simple_cost): Make early
17975 (vect_model_promotion_demotion_cost): Likewise.
17976 (vectorizable_bswap): Guard cost modeling with !slp_node
17977 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
17979 (vectorizable_call): Likewise.
17980 (vectorizable_conversion): Likewise.
17981 (vectorizable_assignment): Likewise.
17982 (vectorizable_shift): Likewise.
17983 (vectorizable_operation): Likewise.
17984 (vectorizable_store): Likewise.
17985 (vectorizable_load): Likewise.
17986 (vectorizable_condition): Likewise.
17987 (vectorizable_comparison): Likewise.
17989 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
17992 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
17993 (ASAN_MARK): Fix fnspec to account for return value, change pointer
17994 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
17996 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17998 PR middle-end/83665
17999 * params.def (inline-min-speedup): Increase from 8 to 15.
18000 (max-inline-insns-auto): Decrease from 40 to 30.
18001 * ipa-split.c (consider_split): Add some buffer for function to
18002 be considered inlining candidate.
18003 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
18006 2018-02-12 Richard Biener <rguenther@suse.de>
18008 PR tree-optimization/84037
18009 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
18010 matched stmts if we cannot swap the non-matched ones.
18012 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18014 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
18015 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
18016 _mm_maskz_scalef_round_ss): New intrinsics.
18017 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
18018 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
18019 __builtin_ia32_scalefss_round): Remove.
18020 (__builtin_ia32_scalefsd_mask_round,
18021 __builtin_ia32_scalefss_mask_round): New intrinsics.
18022 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
18023 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
18024 ((match_operand:VF_128 2 "<round_nimm_predicate>"
18025 "<round_constraint>")): Changed to ...
18026 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
18027 "<round_scalar_constraint>")): ... this.
18028 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
18029 %0, %1, %2<round_op3>}"): Changed to ...
18030 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
18031 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
18032 %2<round_scalar_mask_op3>}"): ... this.
18033 * config/i386/subst.md (round_scalar_nimm_predicate): New.
18035 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
18037 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
18038 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
18039 (_mm_maskz_sqrt_round_ss): New intrinsics.
18040 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
18041 (__builtin_ia32_sqrtsd_mask_round)
18042 (__builtin_ia32_sqrtss_mask_round): New builtins.
18043 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
18044 (__builtin_ia32_sqrtss_round): Remove.
18045 (__builtin_ia32_sqrtsd_mask_round)
18046 (__builtin_ia32_sqrtss_mask_round): New builtins.
18047 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
18048 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
18049 ((match_operand:VF_128 1 "vector_operand"
18050 "xBm,<round_constraint>")): Changed to ...
18051 ((match_operand:VF_128 1 "vector_operand"
18052 "xBm,<round_scalar_constraint>")): ... this.
18053 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
18054 %0, %2, %<iptr>1<round_op3>}): Changed to ...
18055 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
18056 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
18057 %<iptr>1<round_scalar_mask_op3>}): ... this.
18058 ((set_attr "prefix" "<round_prefix>")): Changed to ...
18059 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
18061 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
18064 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
18065 Cast vec_cmpeq result to correct type.
18066 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
18067 Cast vec_cmpgt result to correct type.
18069 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
18071 * final.c (final_scan_insn_1): Renamed from...
18072 (final_scan_insn): ... this. New wrapper, to recover
18073 seen from the outermost call in recursive ones.
18074 * config/sparc/sparc.c (output_return): Drop seen from call.
18075 (output_sibcall): Likewise.
18076 * config/visium/visium.c (output_branch): Likewise.
18078 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
18080 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
18083 2018-02-10 Alan Modra <amodra@gmail.com>
18086 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
18087 Specify LR as an input.
18089 2018-02-10 Jakub Jelinek <jakub@redhat.com>
18092 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
18093 remove_member_access_dummy_vars): New functions.
18094 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
18095 lower_omp_1, execute_lower_omp): Use them.
18097 PR rtl-optimization/84308
18098 * shrink-wrap.c (spread_components): Release todo vector.
18100 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
18102 PR rtl-optimization/57193
18103 * ira-color.c (struct allocno_color_data): Add member
18104 conflict_allocno_hard_prefs.
18105 (update_conflict_allocno_hard_prefs): New.
18106 (bucket_allocno_compare_func): Add a preference based on
18107 conflict_allocno_hard_prefs.
18108 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
18109 (color_allocnos): Remove a dead code. Initiate
18110 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
18112 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18115 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
18116 constraint from =wa to wa. Avoid a subreg on the output operand,
18117 instead use a pseudo and subreg it in a move.
18118 (p9_xxbrd_<mode>): Changed to ...
18119 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
18120 (p9_xxbrd_v2df): New expander.
18121 (p9_xxbrw_<mode>): Changed to ...
18122 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
18123 (p9_xxbrw_v4sf): New expander.
18125 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18127 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
18129 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
18132 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
18133 multiply in 32-bit mode.
18134 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
18135 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
18138 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
18140 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
18141 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
18142 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
18143 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
18145 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
18148 * dwarf2out.c (is_trivial_indirect_ref): New function.
18149 (dwarf2out_late_global_decl): Do not generate a location
18150 attribute for variables that have a non-trivial DECL_VALUE_EXPR
18151 and that are not defined in the current unit.
18153 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18155 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
18156 instead of a libcall for UNORDERED.
18158 2018-02-09 Tamar Christina <tamar.christina@arm.com>
18161 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
18162 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
18164 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18167 * config/s390/s390.c (s390_set_current_function): Invoke
18168 s390_indirect_branch_settings also if fndecl didn't change.
18170 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18172 * config/rs6000/rs6000.md (blockage): Set length to zero.
18174 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
18176 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
18178 2018-02-09 Jakub Jelinek <jakub@redhat.com>
18181 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
18182 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
18186 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
18187 PARALLEL incoming that failed vt_get_decl_and_offset check.
18189 PR middle-end/84237
18190 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
18191 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
18193 (get_variable_section): For decls in named .bss* sections pass true as
18194 second argument to bss_initializer_p.
18196 2018-02-09 Marek Polacek <polacek@redhat.com>
18197 Jakub Jelinek <jakub@redhat.com>
18200 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
18201 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
18202 Sync some changes from cxx_fold_indirect_ref.
18204 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
18206 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
18208 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
18209 (BLOCK_INLINE_ENTRY_LABEL): New.
18210 (dwarf2out_var_location): Disregard inline entry markers.
18211 (inline_entry_data): New struct.
18212 (inline_entry_data_hasher): New hashtable type.
18213 (inline_entry_data_hasher::hash): New.
18214 (inline_entry_data_hasher::equal): New.
18215 (inline_entry_data_table): New variable.
18216 (add_high_low_attributes): Add DW_AT_entry_pc and
18217 DW_AT_GNU_entry_view attributes if a pending entry is found
18218 in inline_entry_data_table. Add old entry_pc attribute only
18219 if debug nonbinding markers are disabled.
18220 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
18221 markers are enabled.
18222 (block_within_block_p, dwarf2out_inline_entry): New.
18223 (dwarf2out_finish): Check that no entries remained in
18224 inline_entry_data_table.
18225 * final.c (reemit_insn_block_notes): Handle inline entry notes.
18226 (final_scan_insn, notice_source_line): Likewise.
18227 (rest_of_clean_state): Skip inline entry markers.
18228 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
18230 * gimple.c (gimple_build_debug_inline_entry): New.
18231 * gimple.h (enum gimple_debug_subcode): Add
18232 GIMPLE_DEBUG_INLINE_ENTRY.
18233 (gimple_build_debug_inline_entry): Declare.
18234 (gimple_debug_inline_entry_p): New.
18235 (gimple_debug_nonbind_marker_p): Adjust.
18236 * insn-notes.def (INLINE_ENTRY): New.
18237 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
18238 inline entry marker notes.
18239 (print_insn): Likewise.
18240 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
18241 (INSN_DEBUG_MARKER_KIND): Likewise.
18242 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
18243 * tree-inline.c (expand_call_inline): Build and insert
18244 debug_inline_entry stmt.
18245 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
18246 inline entry blocks early, if nonbind markers are enabled.
18247 (dump_scope_block): Dump fragment info.
18248 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
18249 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
18250 (gimple_build_debug_inline_entry): New.
18251 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
18252 Enable/disable inline entry points too.
18253 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
18254 (DEBUG_INSN): Describe inline entry markers.
18256 * common.opt (gvariable-location-views): New.
18257 (gvariable-location-views=incompat5): New.
18258 * config.in: Rebuilt.
18259 * configure: Rebuilt.
18260 * configure.ac: Test assembler for view support.
18261 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
18262 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
18263 * dwarf2out.c (var_loc_view): New typedef.
18264 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
18265 (dwarf2out_locviews_in_attribute): New.
18266 (dwarf2out_locviews_in_loclist): New.
18267 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
18268 (enum dw_line_info_opcode): Add LI_adv_address.
18269 (struct dw_line_info_table): Add view.
18270 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
18271 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
18272 (zero_view_p): New variable.
18273 (ZERO_VIEW_P): New macro.
18274 (output_asm_line_debug_info): New.
18275 (struct var_loc_node): Add view.
18276 (add_AT_view_list, AT_loc_list): New.
18277 (add_var_loc_to_decl): Add view param. Test it against last.
18278 (new_loc_list): Add view params. Record them.
18279 (AT_loc_list_ptr): Handle loc and view lists.
18280 (view_list_to_loc_list_val_node): New.
18281 (print_dw_val): Handle dw_val_class_view_list.
18282 (size_of_die): Likewise.
18283 (value_format): Likewise.
18284 (loc_list_has_views): New.
18285 (gen_llsym): Set vl_symbol too.
18286 (maybe_gen_llsym, skip_loc_list_entry): New.
18287 (dwarf2out_maybe_output_loclist_view_pair): New.
18288 (output_loc_list): Output view list or entries too.
18289 (output_view_list_offset): New.
18290 (output_die): Handle dw_val_class_view_list.
18291 (output_dwarf_version): New.
18292 (output_compilation_unit_header): Use it.
18293 (output_skeleton_debug_sections): Likewise.
18294 (output_rnglists, output_line_info): Likewise.
18295 (output_pubnames, output_aranges): Update version comments.
18296 (output_one_line_info_table): Output view numbers in asm comments.
18297 (dw_loc_list): Determine current endview, pass it to new_loc_list.
18298 Call maybe_gen_llsym.
18299 (loc_list_from_tree_1): Adjust.
18300 (add_AT_location_description): Create view list attribute if
18301 needed, check it's absent otherwise.
18302 (convert_cfa_to_fb_loc_list): Adjust.
18303 (maybe_emit_file): Call output_asm_line_debug_info for test.
18304 (dwarf2out_var_location): Reset views as needed. Precompute
18305 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
18306 attribute. Set view.
18307 (new_line_info_table): Reset next view.
18308 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
18309 (dwarf2out_source_line): Likewise. Output view resets and labels to
18310 the assembler, or select appropriate line info opcodes.
18311 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
18312 (optimize_string_length): Catch it. Adjust.
18313 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
18314 dw_val_class_view_list, and remove it if no longer needed.
18315 (hash_loc_list): Hash view numbers.
18316 (loc_list_hasher::equal): Compare them.
18317 (optimize_location_lists): Check whether a view list symbol is
18318 needed, and whether the locview attribute is present, and
18319 whether they match. Remove the locview attribute if no longer
18321 (index_location_lists): Call skip_loc_list_entry for test.
18322 (dwarf2out_finish): Call output_asm_line_debug_info for test.
18323 Use output_dwarf_version.
18324 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
18325 (struct dw_val_node): Add val_view_list.
18326 * final.c (SEEN_NEXT_VIEW): New.
18327 (set_next_view_needed): New.
18328 (clear_next_view_needed): New.
18329 (maybe_output_next_view): New.
18330 (final_start_function): Rename to...
18331 (final_start_function_1): ... this. Take pointer to FIRST,
18332 add SEEN parameter. Emit param bindings in the initial view.
18333 (final_start_function): Reintroduce SEEN-less interface.
18334 (final): Rename to...
18335 (final_1): ... this. Take SEEN parameter. Output final pending
18336 next view at the end.
18337 (final): Reintroduce seen-less interface.
18338 (final_scan_insn): Output pending next view before switching
18339 sections or ending a block. Mark the next view as needed when
18340 outputting variable locations. Notify debug backend of section
18341 changes, and of location view changes.
18342 (rest_of_handle_final): Adjust.
18343 * toplev.c (process_options): Autodetect value for debug variable
18344 location views option. Warn on incompat5 without -gdwarf-5.
18345 * doc/invoke.texi (gvariable-location-views): New.
18346 (gvariable-location-views=incompat5): New.
18347 (gno-variable-location-views): New.
18349 2018-02-08 David Malcolm <dmalcolm@redhat.com>
18351 PR tree-optimization/84136
18352 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
18353 that the result of find_edge is non-NULL.
18355 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18358 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
18359 storing integer register in SImode. Fix cost of 256 and 512
18360 byte aligned SSE register store.
18362 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
18364 * config/i386/i386.c (ix86_multiplication_cost): Fix
18365 multiplication cost for TARGET_AVX512DQ.
18367 2018-02-08 Marek Polacek <polacek@redhat.com>
18369 PR tree-optimization/84238
18370 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
18373 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18375 PR tree-optimization/84265
18376 * tree-vect-stmts.c (vectorizable_store): Don't treat
18377 VMAT_CONTIGUOUS accesses as grouped.
18378 (vectorizable_load): Likewise.
18380 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18382 PR tree-optimization/81635
18383 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
18384 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
18385 (test_round_for_mask): New functions.
18386 (wide_int_cc_tests): Call test_round_for_mask.
18387 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
18388 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
18389 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
18390 range returned by get_range_info.
18392 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18395 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
18396 * symtab.c: Include builtins.h
18397 (symtab_node::output_to_lto_symbol_table_p): Move here
18398 from lto-streamer-out.c:output_symbol_p.
18399 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
18400 (output_symbol_p): Move all logic to symtab.c
18401 (produce_symtab): Update.
18403 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18405 * config/s390/s390-opts.h (enum indirect_branch): Define.
18406 * config/s390/s390-protos.h (s390_return_addr_from_memory)
18407 (s390_indirect_branch_via_thunk)
18408 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
18409 (enum s390_indirect_branch_type): Define.
18410 * config/s390/s390.c (struct s390_frame_layout, struct
18411 machine_function): Remove.
18412 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
18413 (indirect_branch_table_label_no, indirect_branch_table_name):
18415 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
18416 (enum s390_indirect_branch_option): Define.
18417 (s390_return_addr_from_memory): New function.
18418 (s390_handle_string_attribute): New function.
18419 (s390_attribute_table): Add new attribute handler.
18420 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
18421 (s390_indirect_branch_via_thunk): New function.
18422 (s390_indirect_branch_via_inline_thunk): New function.
18423 (s390_function_ok_for_sibcall): When jumping via thunk disallow
18424 sibling call optimization for non z10 compiles.
18425 (s390_emit_call): Force indirect branch target to be a single
18426 register. Add r1 clobber for non-z10 compiles.
18427 (s390_emit_epilogue): Emit return jump via return_use expander.
18428 (s390_reorg): Handle JUMP_INSNs as execute targets.
18429 (s390_option_override_internal): Perform validity checks for the
18430 new command line options.
18431 (s390_indirect_branch_attrvalue): New function.
18432 (s390_indirect_branch_settings): New function.
18433 (s390_set_current_function): Invoke s390_indirect_branch_settings.
18434 (s390_output_indirect_thunk_function): New function.
18435 (s390_code_end): Implement target hook.
18436 (s390_case_values_threshold): Implement target hook.
18437 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
18439 * config/s390/s390.h (struct s390_frame_layout)
18440 (struct machine_function): Move here from s390.c.
18441 (TARGET_INDIRECT_BRANCH_NOBP_RET)
18442 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
18443 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
18444 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
18445 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
18446 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
18447 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
18448 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
18449 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
18450 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
18451 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
18452 (mnemonic attribute): Add values which aren't recognized
18454 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
18455 pattern for branch conversion. Fix mnemonic attribute.
18456 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
18457 indirect branch via thunk if requested.
18458 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
18459 ("*indirect_jump"): Disable for branch conversion using out of
18461 ("indirect_jump_via_thunk<mode>_z10")
18462 ("indirect_jump_via_thunk<mode>")
18463 ("indirect_jump_via_inlinethunk<mode>_z10")
18464 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
18465 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
18466 ("casesi_jump_via_inlinethunk<mode>_z10")
18467 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
18468 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
18469 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
18470 ("*indirect2_jump"): Disable for branch conversion.
18471 ("casesi_jump"): Turn into expander and expand patterns for branch
18473 ("return_use"): New expander.
18474 ("*return"): Emit return via thunk and rename it to ...
18475 ("*return<mode>"): ... this one.
18476 * config/s390/s390.opt: Add new options and and enum for the
18479 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
18481 * lra-constraints.c (match_reload): Unconditionally use
18482 gen_lowpart_SUBREG, rather than selecting between that
18483 and equivalent gen_rtx_SUBREG code.
18485 2018-02-08 Richard Biener <rguenther@suse.de>
18487 PR tree-optimization/84233
18488 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
18489 changed flag instead of boguously re-using phi_inserted.
18491 2018-02-08 Martin Jambor <mjambor@suse.cz>
18493 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
18494 static local variables.
18496 2018-02-08 Richard Biener <rguenther@suse.de>
18498 PR tree-optimization/84278
18499 * tree-vect-stmts.c (vectorizable_store): When looking for
18500 smaller vector types to perform grouped strided loads/stores
18501 make sure the mode is supported by the target.
18502 (vectorizable_load): Likewise.
18504 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18506 * config/aarch64/aarch64.c (aarch64_components_for_bb):
18507 Increase LDP/STP opportunities by adding adjacent callee-saves.
18509 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
18511 PR rtl-optimization/84068
18512 PR rtl-optimization/83459
18513 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
18515 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
18517 PR tree-optimization/84224
18518 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
18519 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
18520 non-zero arguments.
18522 2018-02-07 Iain Sandoe <iain@codesourcery.com>
18525 * config/rs6000/altivec.md (*restore_world): Remove LR use.
18526 * config/rs6000/predicates.md (restore_world_operation): Adjust op
18527 count, remove one USE.
18529 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
18531 * doc/install.texi (Configuration): Document the
18532 --with-long-double-format={ibm,ieee} PowerPC configuration
18536 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
18537 Convert from define_expand to be define_insn_and_split. Rework
18538 float/double/_Float128 conversions to QI/HI/SImode to work with
18539 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
18540 conversions to QI/HImode types did a store and then a load to
18541 truncate the value. For conversions to VSX registers, don't split
18542 the insn, instead emit the code directly. Use the code iterator
18543 any_fix to combine signed and unsigned conversions.
18544 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
18545 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
18546 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18547 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
18548 (fix_<mode>di2_hw): Likewise.
18549 (fixuns_<mode>di2_hw): Likewise.
18550 (fix_<mode>si2_hw): Likewise.
18551 (fixuns_<mode>si2_hw): Likewise.
18552 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
18553 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
18554 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
18555 fix<uns>_trunc<SFDF:mode>si2_p8.
18556 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
18557 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
18558 (fix<uns>_<mode>_mem): Likewise.
18559 (fctiw<u>z_<mode>_mem): Likewise.
18560 (fix<uns>_<mode>_mem): Likewise.
18561 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
18562 the register allocator from doing a direct move to the GPRs to do
18563 a store, and instead use the ISA 3.0 store byte/half-word from
18564 vector register instruction. For IEEE 128-bit floating point,
18565 also optimize stores of 32-bit ints.
18566 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
18568 2018-02-07 Alan Hayward <alan.hayward@arm.com>
18570 * genextract.c (push_pathstr_operand): New function to support
18572 (walk_rtx): Call push_pathstr_operand.
18573 (print_path): Support [a-zA-Z].
18575 2018-02-07 Richard Biener <rguenther@suse.de>
18577 PR tree-optimization/84037
18578 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
18579 (cse_and_gimplify_to_preheader): Declare.
18580 (vect_get_place_in_interleaving_chain): Likewise.
18581 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18583 (_loop_vec_info::~_loop_vec_info): Delete it.
18584 (cse_and_gimplify_to_preheader): New function.
18585 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
18586 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
18587 (vectorizable_load): Likewise. For grouped stores always base
18588 the IV on the first element.
18589 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
18590 condition before gimplifying.
18592 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18594 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
18595 *DIV_EXPR and *MOD_EXPR.
18597 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
18600 * config/i386/i386.c (ix86_option_override_internal): Mask out
18601 the CF_SET bit when checking -fcf-protection.
18603 2018-02-07 Tom de Vries <tom@codesourcery.com>
18606 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
18609 2018-02-07 Richard Biener <rguenther@suse.de>
18611 PR tree-optimization/84204
18612 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
18615 PR tree-optimization/84205
18616 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
18617 special-case isl_ast_op_zdiv_r.
18619 PR tree-optimization/84223
18620 * graphite-scop-detection.c (gather_bbs::before_dom_children):
18621 Only add conditions from within the region.
18622 (gather_bbs::after_dom_children): Adjust.
18624 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
18627 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
18628 * config/avr/avr.md: Only post-reload split REG-REG moves if
18629 either register is GENERAL_REG_P.
18631 2018-02-07 Jakub Jelinek <jakub@redhat.com>
18633 PR tree-optimization/84235
18634 * tree-ssa-scopedtables.c
18635 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
18636 if the subtraction is performed in floating point type where NaNs are
18637 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
18638 build 1. Formatting fix.
18640 2018-02-06 Jakub Jelinek <jakub@redhat.com>
18643 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
18644 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
18645 and skip it regardless of bb boundaries. Use CALL_P macro,
18646 don't test INSN_P (insn) together with CALL_P or JUMP_P check
18647 unnecessarily, formatting fix.
18649 2018-02-06 Michael Collison <michael.collison@arm.com>
18651 * config/arm/thumb2.md:
18652 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
18653 (*thumb_mov_notscc): Ditto.
18655 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
18658 * config/rs6000/rs6000.md (su code attribute): Use "u" for
18659 unsigned_fix, not "s".
18661 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18663 * configure.ac (gcc_fn_eh_frame_ro): New function.
18664 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
18665 correct .eh_frame permissions.
18666 * configure: Regenerate.
18668 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
18670 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
18671 irrelevant options.
18673 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18675 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18676 Display warning message for -mno-speculate-indirect-jumps.
18678 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
18680 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
18682 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
18684 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
18686 PR tree-optimization/84225
18687 * tree-eh.c (find_trapping_overflow): Only call
18688 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
18690 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18693 * config/i386/i386.c: Reimplement the check of possible options
18694 -mibt/-mshstk conbination. Change error messages.
18695 * doc/invoke.texi: Fix a typo: remove extra '='.
18697 2018-02-06 Marek Polacek <polacek@redhat.com>
18699 PR tree-optimization/84228
18700 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
18702 2018-02-06 Tamar Christina <tamar.christina@arm.com>
18705 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
18706 emitted arch directives.
18707 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
18708 __ARM_FEATURE_COPROC before changing architectures.
18710 2018-02-06 Richard Biener <rguenther@suse.de>
18712 * config/i386/i386.c (print_reg): Fix typo.
18713 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
18715 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
18717 * configure: Regenerate.
18719 2018-02-05 Martin Sebor <msebor@redhat.com>
18721 PR tree-optimization/83369
18722 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
18725 2018-02-05 Martin Liska <mliska@suse.cz>
18727 * doc/invoke.texi: Cherry-pick upstream r323995.
18729 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
18731 * ira.c (ira_init_register_move_cost): Adjust comment.
18733 2018-02-05 Martin Liska <mliska@suse.cz>
18735 PR gcov-profile/84137
18736 * doc/gcov.texi: Fix typo in documentation.
18738 2018-02-05 Martin Liska <mliska@suse.cz>
18740 PR gcov-profile/83879
18741 * doc/gcov.texi: Document necessity of --dynamic-list-data when
18742 using dlopen functionality.
18744 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
18746 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
18747 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
18748 _mm_maskz_range_ss, _mm_mask_range_round_ss,
18749 _mm_maskz_range_round_ss): New intrinsics.
18750 (__builtin_ia32_rangesd128_round)
18751 (__builtin_ia32_rangess128_round): Remove.
18752 (__builtin_ia32_rangesd128_mask_round,
18753 __builtin_ia32_rangess128_mask_round): New builtins.
18754 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
18755 __builtin_ia32_rangess128_round): Remove.
18756 (__builtin_ia32_rangesd128_mask_round,
18757 __builtin_ia32_rangess128_mask_round): New builtins.
18758 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
18759 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
18760 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
18761 "<round_saeonly_constraint>")): Changed to ...
18762 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
18763 "<round_saeonly_scalar_constraint>")): ... this.
18764 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
18765 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
18766 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
18767 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
18768 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
18770 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
18772 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
18774 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
18775 Remove all values except native, 8540 and 8548.
18777 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
18779 * config/i386/i386.c (ix86_output_function_return): Pass
18780 INVALID_REGNUM, instead of -1, as invalid register number to
18781 indirect_thunk_name and output_indirect_thunk.
18783 2018-02-02 Julia Koval <julia.koval@intel.com>
18785 * config.gcc: Add -march=icelake.
18786 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
18787 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
18788 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
18789 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
18790 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
18791 (processor_target_table): Add icelake.
18792 (ix86_option_override_internal): Handle new PTAs.
18793 (get_builtin_code_for_version): Handle icelake.
18794 (M_INTEL_COREI7_ICELAKE): New.
18795 (fold_builtin_cpu): Handle icelake.
18796 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
18797 * doc/invoke.texi: Add -march=icelake.
18799 2018-02-02 Julia Koval <julia.koval@intel.com>
18801 * config/i386/i386.c (ix86_option_override_internal): Change flags type
18802 to wide_int_bitmask.
18803 * wide-int-bitmask.h: New.
18805 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18808 * config/i386/i386.md: Replace Pmode with word_mode in
18809 builtin_setjmp_setup and builtin_longjmp to support x32.
18811 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
18815 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
18817 (rs6000_supported_cpu_names): New static variable.
18818 (linux_cpu_translation_table): Likewise.
18819 (elf_platform) <cpu>: Define new static variable and use it.
18820 Translate kernel AT_PLATFORM name to canonical name if needed.
18821 Error if platform name is unknown.
18823 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
18826 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
18828 2018-02-01 Jeff Law <law@redhat.com>
18831 * config/i386/i386.c (release_scratch_register_on_entry): Add new
18832 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
18833 the scratch if RELEASE_VIA_POP is false.
18834 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
18835 If we have to save a temporary register, decrement SIZE appropriately.
18836 Pass new arguments to release_scratch_register_on_entry.
18837 (ix86_adjust_stack_and_probe): Likewise.
18838 (ix86_emit_probe_stack_range): Pass new arguments to
18839 release_scratch_register_on_entry.
18841 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
18843 PR rtl-optimization/84157
18844 * combine.c (change_zero_ext): Use REG_P predicate in
18845 front of HARD_REGISTER_P predicate.
18847 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
18849 * config/avr/avr.c (avr_option_override): Move disabling of
18850 -fdelete-null-pointer-checks to...
18851 * common/config/avr/avr-common.c (avr_option_optimization_table):
18854 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18856 PR tree-optimization/81635
18857 * tree-data-ref.c (split_constant_offset_1): For types that
18858 wrap on overflow, try to use range info to prove that wrapping
18861 2018-02-01 Renlin Li <renlin.li@arm.com>
18864 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
18865 TAILCALL_ADDR_REGS.
18866 (aarch64_register_move_cost): Likewise.
18867 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
18868 TAILCALL_ADDR_REGS.
18869 (REG_CLASS_NAMES): Likewise.
18870 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
18871 TAILCALL_ADDR_REGS. Remove IP registers.
18872 * config/aarch64/aarch64.md (Ucs): Update register constraint.
18874 2018-02-01 Richard Biener <rguenther@suse.de>
18876 * domwalk.h (dom_walker::dom_walker): Add additional constructor
18877 for specifying RPO order and allow NULL for that.
18878 * domwalk.c (dom_walker::dom_walker): Likewise.
18879 (dom_walker::walk): Handle NULL RPO order.
18880 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
18882 (rewrite_update_dom_walker): Likewise.
18883 (mark_def_dom_walker): Likewise.
18885 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18887 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
18888 (aarch64_maybe_expand_sve_subreg_move): Declare.
18889 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
18890 * config/aarch64/predicates.md (aarch64_any_register_operand): New
18892 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
18893 that are semantically a reverse operation.
18894 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
18895 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
18896 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
18898 (aarch64_can_change_mode_class): For big-endian, forbid changes
18899 between two SVE modes if they have different element sizes.
18901 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18903 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
18904 the TImode handling for big-endian targets.
18906 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18908 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
18909 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
18911 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
18912 Remove BSWAP handing for big-endian targets and use the form of
18913 LD1RQ appropariate for the mode.
18915 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18917 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
18918 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
18919 duplicated element.
18921 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
18924 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
18925 check for operands that need to go through aarch64_sve_reload_be.
18927 2018-02-01 Jakub Jelinek <jakub@redhat.com>
18929 PR tree-optimization/81661
18930 PR tree-optimization/84117
18931 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
18932 * tree-eh.c: Include gimplify.h.
18933 (find_trapping_overflow, replace_trapping_overflow,
18934 rewrite_to_non_trapping_overflow): New functions.
18935 * tree-vect-loop.c: Include tree-eh.h.
18936 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
18937 * tree-data-ref.c: Include tree-eh.h.
18938 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
18940 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
18942 PR rtl-optimization/84123
18943 * combine.c (change_zero_ext): Check if hard register satisfies
18944 can_change_dest_mode before calling gen_lowpart_SUBREG.
18946 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
18949 * ira.c (ira_init_register_move_cost): Remove assert.
18951 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18953 PR rtl-optimization/84071
18954 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
18955 * doc/tm.texi: Regenerate.
18957 2018-01-31 Richard Biener <rguenther@suse.de>
18959 PR tree-optimization/84132
18960 * tree-data-ref.c (analyze_miv_subscript): Properly
18961 check whether evolution_function_is_affine_multivariate_p
18962 before calling gcd_of_steps_may_divide_p.
18964 2018-01-31 Julia Koval <julia.koval@intel.com>
18967 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
18968 * config/i386/i386.md (rdpid_rex64) New.
18969 (rdpid): Make 32bit only.
18971 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
18974 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
18975 an IDENTIFIER_NODE for FUNCTION_TYPE's.
18977 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18980 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18982 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18984 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
18986 PR rtl-optimization/84071
18987 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
18988 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
18990 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
18992 * config/arc/arc.c (arc_handle_aux_attribute): New function.
18993 (arc_attribute_table): Add 'aux' attribute.
18994 (arc_in_small_data_p): Consider aux like variables.
18995 (arc_is_aux_reg_p): New function.
18996 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
18997 (arc_get_aux_arg): New function.
18998 (prepare_move_operands): Handle aux-register access.
18999 (arc_handle_aux_attribute): New function.
19000 * doc/extend.texi (ARC Variable attributes): Add subsection.
19002 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19004 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
19005 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
19006 (arc_attribute_table): Add 'uncached' attribute.
19007 (arc_print_operand): Print '.di' flag for uncached memory
19009 (arc_in_small_data_p): Do not consider for small data the uncached
19011 (arc_is_uncached_mem_p): New function.
19012 * config/arc/predicates.md (compact_store_memory_operand): Check
19013 for uncached memory accesses.
19014 (nonvol_nonimm_operand): Likewise.
19015 * doc/extend.texi (ARC Type Attribute): New subsection.
19017 2018-01-31 Jakub Jelinek <jakub@redhat.com>
19020 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
19021 falign-loops=): Add Optimization flag.
19023 2018-01-30 Jeff Law <law@redhat.com>
19026 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
19027 INT_REGISTERS_SAVED. Check it prior to calling
19028 get_scratch_register_on_entry.
19029 (ix86_adjust_stack_and_probe): Similarly.
19030 (ix86_emit_probe_stack_range): Similarly.
19031 (ix86_expand_prologue): Corresponding changes.
19033 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19036 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
19037 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
19039 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
19042 * lra-constraints.c (curr_insn_transform): Process AND in the
19045 2018-01-30 Jakub Jelinek <jakub@redhat.com>
19047 PR rtl-optimization/83986
19048 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
19049 dependence against last_pending_memory_flush in addition to
19050 pending_jump_insns.
19052 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
19054 PR tree-optimization/81611
19055 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
19058 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19061 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
19064 2018-01-30 Richard Biener <rguenther@suse.de>
19065 Jakub Jelinek <jakub@redhat.com>
19067 PR tree-optimization/84111
19068 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
19069 inner loops added during recursion, as they don't have up-to-date
19072 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19075 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
19076 (can_inline_edge_by_limits_p): ... here.
19077 (can_early_inline_edge_p, check_callers,
19078 update_caller_keys, update_callee_keys, recursive_inlining,
19079 add_new_edges_to_heap, speculation_useful_p,
19080 inline_small_functions,
19081 inline_small_functions, flatten_function,
19082 inline_to_all_callers_1): Update.
19084 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
19086 * profile-count.c (profile_count::combine_with_ipa_count): Handle
19089 2018-01-30 Richard Biener <rguenther@suse.de>
19091 PR tree-optimization/83008
19092 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
19093 invariant and constant vector uses in stmts when they need
19094 more than one stmt.
19096 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19099 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
19100 * configure: Regenerate.
19102 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19104 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
19106 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
19107 Use gen_rtx_REG rather than gen_lowpart.
19109 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19111 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
19112 rather than 0 when creating partial subregs.
19114 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
19116 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
19119 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
19122 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
19123 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
19124 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
19125 flags. This restores the settings used before the 2017-07-24.
19126 Turning off pre increment/decrement/modify allows IVOPTS to
19127 optimize DF/SF loops where the index is an int.
19129 2018-01-29 Richard Biener <rguenther@suse.de>
19130 Kelvin Nilsen <kelvin@gcc.gnu.org>
19133 * tree-vect-stmts.c (vectorizable_call): Don't call
19134 targetm.vectorize_builtin_md_vectorized_function if callee is
19137 2018-01-22 Carl Love <cel@us.ibm.com>
19139 * doc/extend.tex: Fix typo in second arg in
19140 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
19142 2018-01-29 Richard Biener <rguenther@suse.de>
19144 PR tree-optimization/84086
19145 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
19146 (flush_ssaname_freelist): When SSA names were released reset
19147 the SCEV hash table.
19149 2018-01-29 Richard Biener <rguenther@suse.de>
19151 PR tree-optimization/84057
19152 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
19153 removed paths when removing edges.
19155 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
19157 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
19158 -mfunction-return=@var{choice}.
19160 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
19162 PR diagnostic/84034
19163 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
19164 Handle CR like TAB.
19165 (layout::print_source_line): Likewise.
19166 (test_get_line_width_without_trailing_whitespace): Add test cases.
19168 2018-01-27 Jakub Jelinek <jakub@redhat.com>
19170 PR middle-end/84040
19171 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
19174 2018-01-26 Jim Wilson <jimw@sifive.com>
19176 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
19178 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
19181 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19183 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
19184 and CMP + SUB-immediate -> SUBS.
19186 2018-01-26 Martin Sebor <msebor@redhat.com>
19188 PR tree-optimization/83896
19189 * tree-ssa-strlen.c (get_string_len): Rename...
19190 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
19191 Avoid assuming length is constant.
19192 (handle_char_store): Use HOST_WIDE_INT for string length.
19194 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
19197 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
19198 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
19200 2018-01-26 Richard Biener <rguenther@suse.de>
19202 PR rtl-optimization/84003
19203 * dse.c (record_store): Only record redundant stores when
19204 the earlier store aliases at least all accesses the later one does.
19206 2018-01-26 Jakub Jelinek <jakub@redhat.com>
19208 PR rtl-optimization/83985
19209 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
19210 REG_CFA_RESTORE insns.
19211 (delete_unmarked_insns): Don't ignore separate shrink wrapping
19212 REG_CFA_RESTORE insns here.
19215 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
19216 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
19218 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19220 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
19221 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
19222 (arc_init): Likewise.
19223 (arc_override_options): Likewise.
19224 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
19226 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
19228 * config/arc/arc.h (TARGET_DBNZ): Define.
19229 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
19230 properly set the tune attribute.
19231 (dbnz): Use TARGET_DBNZ guard.
19232 * config/arc/arc.opt (mtune): Add core3 option.
19234 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19236 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
19237 recognize new pic like addresses.
19238 (arc_delegitimize_address): Clean up.
19240 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19242 * config/arc/arc-arches.def: Option mrf16 valid for all
19244 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
19245 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
19246 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
19247 * config/arc/arc-tables.opt: Regenerate.
19248 * config/arc/arc.c (arc_conditional_register_usage): Handle
19249 reduced register file case.
19250 (arc_file_start): Set must have build attributes.
19251 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
19252 mrf16 option value.
19253 * config/arc/arc.opt (mrf16): Add new option.
19254 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
19255 * config/arc/genmultilib.awk: Handle new mrf16 option.
19256 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
19257 * config/arc/t-multilib: Regenerate.
19258 * doc/invoke.texi (ARC Options): Document mrf16 option.
19260 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19262 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
19263 * config/arc/arc.c (arc_handle_secure_attribute): New function.
19264 (arc_attribute_table): Add 'secure_call' attribute.
19265 (arc_print_operand): Print secure call operand.
19266 (arc_function_ok_for_sibcall): Don't optimize tail calls when
19268 (arc_is_secure_call_p): New function. * config/arc/arc.md
19269 (call_i): Add support for sjli instruction.
19270 (call_value_i): Likewise.
19271 * config/arc/constraints.md (Csc): New constraint.
19273 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
19274 John Eric Martin <John.Martin@emmicro-us.com>
19276 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
19277 * config/arc/arc.c (_arc_jli_section): New struct.
19278 (arc_jli_section): New type.
19279 (rc_jli_sections): New static variable.
19280 (arc_handle_jli_attribute): New function.
19281 (arc_attribute_table): Add jli_always and jli_fixed attribute.
19282 (arc_file_end): New function.
19283 (TARGET_ASM_FILE_END): Define.
19284 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
19285 (arc_add_jli_section): New function.
19286 (jli_call_scan): Likewise.
19287 (arc_reorg): Call jli_call_scan.
19288 (arc_output_addsi): Remove 'S' from printing asm operand.
19289 (arc_is_jli_call_p): New function.
19290 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
19292 (movhi_insn): Likewise.
19293 (movsi_insn): Likewise.
19294 (movsi_set_cc_insn): Likewise.
19295 (loadqi_update): Likewise.
19296 (load_zeroextendqisi_update): Likewise.
19297 (load_signextendqisi_update): Likewise.
19298 (loadhi_update): Likewise.
19299 (load_zeroextendhisi_update): Likewise.
19300 (load_signextendhisi_update): Likewise.
19301 (loadsi_update): Likewise.
19302 (loadsf_update): Likewise.
19303 (movsicc_insn): Likewise.
19304 (bset_insn): Likewise.
19305 (bxor_insn): Likewise.
19306 (bclr_insn): Likewise.
19307 (bmsk_insn): Likewise.
19308 (bicsi3_insn): Likewise.
19309 (cmpsi_cc_c_insn): Likewise.
19310 (movsi_ne): Likewise.
19311 (movsi_cond_exec): Likewise.
19312 (clrsbsi2): Likewise.
19313 (norm_f): Likewise.
19323 (call_i): Remove 'S' asm letter, add jli instruction.
19324 (call_value_i): Likewise.
19325 * config/arc/arc.op (mjli-always): New option.
19326 * config/arc/constraints.md (Cji): New constraint.
19327 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
19329 (subsf3_fpx): Likewise.
19330 (mulsf3_fpx): Likewise.
19331 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
19333 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
19334 function attrbutes.
19335 * doc/invoke.texi (ARC): Document mjli-always option.
19337 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
19339 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
19340 avoid addition with 0 and use incw and decw where possible.
19342 2018-01-26 Richard Biener <rguenther@suse.de>
19344 PR tree-optimization/81082
19345 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
19346 association if it requires casting to unsigned.
19347 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
19348 from fold_plusminus_mult_expr to catch important cases late when
19349 range info is available.
19351 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19353 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
19354 * configure.ac (hidden_linkonce): New test.
19355 * configure: Regenerate.
19356 * config.in: Regenerate.
19358 2018-01-26 Julia Koval <julia.koval@intel.com>
19360 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
19361 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
19362 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
19363 _mm_mask_bitshuffle_epi64_mask): Fix type.
19364 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
19365 USI_FTYPE_V4DI_V4DI_USI): Remove.
19366 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
19367 __builtin_ia32_vpshufbitqmb256_mask,
19368 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
19369 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
19370 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
19372 2018-01-26 Alan Modra <amodra@gmail.com>
19375 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
19376 UNSPEC_VBPERMQ. Sort other unspecs.
19378 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
19380 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
19382 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
19384 PR middle-end/83055
19385 * predict.c (drop_profile): Do not push/pop cfun; update also
19387 (handle_missing_profiles): Fix logic looking for zero profiles.
19389 2018-01-25 Jakub Jelinek <jakub@redhat.com>
19391 PR middle-end/83977
19392 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
19393 on functions with #pragma omp declare simd or functions with simd
19395 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
19396 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
19397 Remove trailing \n from warning_at calls.
19399 2018-01-25 Tom de Vries <tom@codesourcery.com>
19402 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19403 for neutered workers.
19405 2018-01-24 Joseph Myers <joseph@codesourcery.com>
19408 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
19409 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
19411 2018-01-24 Jeff Law <law@redhat.com>
19414 * i386.c (get_probe_interval): Move to earlier point.
19415 (ix86_compute_frame_layout): If -fstack-clash-protection and
19416 the frame is larger than the probe interval, then use pushes
19417 to save registers rather than reg->mem moves.
19418 (ix86_expand_prologue): Remove conditional for int_registers_saved
19421 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
19424 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
19425 min/max for never referenced object.
19427 2018-01-24 Jakub Jelinek <jakub@redhat.com>
19429 PR middle-end/83977
19430 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
19432 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
19433 attributes from DECL_ATTRIBUTES (decl) without affecting
19434 DECL_ATTRIBUTES (current_function_decl).
19435 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
19436 functions with non-NULL DECL_ABSTRACT_ORIGIN.
19438 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
19440 PR tree-optimization/83979
19441 * fold-const.c (fold_comparison): Use constant_boolean_node
19442 instead of boolean_{true,false}_node.
19444 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
19446 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
19449 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19451 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19452 Simplify the clause that sets the length attribute.
19453 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19454 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
19455 clause that sets the length attribute.
19456 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19458 2018-01-24 Tom de Vries <tom@codesourcery.com>
19461 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
19462 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
19463 Add strict parameter.
19464 (prevent_branch_around_nothing): Insert dummy insn between branch to
19465 label and label with no ptx insn inbetween.
19466 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
19468 2018-01-24 Tom de Vries <tom@codesourcery.com>
19471 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
19472 for neutered threads in warp.
19473 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
19475 2018-01-24 Richard Biener <rguenther@suse.de>
19477 PR tree-optimization/83176
19478 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
19481 2018-01-24 Richard Biener <rguenther@suse.de>
19483 PR tree-optimization/82819
19484 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
19485 code generating pluses that are no-ops in the target precision.
19487 2018-01-24 Richard Biener <rguenther@suse.de>
19489 PR middle-end/84000
19490 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
19492 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19494 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
19495 to merge probabilities.
19496 * predict.c (probably_never_executed): Also mark as cold functions
19497 with global 0 profile and guessed local profile.
19498 * profile-count.c (profile_probability::combine_with_count): New
19500 * profile-count.h (profile_probability::operator*,
19501 profile_probability::operator*=, profile_probability::operator/,
19502 profile_probability::operator/=): Reduce precision to adjusted
19503 and set value to guessed on contradictory divisions.
19504 (profile_probability::combine_with_freq): Remove.
19505 (profile_probability::combine_wiht_count): Declare.
19506 (profile_count::force_nonzero):: Set to adjusted.
19507 (profile_count::probability_in):: Set quality to adjusted.
19508 * tree-ssa-tail-merge.c (replace_block_by): Use
19509 combine_with_count.
19511 2018-01-23 Andrew Waterman <andrew@sifive.com>
19512 Jim Wilson <jimw@sifive.com>
19514 * config/riscv/riscv.c (riscv_stack_boundary): New.
19515 (riscv_option_override): Set riscv_stack_boundary. Handle
19516 riscv_preferred_stack_boundary_arg.
19517 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
19518 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
19519 (STACK_BOUNDARY): Set to riscv_stack_boundary.
19520 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
19521 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
19522 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
19524 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
19527 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
19528 of struct ix86_frame.
19529 (ix86_expand_epilogue): Likewise. Add a local variable for
19530 the reg_save_offset field in struct ix86_frame.
19532 2018-01-23 Bin Cheng <bin.cheng@arm.com>
19534 PR tree-optimization/82604
19535 * tree-loop-distribution.c (enum partition_kind): New enum item
19536 PKIND_PARTIAL_MEMSET.
19537 (partition_builtin_p): Support above new enum item.
19538 (generate_code_for_partition): Ditto.
19539 (compute_access_range): Differentiate cases that equality can be
19540 proven at all loops, the innermost loops or no loops.
19541 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
19542 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
19543 (finalize_partitions, distribute_loop): Don't fuse partition of
19544 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
19545 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
19546 parloop is enabled.
19548 2018-01-23 Martin Liska <mliska@suse.cz>
19550 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
19551 order to ignore the predictor.
19552 (PRED_POLYMORPHIC_CALL): Likewise.
19553 (PRED_RECURSIVE_CALL): Likewise.
19555 2018-01-23 Martin Liska <mliska@suse.cz>
19557 * tree-profile.c (tree_profiling): Print function header to
19558 aware reader which function we are working on.
19559 * value-prof.c (gimple_find_values_to_profile): Do not print
19560 not interesting value histograms.
19562 2018-01-23 Martin Liska <mliska@suse.cz>
19564 * profile-count.h (enum profile_quality): Add
19565 profile_uninitialized as the first value. Do not number values
19566 as they are zero based.
19567 (profile_count::verify): Update sanity check.
19568 (profile_probability::verify): Likewise.
19570 2018-01-23 Nathan Sidwell <nathan@acm.org>
19572 * doc/invoke.texi (ffor-scope): Deprecate.
19574 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19576 PR tree-optimization/83510
19577 * domwalk.c (set_all_edges_as_executable): New function.
19578 (dom_walker::dom_walker): Convert bool param
19579 "skip_unreachable_blocks" to enum reachability. Move setup of
19580 edge flags to set_all_edges_as_executable and only do it when
19581 reachability is REACHABLE_BLOCKS.
19582 * domwalk.h (enum dom_walker::reachability): New enum.
19583 (dom_walker::dom_walker): Convert bool param
19584 "skip_unreachable_blocks" to enum reachability.
19585 (set_all_edges_as_executable): New decl.
19586 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
19587 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
19589 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
19590 but converting true to REACHABLE_BLOCKS.
19591 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
19593 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
19594 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
19595 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
19597 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
19598 if check_all_array_refs will be called.
19600 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19602 * tree.c (selftest::test_location_wrappers): Add more test
19605 2018-01-23 David Malcolm <dmalcolm@redhat.com>
19607 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
19608 (selftest::test_bit_in_range): Likewise.
19610 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19613 * doc/sourcebuild.texi (vect_float): Say that the selector
19614 only describes the situation when -funsafe-math-optimizations is on.
19615 (vect_float_strict): Document.
19617 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
19619 PR tree-optimization/83965
19620 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
19621 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
19622 instead of checking only for a reduction.
19623 (vect_recog_widen_sum_pattern): Likewise.
19625 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19627 * predict.c (probably_never_executed): Only use precise profile info.
19628 (compute_function_frequency): Skip after inlining hack since we now
19629 have quality checking.
19631 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
19633 * profile-count.h (profile_probability::very_unlikely,
19634 profile_probability::unlikely, profile_probability::even): Set
19635 precision to guessed.
19637 2018-01-23 Richard Biener <rguenther@suse.de>
19639 PR tree-optimization/83963
19640 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
19641 Properly terminate dominator walk when crossing the exit edge not
19642 when visiting its source block.
19644 2018-01-23 Jakub Jelinek <jakub@redhat.com>
19647 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
19648 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
19650 2018-01-22 Jakub Jelinek <jakub@redhat.com>
19652 PR tree-optimization/83957
19653 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
19654 semicolon after for body surrounded by braces.
19656 PR tree-optimization/83081
19657 * profile-count.h (profile_probability::split): New method.
19658 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
19659 Use profile_probability::split.
19660 (do_compare_rtx_and_jump): Fix adjustment of probabilities
19661 when splitting a single conditional jump into 2.
19663 2018-01-22 David Malcolm <dmalcolm@redhat.com>
19665 PR tree-optimization/69452
19666 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
19669 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19671 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
19672 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
19673 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
19675 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19677 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
19679 * config/rl78/rl78.md (movdi): New define_expand.
19680 * config/rl78/rl78.c (rl78_split_movdi): New function.
19682 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
19685 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
19687 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
19688 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
19689 128-bit to produce an UNSPEC move to get the double word with the
19690 signbit and then a shift directly to do signbit.
19691 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
19692 implementation with a new version that just does either a direct
19693 move or a regular move. Move memory interface to separate insns.
19694 Move insns so they are next to the expander.
19695 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
19696 with signbit move. Split big and little endian case.
19697 (signbit<mode>2_dm_mem_le): Likewise.
19698 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
19699 (signbit<mode>2_dm2): Likewise.
19701 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19703 * config/rl78/rl78.md (anddi3): New define_expand.
19705 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19707 * config/rl78/rl78.md (umindi3): New define_expand.
19709 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19711 * config/rl78/rl78.md (smindi3): New define_expand.
19713 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19715 * config/rl78/rl78.md (smaxdi3): New define_expand.
19717 2018-01-22 Carl Love <cel@us.ibm.com>
19719 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
19720 LVX_V1TI): Add macro expansion.
19721 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
19722 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
19723 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
19724 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
19725 Change check to determine if the instruction is a byte reversing
19726 entry. Fix typo in comment.
19727 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
19728 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
19729 Add def_builtin calls for new builtins.
19730 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
19731 Add define_insn expansion.
19733 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19735 * config/rl78/rl78.md (umaxdi3): New define_expand.
19737 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
19739 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
19740 for non-QImode registers.
19742 2018-01-22 Richard Biener <rguenther@suse.de>
19744 PR tree-optimization/83963
19745 * graphite-scop-detection.c (scop_detection::get_sese): Delay
19746 including the loop exit block.
19747 (scop_detection::merge_sese): Likewise.
19748 (scop_detection::add_scop): Do it here instead.
19750 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19752 * doc/sourcebuild.texi (arm_softfloat): Document.
19754 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
19757 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
19758 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
19759 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
19761 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19762 David Edelsohn <dje.gcc@gmail.com>
19765 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19766 Change "crset eq" to "crset 2".
19767 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19768 (*call_indirect_aix<mode>_nospec): Likewise.
19769 (*call_value_indirect_aix<mode>_nospec): Likewise.
19770 (*call_indirect_elfv2<mode>_nospec): Likewise.
19771 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
19772 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
19773 change assembly output from . to $.
19774 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
19775 (indirect_jump<mode>_nospec): Change assembly output from . to $.
19776 (*tablejump<mode>_internal1_nospec): Likewise.
19778 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
19781 * config/sh/sh_optimize_sett_clrt.cc:
19782 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
19784 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
19786 PR tree-optimization/83940
19787 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
19788 offset_dt to vect_constant_def rather than vect_unknown_def_type.
19789 (vect_check_load_store_mask): Add a mask_dt_out parameter and
19790 use it to pass back the definition type.
19791 (vect_check_store_rhs): Likewise rhs_dt_out.
19792 (vect_build_gather_load_calls): Add a mask_dt argument and use
19793 it instead of a call to vect_is_simple_use.
19794 (vectorizable_store): Update calls to vect_check_load_store_mask
19795 and vect_check_store_rhs. Use the dt returned by the latter instead
19796 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
19797 instead of calls to vect_is_simple_use. Pass the scalar rather
19798 than the vector operand to vect_is_simple_use when handling
19799 second and subsequent copies of an rhs value.
19800 (vectorizable_load): Update calls to vect_check_load_store_mask
19801 and vect_build_gather_load_calls. Use the cached mask_dt and
19802 gs_info.offset_dt instead of calls to vect_is_simple_use.
19804 2018-01-20 Jakub Jelinek <jakub@redhat.com>
19806 PR middle-end/83945
19807 * tree-emutls.c: Include gimplify.h.
19808 (lower_emutls_2): New function.
19809 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
19810 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
19811 it before further processing.
19814 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
19815 UINTVAL (trueop1) instead of INTVAL (op1).
19817 2018-01-19 Jakub Jelinek <jakub@redhat.com>
19821 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
19822 INCOMING_FRAME_SP_OFFSET if not defined.
19823 (scan_trace): Add ENTRY argument. If true and
19824 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
19825 emit a note to adjust the CFA offset.
19826 (create_cfi_notes): Adjust scan_trace callers.
19827 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
19828 INCOMING_FRAME_SP_OFFSET in the CIE.
19829 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
19830 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
19832 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
19833 * doc/tm.texi: Regenerated.
19835 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19837 PR rtl-optimization/83147
19838 * lra-constraints.c (remove_inheritance_pseudos): Use
19839 lra_substitute_pseudo_within_insn.
19841 2018-01-19 Tom de Vries <tom@codesourcery.com>
19842 Cesar Philippidis <cesar@codesourcery.com>
19845 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
19847 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
19850 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
19851 spaces for function labels.
19853 2018-01-19 Martin Liska <mliska@suse.cz>
19855 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
19856 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
19857 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
19858 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
19859 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
19860 (PRED_CONST_RETURN): Change from 69 to 65.
19861 (PRED_NULL_RETURN): Change from 91 to 71.
19862 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
19863 (PRED_LOOP_GUARD): Change from 66 to 73.
19865 2018-01-19 Martin Liska <mliska@suse.cz>
19867 * predict.c (predict_insn_def): Add new assert.
19868 (struct branch_predictor): Change type to signed integer.
19869 (test_prediction_value_range): Amend test to cover
19870 PROB_UNINITIALIZED.
19871 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
19872 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
19873 (PRED_LOOP_ITERATIONS_MAX): Likewise.
19874 (PRED_LOOP_IV_COMPARE): Likewise.
19875 * predict.h (PROB_UNINITIALIZED): Define new constant.
19877 2018-01-19 Martin Liska <mliska@suse.cz>
19879 * predict.c (dump_prediction): Add new format for
19880 analyze_brprob.py script which is enabled with -details
19882 * profile-count.h (precise_p): New function.
19884 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
19886 PR tree-optimization/83922
19887 * tree-vect-loop.c (vect_verify_full_masking): Return false if
19888 there are no statements that need masking.
19889 (vect_active_double_reduction_p): New function.
19890 (vect_analyze_loop_operations): Use it when handling phis that
19891 are not in the loop header.
19893 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
19895 PR tree-optimization/83914
19896 * tree-vect-loop.c (vectorizable_induction): Don't convert
19897 init_expr or apply the peeling adjustment for inductions
19898 that are nested within the vectorized loop.
19900 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19902 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
19905 2018-01-18 Jakub Jelinek <jakub@redhat.com>
19909 * function.h (gimplify_parameters): Add gimple_seq * argument.
19910 * function.c: Include gimple.h and options.h.
19911 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
19912 for the added local temporaries if needed.
19913 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
19914 if there are any parameter cleanups, wrap whole body into a
19915 try/finally with the cleanups.
19917 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
19920 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
19921 Use GET_MODE_CLASS for scalar floating point.
19923 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
19927 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
19928 Fix call of call_cgraph_insertion_hooks.
19930 2018-01-18 Martin Sebor <msebor@redhat.com>
19932 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
19934 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
19937 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
19940 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
19943 * common.opt: (-ffile-prefix-map): New option.
19944 * opts.c (common_handle_option): Defer it.
19945 * opts-global.c (handle_common_deferred_options): Handle it.
19946 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
19947 * file-prefix-map.h: New file.
19948 (remap_debug_filename, add_debug_prefix_map): ...here.
19949 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
19950 * final.c (debug_prefix_map, add_debug_prefix_map
19951 remap_debug_filename): Move to...
19952 * file-prefix-map.c: New file.
19953 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
19954 generalize, get rid of alloca(), use strrchr() instead of strchr().
19955 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
19956 Implement in terms of add_prefix_map().
19957 (remap_macro_filename, remap_debug_filename): Implement in term of
19959 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
19960 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
19961 * dbxout.c: Include file-prefix-map.h.
19962 * varasm.c: Likewise.
19963 * vmsdbgout.c: Likewise.
19964 * xcoffout.c: Likewise.
19965 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
19966 * doc/cppopts.texi (-fmacro-prefix-map): Document.
19967 * doc/invoke.texi (-ffile-prefix-map): Document.
19968 (-fdebug-prefix-map): Update description.
19970 2018-01-18 Martin Liska <mliska@suse.cz>
19972 * config/i386/i386.c (indirect_thunk_name): Document that also
19974 (output_indirect_thunk): Document why both instructions
19975 (pause and lfence) are generated.
19977 2018-01-18 Richard Biener <rguenther@suse.de>
19979 PR tree-optimization/83887
19980 * graphite-scop-detection.c
19981 (scop_detection::get_nearest_dom_with_single_entry): Remove.
19982 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
19983 (scop_detection::merge_sese): Re-implement with a flood-fill
19984 algorithm that properly finds a SESE region if it exists.
19986 2018-01-18 Jakub Jelinek <jakub@redhat.com>
19989 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
19990 pointer_diff optimizations use view_convert instead of convert.
19992 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19994 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
19995 Generate different code for -mno-speculate-indirect-jumps.
19996 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
19997 (*call_indirect_aix<mode>): Disable for
19998 -mno-speculate-indirect-jumps.
19999 (*call_indirect_aix<mode>_nospec): New define_insn.
20000 (*call_value_indirect_aix<mode>): Disable for
20001 -mno-speculate-indirect-jumps.
20002 (*call_value_indirect_aix<mode>_nospec): New define_insn.
20003 (*sibcall_nonlocal_sysv<mode>): Generate different code for
20004 -mno-speculate-indirect-jumps.
20005 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20007 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
20009 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
20010 long double type, set the flags for noting the default long double
20011 type, even if we don't pass or return a long double type.
20013 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
20016 * ipa-inline.c (flatten_function): Do not overwrite final inlining
20019 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
20021 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
20022 support for merge[hl].
20023 (fold_mergehl_helper): New helper function.
20024 (tree-vector-builder.h): New #include for tree_vector_builder usage.
20025 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
20026 (altivec_vmrglw_direct): Add xxmrglw insn.
20028 2018-01-17 Andrew Waterman <andrew@sifive.com>
20030 * config/riscv/riscv.c (riscv_conditional_register_usage): If
20031 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
20033 2018-01-17 David Malcolm <dmalcolm@redhat.com>
20036 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
20037 call the lto_location_cache before reading the
20038 DECL_SOURCE_LOCATION of the types.
20040 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
20041 Richard Sandiford <richard.sandiford@linaro.org>
20043 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
20044 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
20045 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
20046 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
20047 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
20049 * config/aarch64/constraints.md (aarch64_movti_operand):
20051 * config/aarch64/predicates.md (Uti): Add new constraint.
20053 2018-01-17 Carl Love <cel@us.ibm.com>
20055 * config/rs6000/vsx.md (define_expand xl_len_r,
20056 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
20057 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
20059 (define_expand, define_insn): Move the shift left from the
20060 define_insn to the define_expand for lxvl and stxvl instructions.
20061 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
20062 and XL_LEN_R definitions to PURE.
20064 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
20066 * config/i386/i386.c (indirect_thunk_name): Declare regno
20067 as unsigned int. Compare regno with INVALID_REGNUM.
20068 (output_indirect_thunk): Ditto.
20069 (output_indirect_thunk_function): Ditto.
20070 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
20071 in the call to output_indirect_thunk_function.
20073 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
20075 PR middle-end/83884
20076 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
20077 rather than the size of inner_type to determine the stack slot size
20078 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
20080 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
20083 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
20086 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
20088 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
20089 endian Linux systems to optionally enable multilibs for selecting
20090 the long double type if the user configured an explicit type.
20091 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
20092 have no long double multilibs if not defined.
20093 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
20094 warn if the user used -mabi={ieee,ibm}longdouble and we built
20095 multilibs for long double.
20096 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
20097 appropriate multilib option.
20098 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
20100 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
20101 for building long double multilibs.
20102 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
20104 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
20106 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
20109 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
20111 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
20114 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
20117 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
20120 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20122 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
20123 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
20125 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
20127 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
20128 different rtl trees depending on TARGET_64BIT.
20129 (rs6000_gen_lvx): Likewise.
20131 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
20133 * config/visium/visium.md (nop): Tweak comment.
20134 (hazard_nop): Likewise.
20136 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20138 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
20139 -mspeculate-indirect-jumps.
20140 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
20141 for -mno-speculate-indirect-jumps.
20142 (*call_indirect_elfv2<mode>_nospec): New define_insn.
20143 (*call_value_indirect_elfv2<mode>): Disable for
20144 -mno-speculate-indirect-jumps.
20145 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
20146 (indirect_jump): Emit different RTL for
20147 -mno-speculate-indirect-jumps.
20148 (*indirect_jump<mode>): Disable for
20149 -mno-speculate-indirect-jumps.
20150 (*indirect_jump<mode>_nospec): New define_insn.
20151 (tablejump): Emit different RTL for
20152 -mno-speculate-indirect-jumps.
20153 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
20154 (tablejumpsi_nospec): New define_expand.
20155 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
20156 (tablejumpdi_nospec): New define_expand.
20157 (*tablejump<mode>_internal1): Disable for
20158 -mno-speculate-indirect-jumps.
20159 (*tablejump<mode>_internal1_nospec): New define_insn.
20160 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
20163 2018-01-16 Artyom Skrobov tyomitch@gmail.com
20165 * caller-save.c (insert_save): Drop unnecessary parameter. All
20168 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20169 Richard Biener <rguenth@suse.de>
20172 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
20173 return early, inline manually is_gimple_sizepos. Make sure if we
20174 call gimplify_expr we don't end up with a gimple constant.
20175 * tree.c (variably_modified_type_p): Don't return true for
20176 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
20177 * gimplify.h (is_gimple_sizepos): Remove.
20179 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20181 PR tree-optimization/83857
20182 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
20183 vectorizable_live_operation for pure SLP statements.
20184 (vectorizable_live_operation): Handle PHIs.
20186 2018-01-16 Richard Biener <rguenther@suse.de>
20188 PR tree-optimization/83867
20189 * tree-vect-stmts.c (vect_transform_stmt): Precompute
20190 nested_in_vect_loop_p since the scalar stmt may get invalidated.
20192 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20195 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
20196 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
20197 If off is not INTEGER_CST, issue a may not be aligned warning
20198 rather than isn't aligned. Use isn%'t rather than isn't.
20199 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
20201 <case MULT_EXPR>: Improve the case when bottom and one of the
20202 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
20203 operand, in that case check if the other operand is multiple of
20204 bottom divided by the INTEGER_CST operand.
20206 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20209 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
20210 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
20211 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
20212 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
20213 * config/pa/pa.c (pa_function_arg_advance): Likewise.
20214 (pa_function_arg, pa_arg_partial_bytes): Likewise.
20215 (pa_function_arg_size): New function.
20217 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20219 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
20220 in a separate statement.
20222 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
20224 PR tree-optimization/83847
20225 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
20226 group gathers and scatters.
20228 2018-01-16 Jakub Jelinek <jakub@redhat.com>
20230 PR rtl-optimization/86620
20231 * params.def (max-sched-ready-insns): Bump minimum value to 1.
20233 PR rtl-optimization/83213
20234 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
20235 to last if both are JUMP_INSNs.
20237 PR tree-optimization/83843
20238 * gimple-ssa-store-merging.c
20239 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
20240 store_immediate_info for bswap/nop orig_stores.
20242 2018-01-15 Andrew Waterman <andrew@sifive.com>
20244 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
20246 <UDIV>: Increase cost if !TARGET_DIV.
20248 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
20250 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
20251 (define_attr "cr_logical_3op"): New.
20252 (cceq_ior_compare): Adjust.
20253 (cceq_ior_compare_complement): Adjust.
20254 (*cceq_rev_compare): Adjust.
20255 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
20256 (is_cracked_insn): Adjust.
20257 (insn_must_be_first_in_group): Adjust.
20258 * config/rs6000/40x.md: Adjust.
20259 * config/rs6000/440.md: Adjust.
20260 * config/rs6000/476.md: Adjust.
20261 * config/rs6000/601.md: Adjust.
20262 * config/rs6000/603.md: Adjust.
20263 * config/rs6000/6xx.md: Adjust.
20264 * config/rs6000/7450.md: Adjust.
20265 * config/rs6000/7xx.md: Adjust.
20266 * config/rs6000/8540.md: Adjust.
20267 * config/rs6000/cell.md: Adjust.
20268 * config/rs6000/e300c2c3.md: Adjust.
20269 * config/rs6000/e500mc.md: Adjust.
20270 * config/rs6000/e500mc64.md: Adjust.
20271 * config/rs6000/e5500.md: Adjust.
20272 * config/rs6000/e6500.md: Adjust.
20273 * config/rs6000/mpc.md: Adjust.
20274 * config/rs6000/power4.md: Adjust.
20275 * config/rs6000/power5.md: Adjust.
20276 * config/rs6000/power6.md: Adjust.
20277 * config/rs6000/power7.md: Adjust.
20278 * config/rs6000/power8.md: Adjust.
20279 * config/rs6000/power9.md: Adjust.
20280 * config/rs6000/rs64.md: Adjust.
20281 * config/rs6000/titan.md: Adjust.
20283 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20285 * config/i386/predicates.md (indirect_branch_operand): Rewrite
20286 ix86_indirect_branch_register logic.
20288 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20290 * config/i386/constraints.md (Bs): Update
20291 ix86_indirect_branch_register check. Don't check
20292 ix86_indirect_branch_register with GOT_memory_operand.
20294 * config/i386/predicates.md (GOT_memory_operand): Don't check
20295 ix86_indirect_branch_register here.
20296 (GOT32_symbol_operand): Likewise.
20298 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20300 * config/i386/predicates.md (constant_call_address_operand):
20301 Rewrite ix86_indirect_branch_register logic.
20302 (sibcall_insn_operand): Likewise.
20304 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20306 * config/i386/constraints.md (Bs): Replace
20307 ix86_indirect_branch_thunk_register with
20308 ix86_indirect_branch_register.
20310 * config/i386/i386.md (indirect_jump): Likewise.
20311 (tablejump): Likewise.
20312 (*sibcall_memory): Likewise.
20313 (*sibcall_value_memory): Likewise.
20314 Peepholes of indirect call and jump via memory: Likewise.
20315 * config/i386/i386.opt: Likewise.
20316 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20317 (GOT_memory_operand): Likewise.
20318 (call_insn_operand): Likewise.
20319 (sibcall_insn_operand): Likewise.
20320 (GOT32_symbol_operand): Likewise.
20322 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20324 PR middle-end/83837
20325 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
20326 type rather than type addr's type points to.
20327 (expand_omp_atomic_mutex): Likewise.
20328 (expand_omp_atomic): Likewise.
20330 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
20333 * config/i386/i386.c (output_indirect_thunk_function): Use
20334 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
20335 for __x86_return_thunk.
20337 2018-01-15 Richard Biener <rguenther@suse.de>
20339 PR middle-end/83850
20340 * expmed.c (extract_bit_field_1): Fix typo.
20342 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20345 * config/arm/iterators.md (VF): New mode iterator.
20346 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
20347 Remove integer-related logic from pattern.
20348 (neon_vabd<mode>_3): Likewise.
20350 2018-01-15 Jakub Jelinek <jakub@redhat.com>
20352 PR middle-end/82694
20353 * common.opt (fstrict-overflow): No longer an alias.
20354 (fwrapv-pointer): New option.
20355 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
20356 also for pointer types based on flag_wrapv_pointer.
20357 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
20358 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
20359 opts->x_flag_wrapv got set.
20360 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
20361 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
20362 POINTER_TYPE_OVERFLOW_UNDEFINED.
20363 * match.pd: Likewise in address comparison pattern.
20364 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
20366 2018-01-15 Richard Biener <rguenther@suse.de>
20369 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
20370 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
20371 Reset type names to their identifier if their TYPE_DECL doesn't
20372 have linkage (and thus is used for ODR and devirt).
20373 (save_debug_info_for_decl): Remove.
20374 (save_debug_info_for_type): Likewise.
20375 (add_tree_to_fld_list): Adjust.
20376 * tree-pretty-print.c (dump_generic_node): Make dumping of
20377 type names more robust.
20379 2018-01-15 Richard Biener <rguenther@suse.de>
20381 * BASE-VER: Bump to 8.0.1.
20383 2018-01-14 Martin Sebor <msebor@redhat.com>
20386 * builtins.c (check_access): Avoid warning when the no-warning bit
20389 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
20391 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
20392 * ira-color (allocno_hard_regs_compare): Likewise.
20394 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
20397 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
20398 Use .pushsection/.popsection.
20400 2018-01-14 Martin Sebor <msebor@redhat.com>
20403 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
20405 2018-01-14 Jakub Jelinek <jakub@redhat.com>
20407 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
20408 entry from extra_headers.
20409 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
20410 extra_headers, make the list bitwise identical to the i?86-*-* one.
20412 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20414 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
20415 -mcmodel=large with -mindirect-branch=thunk,
20416 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
20417 -mfunction-return=thunk-extern.
20418 * doc/invoke.texi: Document -mcmodel=large is incompatible with
20419 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
20420 -mfunction-return=thunk and -mfunction-return=thunk-extern.
20422 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20424 * config/i386/i386.c (print_reg): Print the name of the full
20425 integer register without '%'.
20426 (ix86_print_operand): Handle 'V'.
20427 * doc/extend.texi: Document 'V' modifier.
20429 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20431 * config/i386/constraints.md (Bs): Disallow memory operand for
20432 -mindirect-branch-register.
20434 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20435 (GOT_memory_operand): Likewise.
20436 (call_insn_operand): Likewise.
20437 (sibcall_insn_operand): Likewise.
20438 (GOT32_symbol_operand): Likewise.
20439 * config/i386/i386.md (indirect_jump): Call convert_memory_address
20440 for -mindirect-branch-register.
20441 (tablejump): Likewise.
20442 (*sibcall_memory): Likewise.
20443 (*sibcall_value_memory): Likewise.
20444 Disallow peepholes of indirect call and jump via memory for
20445 -mindirect-branch-register.
20446 (*call_pop): Replace m with Bw.
20447 (*call_value_pop): Likewise.
20448 (*sibcall_pop_memory): Replace m with Bs.
20449 * config/i386/i386.opt (mindirect-branch-register): New option.
20450 * doc/invoke.texi: Document -mindirect-branch-register option.
20452 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20454 * config/i386/i386-protos.h (ix86_output_function_return): New.
20455 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
20456 set function_return_type.
20457 (indirect_thunk_name): Add ret_p to indicate thunk for function
20459 (output_indirect_thunk_function): Pass false to
20460 indirect_thunk_name.
20461 (ix86_output_indirect_branch_via_reg): Likewise.
20462 (ix86_output_indirect_branch_via_push): Likewise.
20463 (output_indirect_thunk_function): Create alias for function
20464 return thunk if regno < 0.
20465 (ix86_output_function_return): New function.
20466 (ix86_handle_fndecl_attribute): Handle function_return.
20467 (ix86_attribute_table): Add function_return.
20468 * config/i386/i386.h (machine_function): Add
20469 function_return_type.
20470 * config/i386/i386.md (simple_return_internal): Use
20471 ix86_output_function_return.
20472 (simple_return_internal_long): Likewise.
20473 * config/i386/i386.opt (mfunction-return=): New option.
20474 (indirect_branch): Mention -mfunction-return=.
20475 * doc/extend.texi: Document function_return function attribute.
20476 * doc/invoke.texi: Document -mfunction-return= option.
20478 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
20480 * config/i386/i386-opts.h (indirect_branch): New.
20481 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
20482 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
20483 with local indirect jump when converting indirect call and jump.
20484 (ix86_set_indirect_branch_type): New.
20485 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
20486 (indirectlabelno): New.
20487 (indirect_thunk_needed): Likewise.
20488 (indirect_thunk_bnd_needed): Likewise.
20489 (indirect_thunks_used): Likewise.
20490 (indirect_thunks_bnd_used): Likewise.
20491 (INDIRECT_LABEL): Likewise.
20492 (indirect_thunk_name): Likewise.
20493 (output_indirect_thunk): Likewise.
20494 (output_indirect_thunk_function): Likewise.
20495 (ix86_output_indirect_branch_via_reg): Likewise.
20496 (ix86_output_indirect_branch_via_push): Likewise.
20497 (ix86_output_indirect_branch): Likewise.
20498 (ix86_output_indirect_jmp): Likewise.
20499 (ix86_code_end): Call output_indirect_thunk_function if needed.
20500 (ix86_output_call_insn): Call ix86_output_indirect_branch if
20502 (ix86_handle_fndecl_attribute): Handle indirect_branch.
20503 (ix86_attribute_table): Add indirect_branch.
20504 * config/i386/i386.h (machine_function): Add indirect_branch_type
20505 and has_local_indirect_jump.
20506 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
20508 (tablejump): Likewise.
20509 (*indirect_jump): Use ix86_output_indirect_jmp.
20510 (*tablejump_1): Likewise.
20511 (simple_return_indirect_internal): Likewise.
20512 * config/i386/i386.opt (mindirect-branch=): New option.
20513 (indirect_branch): New.
20516 (thunk-inline): Likewise.
20517 (thunk-extern): Likewise.
20518 * doc/extend.texi: Document indirect_branch function attribute.
20519 * doc/invoke.texi: Document -mindirect-branch= option.
20521 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
20524 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
20526 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
20528 * ipa-inline.c (want_inline_small_function_p): Return false if
20529 inlining has already failed with CIF_FINAL_ERROR.
20530 (update_caller_keys): Call want_inline_small_function_p before
20532 (update_callee_keys): Likewise.
20534 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
20536 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
20538 (rs6000_quadword_masked_address_p): Likewise.
20539 (quad_aligned_load_p): Likewise.
20540 (quad_aligned_store_p): Likewise.
20541 (const_load_sequence_p): Add comment to describe the outer-most loop.
20542 (mimic_memory_attributes_and_flags): New function.
20543 (rs6000_gen_stvx): Likewise.
20544 (replace_swapped_aligned_store): Likewise.
20545 (rs6000_gen_lvx): Likewise.
20546 (replace_swapped_aligned_load): Likewise.
20547 (replace_swapped_load_constant): Capitalize argument name in
20548 comment describing this function.
20549 (rs6000_analyze_swaps): Add a third pass to search for vector loads
20550 and stores that access quad-word aligned addresses and replace
20551 with stvx or lvx instructions when appropriate.
20552 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
20553 New function prototype.
20554 (rs6000_quadword_masked_address_p): Likewise.
20555 (rs6000_gen_lvx): Likewise.
20556 (rs6000_gen_stvx): Likewise.
20557 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
20558 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
20559 when memory address is aligned.
20560 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
20561 this split to select lvx instruction when memory address is aligned.
20562 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
20563 instruction when memory address is aligned.
20564 (*vsx_le_perm_load_v16qi): Likewise.
20565 (four unnamed splitters): Modify to select the stvx instruction
20566 when memory is aligned.
20568 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
20570 * predict.c (determine_unlikely_bbs): Handle correctly BBs
20571 which appears in the queue multiple times.
20573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20574 Alan Hayward <alan.hayward@arm.com>
20575 David Sherwood <david.sherwood@arm.com>
20577 * tree-vectorizer.h (vec_lower_bound): New structure.
20578 (_loop_vec_info): Add check_nonzero and lower_bounds.
20579 (LOOP_VINFO_CHECK_NONZERO): New macro.
20580 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
20581 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
20582 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
20583 fields. Make seg_len the distance travelled, not including the
20585 (dr_direction_indicator): Declare.
20586 (dr_zero_step_indicator): Likewise.
20587 (dr_known_forward_stride_p): Likewise.
20588 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
20590 (runtime_alias_check_p): Allow runtime alias checks with
20592 (operator ==): Compare access_size and align.
20593 (prune_runtime_alias_test_list): Rework for new distinction between
20594 the access_size and seg_len.
20595 (create_intersect_range_checks_index): Likewise. Cope with polynomial
20597 (get_segment_min_max): New function.
20598 (create_intersect_range_checks): Use it.
20599 (dr_step_indicator): New function.
20600 (dr_direction_indicator): Likewise.
20601 (dr_zero_step_indicator): Likewise.
20602 (dr_known_forward_stride_p): Likewise.
20603 * tree-loop-distribution.c (data_ref_segment_size): Return
20604 DR_STEP * (niters - 1).
20605 (compute_alias_check_pairs): Update call to the dr_with_seg_len
20607 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
20608 (vect_preserves_scalar_order_p): New function, split out from...
20609 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
20610 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
20611 (vect_vfa_access_size): New function.
20612 (vect_vfa_align): Likewise.
20613 (vect_compile_time_alias): Take access_size_a and access_b arguments.
20614 (dump_lower_bound): New function.
20615 (vect_check_lower_bound): Likewise.
20616 (vect_small_gap_p): Likewise.
20617 (vectorizable_with_step_bound_p): Likewise.
20618 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
20619 depencies if the vectorization factor is 1. Convert the checks
20620 for nonzero steps into checks on the bounds of DR_STEP. Try using
20621 a bunds check for variable steps if the minimum required step is
20622 relatively small. Update calls to the dr_with_seg_len
20623 constructor and to vect_compile_time_alias.
20624 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
20626 (vect_loop_versioning): Call it.
20627 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
20629 (vect_estimate_min_profitable_iters): Account for any bounds checks.
20631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20632 Alan Hayward <alan.hayward@arm.com>
20633 David Sherwood <david.sherwood@arm.com>
20635 * doc/sourcebuild.texi (vect_scatter_store): Document.
20636 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
20638 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
20640 * genopinit.c (main): Add supports_vec_scatter_store and
20641 supports_vec_scatter_store_cached to target_optabs.
20642 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
20643 IFN_MASK_SCATTER_STORE.
20644 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
20646 * internal-fn.h (internal_store_fn_p): Declare.
20647 (internal_fn_stored_value_index): Likewise.
20648 * internal-fn.c (scatter_store_direct): New macro.
20649 (expand_scatter_store_optab_fn): New function.
20650 (direct_scatter_store_optab_supported_p): New macro.
20651 (internal_store_fn_p): New function.
20652 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
20653 IFN_MASK_SCATTER_STORE.
20654 (internal_fn_mask_index): Likewise.
20655 (internal_fn_stored_value_index): New function.
20656 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
20657 for scatter stores.
20658 * optabs-query.h (supports_vec_scatter_store_p): Declare.
20659 * optabs-query.c (supports_vec_scatter_store_p): New function.
20660 * tree-vectorizer.h (vect_get_store_rhs): Declare.
20661 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
20662 true for scatter stores.
20663 (vect_gather_scatter_fn_p): Handle scatter stores too.
20664 (vect_check_gather_scatter): Consider using scatter stores if
20665 supports_vec_scatter_store_p.
20666 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
20667 scatter stores too.
20668 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20669 internal_fn_stored_value_index.
20670 (check_load_store_masking): Handle scatter stores too.
20671 (vect_get_store_rhs): Make public.
20672 (vectorizable_call): Use internal_store_fn_p.
20673 (vectorizable_store): Handle scatter store internal functions.
20674 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
20675 when deciding whether the end of the group has been reached.
20676 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
20677 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
20678 (mask_scatter_store<mode>): New insns.
20680 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20681 Alan Hayward <alan.hayward@arm.com>
20682 David Sherwood <david.sherwood@arm.com>
20684 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
20685 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
20686 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
20688 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
20689 Use vect_truncate_gather_scatter_offset if we can't treat the
20690 operation as a normal gather load or scatter store.
20691 (get_group_load_store_type): Take the gather_scatter_info
20692 as argument. Try using a gather load or scatter store for
20693 single-element groups.
20694 (get_load_store_type): Update calls to get_group_load_store_type
20695 and vect_use_strided_gather_scatters_p.
20697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20698 Alan Hayward <alan.hayward@arm.com>
20699 David Sherwood <david.sherwood@arm.com>
20701 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
20702 optional tree argument.
20703 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
20705 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
20706 but continue to use the current value as a fallback.
20707 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
20708 to compare the updates.
20709 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
20710 (get_load_store_type): Use it when handling a strided access.
20711 (vect_get_strided_load_store_ops): New function.
20712 (vect_get_data_ptr_increment): Likewise.
20713 (vectorizable_load): Handle strided gather loads. Always pass
20714 a step to vect_create_data_ref_ptr and bump_vector_ptr.
20716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20717 Alan Hayward <alan.hayward@arm.com>
20718 David Sherwood <david.sherwood@arm.com>
20720 * doc/md.texi (gather_load@var{m}): Document.
20721 (mask_gather_load@var{m}): Likewise.
20722 * genopinit.c (main): Add supports_vec_gather_load and
20723 supports_vec_gather_load_cached to target_optabs.
20724 * optabs-tree.c (init_tree_optimization_optabs): Use
20725 ggc_cleared_alloc to allocate target_optabs.
20726 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
20727 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
20729 * internal-fn.h (internal_load_fn_p): Declare.
20730 (internal_gather_scatter_fn_p): Likewise.
20731 (internal_fn_mask_index): Likewise.
20732 (internal_gather_scatter_fn_supported_p): Likewise.
20733 * internal-fn.c (gather_load_direct): New macro.
20734 (expand_gather_load_optab_fn): New function.
20735 (direct_gather_load_optab_supported_p): New macro.
20736 (direct_internal_fn_optab): New function.
20737 (internal_load_fn_p): Likewise.
20738 (internal_gather_scatter_fn_p): Likewise.
20739 (internal_fn_mask_index): Likewise.
20740 (internal_gather_scatter_fn_supported_p): Likewise.
20741 * optabs-query.c (supports_at_least_one_mode_p): New function.
20742 (supports_vec_gather_load_p): Likewise.
20743 * optabs-query.h (supports_vec_gather_load_p): Declare.
20744 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
20745 and memory_type field.
20746 (NUM_PATTERNS): Bump to 15.
20747 * tree-vect-data-refs.c: Include internal-fn.h.
20748 (vect_gather_scatter_fn_p): New function.
20749 (vect_describe_gather_scatter_call): Likewise.
20750 (vect_check_gather_scatter): Try using internal functions for
20751 gather loads. Recognize existing calls to a gather load function.
20752 (vect_analyze_data_refs): Consider using gather loads if
20753 supports_vec_gather_load_p.
20754 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
20755 (vect_get_gather_scatter_offset_type): Likewise.
20756 (vect_convert_mask_for_vectype): Likewise.
20757 (vect_add_conversion_to_patterm): Likewise.
20758 (vect_try_gather_scatter_pattern): Likewise.
20759 (vect_recog_gather_scatter_pattern): New pattern recognizer.
20760 (vect_vect_recog_func_ptrs): Add it.
20761 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
20762 internal_fn_mask_index and internal_gather_scatter_fn_p.
20763 (check_load_store_masking): Take the gather_scatter_info as an
20764 argument and handle gather loads.
20765 (vect_get_gather_scatter_ops): New function.
20766 (vectorizable_call): Check internal_load_fn_p.
20767 (vectorizable_load): Likewise. Handle gather load internal
20769 (vectorizable_store): Update call to check_load_store_masking.
20770 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
20771 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
20772 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
20773 (aarch64_gather_scale_operand_d): New predicates.
20774 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
20775 (mask_gather_load<mode>): New insns.
20777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20778 Alan Hayward <alan.hayward@arm.com>
20779 David Sherwood <david.sherwood@arm.com>
20781 * optabs.def (fold_left_plus_optab): New optab.
20782 * doc/md.texi (fold_left_plus_@var{m}): Document.
20783 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
20784 * internal-fn.c (fold_left_direct): Define.
20785 (expand_fold_left_optab_fn): Likewise.
20786 (direct_fold_left_optab_supported_p): Likewise.
20787 * fold-const-call.c (fold_const_fold_left): New function.
20788 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
20789 * tree-parloops.c (valid_reduction_p): New function.
20790 (gather_scalar_reductions): Use it.
20791 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
20792 (vect_finish_replace_stmt): Declare.
20793 * tree-vect-loop.c (fold_left_reduction_fn): New function.
20794 (needs_fold_left_reduction_p): New function, split out from...
20795 (vect_is_simple_reduction): ...here. Accept reductions that
20796 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
20797 (vect_force_simple_reduction): Also store the reduction type in
20798 the assignment's STMT_VINFO_REDUC_TYPE.
20799 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
20800 (merge_with_identity): New function.
20801 (vect_expand_fold_left): Likewise.
20802 (vectorize_fold_left_reduction): Likewise.
20803 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
20804 scalar phi in place for it. Check for target support and reject
20805 cases that would reassociate the operation. Defer the transform
20806 phase to vectorize_fold_left_reduction.
20807 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
20808 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
20809 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
20811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20813 * tree-if-conv.c (predicate_mem_writes): Remove redundant
20814 call to ifc_temp_var.
20816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20817 Alan Hayward <alan.hayward@arm.com>
20818 David Sherwood <david.sherwood@arm.com>
20820 * target.def (legitimize_address_displacement): Take the original
20821 offset as a poly_int.
20822 * targhooks.h (default_legitimize_address_displacement): Update
20824 * targhooks.c (default_legitimize_address_displacement): Likewise.
20825 * doc/tm.texi: Regenerate.
20826 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
20827 as an argument, moving assert of ad->disp == ad->disp_term to...
20828 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
20829 Try calling targetm.legitimize_address_displacement before expanding
20830 the address rather than afterwards, and adjust for the new interface.
20831 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
20832 Match the new hook interface. Handle SVE addresses.
20833 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
20834 new hook interface.
20836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20838 * Makefile.in (OBJS): Add early-remat.o.
20839 * target.def (select_early_remat_modes): New hook.
20840 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
20841 * doc/tm.texi: Regenerate.
20842 * targhooks.h (default_select_early_remat_modes): Declare.
20843 * targhooks.c (default_select_early_remat_modes): New function.
20844 * timevar.def (TV_EARLY_REMAT): New timevar.
20845 * passes.def (pass_early_remat): New pass.
20846 * tree-pass.h (make_pass_early_remat): Declare.
20847 * early-remat.c: New file.
20848 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
20850 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
20852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20853 Alan Hayward <alan.hayward@arm.com>
20854 David Sherwood <david.sherwood@arm.com>
20856 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
20857 vfm1 with a bound_epilog parameter.
20858 (vect_do_peeling): Update calls accordingly, and move the prologue
20859 call earlier in the function. Treat the base bound_epilog as 0 for
20860 fully-masked loops and retain vf - 1 for other loops. Add 1 to
20861 this base when peeling for gaps.
20862 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
20863 with fully-masked loops.
20864 (vect_estimate_min_profitable_iters): Handle the single peeled
20865 iteration in that case.
20867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20868 Alan Hayward <alan.hayward@arm.com>
20869 David Sherwood <david.sherwood@arm.com>
20871 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
20872 single-element interleaving even if the size is not a power of 2.
20873 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
20874 accesses for single-element interleaving if the group size is
20877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20878 Alan Hayward <alan.hayward@arm.com>
20879 David Sherwood <david.sherwood@arm.com>
20881 * doc/md.texi (fold_extract_last_@var{m}): Document.
20882 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
20883 * optabs.def (fold_extract_last_optab): New optab.
20884 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
20885 * internal-fn.c (fold_extract_direct): New macro.
20886 (expand_fold_extract_optab_fn): Likewise.
20887 (direct_fold_extract_optab_supported_p): Likewise.
20888 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
20889 * tree-vect-loop.c (vect_model_reduction_cost): Handle
20890 EXTRACT_LAST_REDUCTION.
20891 (get_initial_def_for_reduction): Do not create an initial vector
20892 for EXTRACT_LAST_REDUCTION reductions.
20893 (vectorizable_reduction): Leave the scalar phi in place for
20894 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
20895 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
20896 epilogue code for EXTRACT_LAST_REDUCTION and defer the
20897 transform phase to vectorizable_condition.
20898 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
20900 (vect_finish_stmt_generation): ...here.
20901 (vect_finish_replace_stmt): New function.
20902 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
20903 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
20905 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
20907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20908 Alan Hayward <alan.hayward@arm.com>
20909 David Sherwood <david.sherwood@arm.com>
20911 * doc/md.texi (extract_last_@var{m}): Document.
20912 * optabs.def (extract_last_optab): New optab.
20913 * internal-fn.def (EXTRACT_LAST): New internal function.
20914 * internal-fn.c (cond_unary_direct): New macro.
20915 (expand_cond_unary_optab_fn): Likewise.
20916 (direct_cond_unary_optab_supported_p): Likewise.
20917 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
20918 loops using EXTRACT_LAST.
20919 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
20920 (extract_last_<mode>): ...this optab.
20921 (vec_extract<mode><Vel>): Update accordingly.
20923 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20924 Alan Hayward <alan.hayward@arm.com>
20925 David Sherwood <david.sherwood@arm.com>
20927 * target.def (empty_mask_is_expensive): New hook.
20928 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
20929 * doc/tm.texi: Regenerate.
20930 * targhooks.h (default_empty_mask_is_expensive): Declare.
20931 * targhooks.c (default_empty_mask_is_expensive): New function.
20932 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
20933 if the target says that empty masks are expensive.
20934 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
20936 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
20938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20939 Alan Hayward <alan.hayward@arm.com>
20940 David Sherwood <david.sherwood@arm.com>
20942 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
20943 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
20944 (vect_use_loop_mask_for_alignment_p): New function.
20945 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
20946 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
20947 niters_skip argument. Make sure that the first niters_skip elements
20948 of the first iteration are inactive.
20949 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
20950 Update call to vect_set_loop_masks_directly.
20951 (get_misalign_in_elems): New function, split out from...
20952 (vect_gen_prolog_loop_niters): ...here.
20953 (vect_update_init_of_dr): Take a code argument that specifies whether
20954 the adjustment should be added or subtracted.
20955 (vect_update_init_of_drs): Likewise.
20956 (vect_prepare_for_masked_peels): New function.
20957 (vect_do_peeling): Skip prologue peeling if we're using a mask
20958 instead. Update call to vect_update_inits_of_drs.
20959 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20961 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
20962 alignment. Do not include the number of peeled iterations in
20963 the minimum threshold in that case.
20964 (vectorizable_induction): Adjust the start value down by
20965 LOOP_VINFO_MASK_SKIP_NITERS iterations.
20966 (vect_transform_loop): Call vect_prepare_for_masked_peels.
20967 Take the number of skipped iterations into account when calculating
20969 * tree-vect-stmts.c (vect_gen_while_not): New function.
20971 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20972 Alan Hayward <alan.hayward@arm.com>
20973 David Sherwood <david.sherwood@arm.com>
20975 * doc/sourcebuild.texi (vect_fully_masked): Document.
20976 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
20977 default value to 0.
20978 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
20980 (vect_analyze_loop_2): ...here. Don't check the vectorization
20981 factor against the number of loop iterations if the loop is
20984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20985 Alan Hayward <alan.hayward@arm.com>
20986 David Sherwood <david.sherwood@arm.com>
20988 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
20989 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
20990 (dump_groups): Update accordingly.
20991 (iv_use::mem_type): New member variable.
20992 (address_p): New function.
20993 (record_use): Add a mem_type argument and initialize the new
20995 (record_group_use): Add a mem_type argument. Use address_p.
20996 Remove obsolete null checks of base_object. Update call to record_use.
20997 (find_interesting_uses_op): Update call to record_group_use.
20998 (find_interesting_uses_cond): Likewise.
20999 (find_interesting_uses_address): Likewise.
21000 (get_mem_type_for_internal_fn): New function.
21001 (find_address_like_use): Likewise.
21002 (find_interesting_uses_stmt): Try find_address_like_use before
21003 calling find_interesting_uses_op.
21004 (addr_offset_valid_p): Use the iv mem_type field as the type
21005 of the addressed memory.
21006 (add_autoinc_candidates): Likewise.
21007 (get_address_cost): Likewise.
21008 (split_small_address_groups_p): Use address_p.
21009 (split_address_groups): Likewise.
21010 (add_iv_candidate_for_use): Likewise.
21011 (autoinc_possible_for_pair): Likewise.
21012 (rewrite_groups): Likewise.
21013 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
21014 (determine_group_iv_cost): Update after split of USE_ADDRESS.
21015 (get_alias_ptr_type_for_ptr_address): New function.
21016 (rewrite_use_address): Rewrite address uses in calls that were
21017 identified by find_address_like_use.
21019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21020 Alan Hayward <alan.hayward@arm.com>
21021 David Sherwood <david.sherwood@arm.com>
21023 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
21025 * gimple-expr.h (is_gimple_addressable: Likewise.
21026 * gimple-expr.c (is_gimple_address): Likewise.
21027 * internal-fn.c (expand_call_mem_ref): New function.
21028 (expand_mask_load_optab_fn): Use it.
21029 (expand_mask_store_optab_fn): Likewise.
21031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21032 Alan Hayward <alan.hayward@arm.com>
21033 David Sherwood <david.sherwood@arm.com>
21035 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
21036 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
21037 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
21038 (cond_umax@var{mode}): Document.
21039 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
21040 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
21041 (cond_umin_optab, cond_umax_optab): New optabs.
21042 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
21043 (COND_IOR, COND_XOR): New internal functions.
21044 * internal-fn.h (get_conditional_internal_fn): Declare.
21045 * internal-fn.c (cond_binary_direct): New macro.
21046 (expand_cond_binary_optab_fn): Likewise.
21047 (direct_cond_binary_optab_supported_p): Likewise.
21048 (get_conditional_internal_fn): New function.
21049 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
21050 Cope with reduction statements that are vectorized as calls rather
21052 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
21053 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
21054 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
21055 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21056 (UNSPEC_COND_EOR): New unspecs.
21057 (optab): Add mappings for them.
21058 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
21059 (sve_int_op, sve_fp_op): New int attributes.
21061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21062 Alan Hayward <alan.hayward@arm.com>
21063 David Sherwood <david.sherwood@arm.com>
21065 * optabs.def (while_ult_optab): New optab.
21066 * doc/md.texi (while_ult@var{m}@var{n}): Document.
21067 * internal-fn.def (WHILE_ULT): New internal function.
21068 * internal-fn.h (direct_internal_fn_supported_p): New override
21069 that takes two types as argument.
21070 * internal-fn.c (while_direct): New macro.
21071 (expand_while_optab_fn): New function.
21072 (convert_optab_supported_p): Likewise.
21073 (direct_while_optab_supported_p): New macro.
21074 * wide-int.h (wi::udiv_ceil): New function.
21075 * tree-vectorizer.h (rgroup_masks): New structure.
21076 (vec_loop_masks): New typedef.
21077 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
21078 and fully_masked_p.
21079 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
21080 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
21081 (vect_max_vf): New function.
21082 (slpeel_make_loop_iterate_ntimes): Delete.
21083 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
21084 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
21085 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
21086 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
21087 internal-fn.h, stor-layout.h and optabs-query.h.
21088 (vect_set_loop_mask): New function.
21089 (add_preheader_seq): Likewise.
21090 (add_header_seq): Likewise.
21091 (interleave_supported_p): Likewise.
21092 (vect_maybe_permute_loop_masks): Likewise.
21093 (vect_set_loop_masks_directly): Likewise.
21094 (vect_set_loop_condition_masked): Likewise.
21095 (vect_set_loop_condition_unmasked): New function, split out from
21096 slpeel_make_loop_iterate_ntimes.
21097 (slpeel_make_loop_iterate_ntimes): Rename to..
21098 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
21099 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
21100 (vect_do_peeling): Update call accordingly.
21101 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
21103 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21104 mask_compare_type, can_fully_mask_p and fully_masked_p.
21105 (release_vec_loop_masks): New function.
21106 (_loop_vec_info): Use it to free the loop masks.
21107 (can_produce_all_loop_masks_p): New function.
21108 (vect_get_max_nscalars_per_iter): Likewise.
21109 (vect_verify_full_masking): Likewise.
21110 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
21111 retries, and free the mask rgroups before retrying. Check loop-wide
21112 reasons for disallowing fully-masked loops. Make the final decision
21113 about whether use a fully-masked loop or not.
21114 (vect_estimate_min_profitable_iters): Do not assume that peeling
21115 for the number of iterations will be needed for fully-masked loops.
21116 (vectorizable_reduction): Disable fully-masked loops.
21117 (vectorizable_live_operation): Likewise.
21118 (vect_halve_mask_nunits): New function.
21119 (vect_double_mask_nunits): Likewise.
21120 (vect_record_loop_mask): Likewise.
21121 (vect_get_loop_mask): Likewise.
21122 (vect_transform_loop): Handle the case in which the final loop
21123 iteration might handle a partial vector. Call vect_set_loop_condition
21124 instead of slpeel_make_loop_iterate_ntimes.
21125 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
21126 (check_load_store_masking): New function.
21127 (prepare_load_store_mask): Likewise.
21128 (vectorizable_store): Handle fully-masked loops.
21129 (vectorizable_load): Likewise.
21130 (supportable_widening_operation): Use vect_halve_mask_nunits for
21132 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
21133 (vect_gen_while): New function.
21134 * config/aarch64/aarch64.md (umax<mode>3): New expander.
21135 (aarch64_uqdec<mode>): New insn.
21137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21138 Alan Hayward <alan.hayward@arm.com>
21139 David Sherwood <david.sherwood@arm.com>
21141 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
21142 (reduc_xor_scal_optab): New optabs.
21143 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
21144 (reduc_xor_scal_@var{m}): Document.
21145 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
21146 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
21147 internal functions.
21148 * fold-const-call.c (fold_const_call): Handle them.
21149 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
21150 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
21151 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
21152 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
21153 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
21154 (UNSPEC_XORV): New unspecs.
21155 (optab): Add entries for them.
21156 (BITWISEV): New int iterator.
21157 (bit_reduc_op): New int attributes.
21159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21160 Alan Hayward <alan.hayward@arm.com>
21161 David Sherwood <david.sherwood@arm.com>
21163 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
21164 * internal-fn.def (VEC_SHL_INSERT): New internal function.
21165 * optabs.def (vec_shl_insert_optab): New optab.
21166 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
21167 (duplicate_and_interleave): Likewise.
21168 * tree-vect-loop.c: Include internal-fn.h.
21169 (neutral_op_for_slp_reduction): New function, split out from
21170 get_initial_defs_for_reduction.
21171 (get_initial_def_for_reduction): Handle option 2 for variable-length
21172 vectors by loading the neutral value into a vector and then shifting
21173 the initial value into element 0.
21174 (get_initial_defs_for_reduction): Replace the code argument with
21175 the neutral value calculated by neutral_op_for_slp_reduction.
21176 Use gimple_build_vector for constant-length vectors.
21177 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
21178 but the first group_size elements have a neutral value.
21179 Use duplicate_and_interleave otherwise.
21180 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
21181 Update call to get_initial_defs_for_reduction. Handle SLP
21182 reductions for variable-length vectors by creating one vector
21183 result for each scalar result, with the elements associated
21184 with other scalar results stubbed out with the neutral value.
21185 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
21186 Require IFN_VEC_SHL_INSERT for double reductions on
21187 variable-length vectors, or SLP reductions that have
21188 a neutral value. Require can_duplicate_and_interleave_p
21189 support for variable-length unchained SLP reductions if there
21190 is no neutral value, such as for MIN/MAX reductions. Also require
21191 the number of vector elements to be a multiple of the number of
21192 SLP statements when doing variable-length unchained SLP reductions.
21193 Update call to vect_create_epilog_for_reduction.
21194 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
21195 and remove initial values.
21196 (duplicate_and_interleave): Make public.
21197 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
21198 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
21200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21201 Alan Hayward <alan.hayward@arm.com>
21202 David Sherwood <david.sherwood@arm.com>
21204 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
21205 (can_duplicate_and_interleave_p): New function.
21206 (vect_get_and_check_slp_defs): Take the vector of statements
21207 rather than just the current one. Remove excess parentheses.
21208 Restriction rejectinon of vect_constant_def and vect_external_def
21209 for variable-length vectors to boolean types, or types for which
21210 can_duplicate_and_interleave_p is false.
21211 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
21212 (duplicate_and_interleave): New function.
21213 (vect_get_constant_vectors): Use gimple_build_vector for
21214 constant-length vectors and suitable variable-length constant
21215 vectors. Use duplicate_and_interleave for other variable-length
21216 vectors. Don't defer the update when inserting new statements.
21218 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21219 Alan Hayward <alan.hayward@arm.com>
21220 David Sherwood <david.sherwood@arm.com>
21222 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
21223 min_profitable_iters doesn't go negative.
21225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21226 Alan Hayward <alan.hayward@arm.com>
21227 David Sherwood <david.sherwood@arm.com>
21229 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
21230 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
21231 * optabs.def (vec_mask_load_lanes_optab): New optab.
21232 (vec_mask_store_lanes_optab): Likewise.
21233 * internal-fn.def (MASK_LOAD_LANES): New internal function.
21234 (MASK_STORE_LANES): Likewise.
21235 * internal-fn.c (mask_load_lanes_direct): New macro.
21236 (mask_store_lanes_direct): Likewise.
21237 (expand_mask_load_optab_fn): Handle masked operations.
21238 (expand_mask_load_lanes_optab_fn): New macro.
21239 (expand_mask_store_optab_fn): Handle masked operations.
21240 (expand_mask_store_lanes_optab_fn): New macro.
21241 (direct_mask_load_lanes_optab_supported_p): Likewise.
21242 (direct_mask_store_lanes_optab_supported_p): Likewise.
21243 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
21245 (vect_load_lanes_supported): Likewise.
21246 * tree-vect-data-refs.c (strip_conversion): New function.
21247 (can_group_stmts_p): Likewise.
21248 (vect_analyze_data_ref_accesses): Use it instead of checking
21249 for a pair of assignments.
21250 (vect_store_lanes_supported): Take a masked_p parameter.
21251 (vect_load_lanes_supported): Likewise.
21252 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
21253 vect_store_lanes_supported and vect_load_lanes_supported.
21254 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
21255 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
21256 parameter. Don't allow gaps for masked accesses.
21257 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
21258 and vect_load_lanes_supported.
21259 (get_load_store_type): Take a masked_p parameter and update
21260 call to get_group_load_store_type.
21261 (vectorizable_store): Update call to get_load_store_type.
21262 Handle IFN_MASK_STORE_LANES.
21263 (vectorizable_load): Update call to get_load_store_type.
21264 Handle IFN_MASK_LOAD_LANES.
21266 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21267 Alan Hayward <alan.hayward@arm.com>
21268 David Sherwood <david.sherwood@arm.com>
21270 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
21272 * config/aarch64/aarch64-protos.h
21273 (aarch64_sve_struct_memory_operand_p): Declare.
21274 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
21275 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
21276 (VPRED, vpred): Handle SVE structure modes.
21277 * config/aarch64/constraints.md (Utx): New constraint.
21278 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
21279 (aarch64_sve_struct_nonimmediate_operand): New predicates.
21280 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
21281 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
21282 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
21283 structure modes. Split into pieces after RA.
21284 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
21285 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
21287 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
21288 SVE structure modes.
21289 (aarch64_classify_address): Likewise.
21290 (sizetochar): Move earlier in file.
21291 (aarch64_print_operand): Handle SVE register lists.
21292 (aarch64_array_mode): New function.
21293 (aarch64_sve_struct_memory_operand_p): Likewise.
21294 (TARGET_ARRAY_MODE): Redefine.
21296 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21297 Alan Hayward <alan.hayward@arm.com>
21298 David Sherwood <david.sherwood@arm.com>
21300 * target.def (array_mode): New target hook.
21301 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
21302 * doc/tm.texi: Regenerate.
21303 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
21304 * hooks.c (hook_optmode_mode_uhwi_none): New function.
21305 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
21306 targetm.array_mode.
21307 * stor-layout.c (mode_for_array): Likewise. Support polynomial
21310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21311 Alan Hayward <alan.hayward@arm.com>
21312 David Sherwood <david.sherwood@arm.com>
21314 * fold-const.c (fold_binary_loc): Check the argument types
21315 rather than the result type when testing for a vector operation.
21317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21319 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
21320 * doc/tm.texi: Regenerate.
21322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21323 Alan Hayward <alan.hayward@arm.com>
21324 David Sherwood <david.sherwood@arm.com>
21326 * doc/invoke.texi (-msve-vector-bits=): Document new option.
21327 (sve): Document new AArch64 extension.
21328 * doc/md.texi (w): Extend the description of the AArch64
21329 constraint to include SVE vectors.
21330 (Upl, Upa): Document new AArch64 predicate constraints.
21331 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
21333 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
21334 (msve-vector-bits=): New option.
21335 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
21336 SVE when these are disabled.
21337 (sve): New extension.
21338 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
21339 modes. Adjust their number of units based on aarch64_sve_vg.
21340 (MAX_BITSIZE_MODE_ANY_MODE): Define.
21341 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
21342 aarch64_addr_query_type.
21343 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
21344 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
21345 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
21346 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
21347 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
21348 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
21349 (aarch64_simd_imm_zero_p): Delete.
21350 (aarch64_check_zero_based_sve_index_immediate): Declare.
21351 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21352 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21353 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21354 (aarch64_sve_float_mul_immediate_p): Likewise.
21355 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21356 rather than an rtx.
21357 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
21358 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
21359 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
21360 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
21361 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
21362 (aarch64_regmode_natural_size): Likewise.
21363 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
21364 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
21366 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
21367 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
21368 for VG and the SVE predicate registers.
21369 (V_ALIASES): Add a "z"-prefixed alias.
21370 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
21371 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
21372 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
21373 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
21374 (REG_CLASS_NAMES): Add entries for them.
21375 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
21376 and the predicate registers.
21377 (aarch64_sve_vg): Declare.
21378 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
21379 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
21380 (REGMODE_NATURAL_SIZE): Define.
21381 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
21383 * config/aarch64/aarch64.c: Include cfgrtl.h.
21384 (simd_immediate_info): Add a constructor for series vectors,
21385 and an associated step field.
21386 (aarch64_sve_vg): New variable.
21387 (aarch64_dbx_register_number): Handle VG and the predicate registers.
21388 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
21389 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
21390 (VEC_ANY_DATA, VEC_STRUCT): New constants.
21391 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
21392 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
21393 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
21394 (aarch64_get_mask_mode): New functions.
21395 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
21396 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
21397 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
21398 predicate modes and predicate registers. Explicitly restrict
21399 GPRs to modes of 16 bytes or smaller. Only allow FP registers
21400 to store a vector mode if it is recognized by
21401 aarch64_classify_vector_mode.
21402 (aarch64_regmode_natural_size): New function.
21403 (aarch64_hard_regno_caller_save_mode): Return the original mode
21405 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
21406 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
21407 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
21408 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
21410 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
21411 does not overlap dest if the function is frame-related. Handle
21413 (aarch64_split_add_offset): New function.
21414 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
21415 them aarch64_add_offset.
21416 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
21417 and update call to aarch64_sub_sp.
21418 (aarch64_add_cfa_expression): New function.
21419 (aarch64_expand_prologue): Pass extra temporary registers to the
21420 functions above. Handle the case in which we need to emit new
21421 DW_CFA_expressions for registers that were originally saved
21422 relative to the stack pointer, but now have to be expressed
21423 relative to the frame pointer.
21424 (aarch64_output_mi_thunk): Pass extra temporary registers to the
21426 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
21427 IP0 and IP1 values for SVE frames.
21428 (aarch64_expand_vec_series): New function.
21429 (aarch64_expand_sve_widened_duplicate): Likewise.
21430 (aarch64_expand_sve_const_vector): Likewise.
21431 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
21432 Handle SVE constants. Use emit_move_insn to move a force_const_mem
21433 into the register, rather than emitting a SET directly.
21434 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
21435 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
21436 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
21437 (offset_9bit_signed_scaled_p): New functions.
21438 (aarch64_replicate_bitmask_imm): New function.
21439 (aarch64_bitmask_imm): Use it.
21440 (aarch64_cannot_force_const_mem): Reject expressions involving
21441 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
21442 (aarch64_classify_index): Handle SVE indices, by requiring
21443 a plain register index with a scale that matches the element size.
21444 (aarch64_classify_address): Handle SVE addresses. Assert that
21445 the mode of the address is VOIDmode or an integer mode.
21446 Update call to aarch64_classify_symbol.
21447 (aarch64_classify_symbolic_expression): Update call to
21448 aarch64_classify_symbol.
21449 (aarch64_const_vec_all_in_range_p): New function.
21450 (aarch64_print_vector_float_operand): Likewise.
21451 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
21452 "vN" for FP registers with SVE modes. Handle (const ...) vectors
21453 and the FP immediates 1.0 and 0.5.
21454 (aarch64_print_address_internal): Handle SVE addresses.
21455 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
21456 (aarch64_regno_regclass): Handle predicate registers.
21457 (aarch64_secondary_reload): Handle big-endian reloads of SVE
21459 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
21460 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
21461 (aarch64_convert_sve_vector_bits): New function.
21462 (aarch64_override_options): Use it to handle -msve-vector-bits=.
21463 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
21464 rather than an rtx.
21465 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
21466 Handle SVE vector and predicate modes. Accept VL-based constants
21467 that need only one temporary register, and VL offsets that require
21468 no temporary registers.
21469 (aarch64_conditional_register_usage): Mark the predicate registers
21470 as fixed if SVE isn't available.
21471 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
21472 Return true for SVE vector and predicate modes.
21473 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
21474 rather than an unsigned int. Handle SVE modes.
21475 (aarch64_preferred_simd_mode): Update call accordingly. Handle
21477 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
21479 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
21480 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
21481 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
21482 (aarch64_sve_float_mul_immediate_p): New functions.
21483 (aarch64_sve_valid_immediate): New function.
21484 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
21485 Explicitly reject structure modes. Check for INDEX constants.
21486 Handle PTRUE and PFALSE constants.
21487 (aarch64_check_zero_based_sve_index_immediate): New function.
21488 (aarch64_simd_imm_zero_p): Delete.
21489 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
21490 vector modes. Accept constants in the range of CNT[BHWD].
21491 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
21492 ask for an Advanced SIMD mode.
21493 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
21494 (aarch64_simd_vector_alignment): Handle SVE predicates.
21495 (aarch64_vectorize_preferred_vector_alignment): New function.
21496 (aarch64_simd_vector_alignment_reachable): Use it instead of
21498 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
21499 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
21501 (MAX_VECT_LEN): Delete.
21502 (expand_vec_perm_d): Add a vec_flags field.
21503 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
21504 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
21505 (aarch64_evpc_ext): Don't apply a big-endian lane correction
21507 (aarch64_evpc_rev): Rename to...
21508 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
21509 (aarch64_evpc_rev_global): New function.
21510 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
21511 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
21513 (aarch64_evpc_sve_tbl): New function.
21514 (aarch64_expand_vec_perm_const_1): Update after rename of
21515 aarch64_evpc_rev. Handle SVE permutes too, trying
21516 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
21517 than aarch64_evpc_tbl.
21518 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
21519 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
21520 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
21521 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
21522 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
21523 (aarch64_expand_sve_vcond): New functions.
21524 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
21525 of aarch64_vector_mode_p.
21526 (aarch64_dwarf_poly_indeterminate_value): New function.
21527 (aarch64_compute_pressure_classes): Likewise.
21528 (aarch64_can_change_mode_class): Likewise.
21529 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
21530 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
21531 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
21532 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
21533 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
21534 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
21535 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
21536 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
21538 (Dn, Dl, Dr): Accept const as well as const_vector.
21539 (Dz): Likewise. Compare against CONST0_RTX.
21540 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
21541 of "vector" where appropriate.
21542 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
21543 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
21544 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
21545 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
21546 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
21547 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
21548 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
21549 (v_int_equiv): Extend to SVE modes.
21550 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
21552 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
21553 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
21554 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
21555 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
21556 (SVE_COND_FP_CMP): New int iterators.
21557 (perm_hilo): Handle the new unpack unspecs.
21558 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
21560 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
21561 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
21562 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
21563 (aarch64_equality_operator, aarch64_constant_vector_operand)
21564 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
21565 (aarch64_sve_nonimmediate_operand): Likewise.
21566 (aarch64_sve_general_operand): Likewise.
21567 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
21568 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
21569 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
21570 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
21571 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
21572 (aarch64_sve_float_arith_immediate): Likewise.
21573 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
21574 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
21575 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
21576 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
21577 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
21578 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
21579 (aarch64_sve_float_arith_operand): Likewise.
21580 (aarch64_sve_float_arith_with_sub_operand): Likewise.
21581 (aarch64_sve_float_mul_operand): Likewise.
21582 (aarch64_sve_vec_perm_operand): Likewise.
21583 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
21584 (aarch64_mov_operand): Accept const_poly_int and const_vector.
21585 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
21586 as well as const_vector.
21587 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
21588 in file. Use CONST0_RTX and CONSTM1_RTX.
21589 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
21590 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
21591 Use aarch64_simd_imm_zero.
21592 * config/aarch64/aarch64-sve.md: New file.
21593 * config/aarch64/aarch64.md: Include it.
21594 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
21595 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
21596 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
21597 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
21598 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
21599 (sve): New attribute.
21600 (enabled): Disable instructions with the sve attribute unless
21602 (movqi, movhi): Pass CONST_POLY_INT operaneds through
21603 aarch64_expand_mov_immediate.
21604 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
21605 CNT[BHSD] immediates.
21606 (movti): Split CONST_POLY_INT moves into two halves.
21607 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
21608 Split additions that need a temporary here if the destination
21609 is the stack pointer.
21610 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
21611 (*add<mode>3_poly_1): New instruction.
21612 (set_clobber_cc): New expander.
21614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21616 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
21617 parameter and use it instead of GET_MODE_SIZE (innermode). Use
21618 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
21619 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
21620 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
21621 Change innermode from fixed_mode_size to machine_mode.
21622 (simplify_subreg): Update call accordingly. Handle a constant-sized
21623 subreg of a variable-length CONST_VECTOR.
21625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21626 Alan Hayward <alan.hayward@arm.com>
21627 David Sherwood <david.sherwood@arm.com>
21629 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
21630 (add_offset_to_base): New function, split out from...
21631 (create_mem_ref): ...here. When handling a scale other than 1,
21632 check first whether the address is valid without the offset.
21633 Add it into the base if so, leaving the index and scale as-is.
21635 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21638 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
21639 fold_for_warn before checking if arg2 is INTEGER_CST.
21641 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
21643 * config/rs6000/predicates.md (load_multiple_operation): Delete.
21644 (store_multiple_operation): Delete.
21645 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
21646 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
21647 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
21648 guarded by TARGET_STRING.
21649 (rs6000_output_load_multiple): Delete.
21650 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
21651 OPTION_MASK_STRING / TARGET_STRING handling.
21652 (print_operand) <'N', 'O'>: Add comment that these are unused now.
21653 (const rs6000_opt_masks) <"string">: Change mask to 0.
21654 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
21655 (MASK_STRING): Delete.
21656 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
21658 (load_multiple): Delete.
21665 (store_multiple): Delete.
21672 (movmemsi_8reg): Delete.
21673 (corresponding unnamed define_insn): Delete.
21674 (movmemsi_6reg): Delete.
21675 (corresponding unnamed define_insn): Delete.
21676 (movmemsi_4reg): Delete.
21677 (corresponding unnamed define_insn): Delete.
21678 (movmemsi_2reg): Delete.
21679 (corresponding unnamed define_insn): Delete.
21680 (movmemsi_1reg): Delete.
21681 (corresponding unnamed define_insn): Delete.
21682 * config/rs6000/rs6000.opt (mno-string): New.
21683 (mstring): Replace by deprecation warning stub.
21684 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
21686 2018-01-12 Jakub Jelinek <jakub@redhat.com>
21688 * regrename.c (regrename_do_replace): If replacing the same
21689 reg multiple times, try to reuse last created gen_raw_REG.
21692 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
21693 main to workaround a bug in GDB.
21695 2018-01-12 Tom de Vries <tom@codesourcery.com>
21698 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
21700 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
21702 PR rtl-optimization/80481
21703 * ira-color.c (get_cap_member): New function.
21704 (allocnos_conflict_by_live_ranges_p): Use it.
21705 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
21706 (setup_slot_coalesced_allocno_live_ranges): Ditto.
21708 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
21711 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
21712 (*saddl_se_1): Ditto.
21713 (*ssubsi_1): Ditto.
21714 (*ssubl_se_1): Ditto.
21716 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21718 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
21719 rather than wi::to_widest for DR_INITs.
21720 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
21721 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
21722 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
21724 (vect_analyze_group_access_1): Note that here.
21726 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21728 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
21729 polynomial type sizes.
21731 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
21733 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
21734 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
21735 (gimple_add_tmp_var): Likewise.
21737 2018-01-12 Martin Liska <mliska@suse.cz>
21739 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
21740 (gimple_alloc_sizes): Likewise.
21741 (dump_gimple_statistics): Use PRIu64 in printf format.
21742 * gimple.h: Change uint64_t to int.
21744 2018-01-12 Martin Liska <mliska@suse.cz>
21746 * tree-core.h: Use uint64_t instead of int.
21747 * tree.c (tree_node_counts): Likewise.
21748 (tree_node_sizes): Likewise.
21749 (dump_tree_statistics): Use PRIu64 in printf format.
21751 2018-01-12 Martin Liska <mliska@suse.cz>
21753 * Makefile.in: As qsort_chk is implemented in vec.c, add
21754 vec.o to linkage of gencfn-macros.
21755 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
21756 passing the info to record_node_allocation_statistics.
21757 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
21759 * ggc-common.c (struct ggc_usage): Add operator== and use
21760 it in operator< and compare function.
21761 * mem-stats.h (struct mem_usage): Likewise.
21762 * vec.c (struct vec_usage): Remove operator< and compare
21763 function. Can be simply inherited.
21765 2018-01-12 Martin Jambor <mjambor@suse.cz>
21768 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
21769 * tree-ssa-math-opts.c: Include domwalk.h.
21770 (convert_mult_to_fma_1): New function.
21771 (fma_transformation_info): New type.
21772 (fma_deferring_state): Likewise.
21773 (cancel_fma_deferring): New function.
21774 (result_of_phi): Likewise.
21775 (last_fma_candidate_feeds_initial_phi): Likewise.
21776 (convert_mult_to_fma): Added deferring logic, split actual
21777 transformation to convert_mult_to_fma_1.
21778 (math_opts_dom_walker): New type.
21779 (math_opts_dom_walker::after_dom_children): New method, body moved
21780 here from pass_optimize_widening_mul::execute, added deferring logic
21782 (pass_optimize_widening_mul::execute): Moved most of code to
21783 math_opts_dom_walker::after_dom_children.
21784 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
21785 * config/i386/i386.c (ix86_option_override_internal): Added
21786 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
21788 2018-01-12 Richard Biener <rguenther@suse.de>
21791 * dwarf2out.c (gen_variable_die): Do not reset old_die for
21792 inline instance vars.
21794 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
21797 * config/rx/rx.c (rx_is_restricted_memory_address):
21798 Handle SUBREG case.
21800 2018-01-12 Richard Biener <rguenther@suse.de>
21802 PR tree-optimization/80846
21803 * target.def (split_reduction): New target hook.
21804 * targhooks.c (default_split_reduction): New function.
21805 * targhooks.h (default_split_reduction): Declare.
21806 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
21807 target requests first reduce vectors by combining low and high
21809 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
21810 (get_vectype_for_scalar_type_and_size): Export.
21811 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
21812 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
21813 * doc/tm.texi: Regenerate.
21814 * config/i386/i386.c (ix86_split_reduction): Implement
21815 TARGET_VECTORIZE_SPLIT_REDUCTION.
21817 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21820 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
21821 in PIC mode except for TARGET_VXWORKS_RTP.
21822 * config/sparc/sparc.c: Include cfgrtl.h.
21823 (TARGET_INIT_PIC_REG): Define.
21824 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
21825 (sparc_pic_register_p): New predicate.
21826 (sparc_legitimate_address_p): Use it.
21827 (sparc_legitimize_pic_address): Likewise.
21828 (sparc_delegitimize_address): Likewise.
21829 (sparc_mode_dependent_address_p): Likewise.
21830 (gen_load_pcrel_sym): Remove 4th parameter.
21831 (load_got_register): Adjust call to above. Remove obsolete stuff.
21832 (sparc_expand_prologue): Do not call load_got_register here.
21833 (sparc_flat_expand_prologue): Likewise.
21834 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
21835 (sparc_use_pseudo_pic_reg): New function.
21836 (sparc_init_pic_reg): Likewise.
21837 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21838 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
21840 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
21842 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
21843 Add item for branch_cost.
21845 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21847 PR rtl-optimization/83565
21848 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
21849 not extend the result to a larger mode for rotate operations.
21850 (num_sign_bit_copies1): Likewise.
21852 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21855 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
21857 Use values-Xc.o for -pedantic.
21858 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
21860 2018-01-12 Martin Liska <mliska@suse.cz>
21863 * ipa-devirt.c (final_warning_record::grow_type_warnings):
21865 (possible_polymorphic_call_targets): Use it.
21866 (ipa_devirt): Likewise.
21868 2018-01-12 Martin Liska <mliska@suse.cz>
21870 * profile-count.h (enum profile_quality): Use 0 as invalid
21871 enum value of profile_quality.
21873 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
21875 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
21876 -mext-string options.
21878 2018-01-12 Richard Biener <rguenther@suse.de>
21880 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
21881 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
21882 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21884 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21886 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
21888 * configure.ac (--with-long-double-format): Add support for the
21889 configuration option to change the default long double format on
21891 * config.gcc (powerpc*-linux*-*): Likewise.
21892 * configure: Regenerate.
21893 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
21894 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
21895 used without modification.
21897 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21899 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
21900 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
21901 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
21902 MISC_BUILTIN_SPEC_BARRIER.
21903 (rs6000_init_builtins): Likewise.
21904 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
21906 (speculation_barrier): New define_insn.
21907 * doc/extend.texi: Document __builtin_speculation_barrier.
21909 2018-01-11 Jakub Jelinek <jakub@redhat.com>
21912 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
21913 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
21914 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
21916 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
21917 integral modes instead of "ss" and "sd".
21918 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
21919 vectors with 32-bit and 64-bit elements.
21920 (vecdupssescalarmodesuffix): New mode attribute.
21921 (vec_dup<mode>): Use it.
21923 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
21926 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
21927 frame if argument is passed on stack.
21929 2018-01-11 Jakub Jelinek <jakub@redhat.com>
21932 * ree.c (combine_reaching_defs): Optimize also
21933 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
21934 reg2=any_extend(exp); reg1=reg2;, formatting fix.
21936 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
21938 PR middle-end/83189
21939 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
21941 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
21943 PR middle-end/83718
21944 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
21945 after they are computed.
21947 2018-01-11 Bin Cheng <bin.cheng@arm.com>
21949 PR tree-optimization/83695
21950 * gimple-loop-linterchange.cc
21951 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
21952 reset cached scev information after interchange.
21953 (pass_linterchange::execute): Remove call to scev_reset_htab.
21955 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21957 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
21958 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
21959 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
21960 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
21961 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
21962 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
21963 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
21964 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
21965 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
21966 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
21967 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
21968 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
21969 (V_lane_reg): Likewise.
21970 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
21972 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
21973 (vfmal_lane_low<mode>_intrinsic,
21974 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
21975 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
21976 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
21977 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
21978 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
21979 vfmsl_lane_high<mode>_intrinsic): New define_insns.
21981 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21983 * config/arm/arm-cpus.in (fp16fml): New feature.
21984 (ALL_SIMD): Add fp16fml.
21985 (armv8.2-a): Add fp16fml as an option.
21986 (armv8.3-a): Likewise.
21987 (armv8.4-a): Add fp16fml as part of fp16.
21988 * config/arm/arm.h (TARGET_FP16FML): Define.
21989 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
21991 * config/arm/arm-modes.def (V2HF): Define.
21992 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
21993 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
21994 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
21995 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
21996 vfmsl_low, vfmsl_high): New set of builtins.
21997 * config/arm/iterators.md (PLUSMINUS): New code iterator.
21998 (vfml_op): New code attribute.
21999 (VFMLHALVES): New int iterator.
22000 (VFML, VFMLSEL): New mode attributes.
22001 (V_reg): Define mapping for V2HF.
22002 (V_hi, V_lo): New mode attributes.
22003 (VF_constraint): Likewise.
22004 (vfml_half, vfml_half_selector): New int attributes.
22005 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
22007 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
22008 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
22010 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
22011 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
22012 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
22013 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
22015 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
22016 Document new effective target and option set.
22018 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22020 * config/arm/arm-cpus.in (armv8_4): New feature.
22021 (ARMv8_4a): New fgroup.
22022 (armv8.4-a): New arch.
22023 * config/arm/arm-tables.opt: Regenerate.
22024 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
22025 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
22026 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
22027 Add matching rules for -march=armv8.4-a and extensions.
22028 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
22030 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
22033 * config/rx/rx.md (BW): New mode attribute.
22034 (sync_lock_test_and_setsi): Add mode suffix to insn output.
22036 2018-01-11 Richard Biener <rguenther@suse.de>
22038 PR tree-optimization/83435
22039 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
22040 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
22041 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
22043 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22044 Alan Hayward <alan.hayward@arm.com>
22045 David Sherwood <david.sherwood@arm.com>
22047 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
22049 (aarch64_classify_address): Initialize it. Track polynomial offsets.
22050 (aarch64_print_address_internal): Use it to check for a zero offset.
22052 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22053 Alan Hayward <alan.hayward@arm.com>
22054 David Sherwood <david.sherwood@arm.com>
22056 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
22057 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
22058 Return a poly_int64 rather than a HOST_WIDE_INT.
22059 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
22060 rather than a HOST_WIDE_INT.
22061 * config/aarch64/aarch64.h (aarch64_frame): Protect with
22062 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
22063 hard_fp_offset, frame_size, initial_adjust, callee_offset and
22064 final_offset from HOST_WIDE_INT to poly_int64.
22065 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
22066 to_constant when getting the number of units in an Advanced SIMD
22068 (aarch64_builtin_vectorized_function): Check for a constant number
22070 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
22072 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
22073 attribute instead of GET_MODE_NUNITS.
22074 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
22075 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
22076 GET_MODE_SIZE for fixed-size registers.
22077 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
22078 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
22079 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
22080 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
22081 (aarch64_print_operand, aarch64_print_address_internal)
22082 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
22083 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
22084 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
22085 Handle polynomial GET_MODE_SIZE.
22086 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
22087 wider than SImode without modification.
22088 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
22089 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
22090 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
22091 passing and returning SVE modes.
22092 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
22093 rather than GEN_INT.
22094 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
22095 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
22096 (aarch64_allocate_and_probe_stack_space): Likewise.
22097 (aarch64_layout_frame): Cope with polynomial offsets.
22098 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
22099 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
22100 polynomial offsets.
22101 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
22102 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
22103 poly_int64 rather than a HOST_WIDE_INT.
22104 (aarch64_get_separate_components, aarch64_process_components)
22105 (aarch64_expand_prologue, aarch64_expand_epilogue)
22106 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
22107 (aarch64_anchor_offset): New function, split out from...
22108 (aarch64_legitimize_address): ...here.
22109 (aarch64_builtin_vectorization_cost): Handle polynomial
22110 TYPE_VECTOR_SUBPARTS.
22111 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
22113 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
22114 number of elements from the PARALLEL rather than the mode.
22115 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
22116 rather than GET_MODE_BITSIZE.
22117 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
22118 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
22119 (aarch64_expand_vec_perm_const_1): Handle polynomial
22120 d->perm.length () and d->perm elements.
22121 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
22122 Apply to_constant to d->perm elements.
22123 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
22124 polynomial CONST_VECTOR_NUNITS.
22125 (aarch64_move_pointer): Take amount as a poly_int64 rather
22127 (aarch64_progress_pointer): Avoid temporary variable.
22128 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
22129 the mode attribute instead of GET_MODE.
22131 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22132 Alan Hayward <alan.hayward@arm.com>
22133 David Sherwood <david.sherwood@arm.com>
22135 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
22136 x exists before using it.
22137 (aarch64_add_constant_internal): Rename to...
22138 (aarch64_add_offset_1): ...this. Replace regnum with separate
22139 src and dest rtxes. Handle the case in which they're different,
22140 including when the offset is zero. Replace scratchreg with an rtx.
22141 Use 2 additions if there is no spare register into which we can
22142 move a 16-bit constant.
22143 (aarch64_add_constant): Delete.
22144 (aarch64_add_offset): Replace reg with separate src and dest
22145 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
22146 Use aarch64_add_offset_1.
22147 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
22148 an rtx rather than an int. Take the delta as a poly_int64
22149 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
22150 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
22151 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
22152 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
22153 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
22154 and aarch64_add_sp.
22155 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
22156 aarch64_add_constant.
22158 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22160 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22161 Use scalar_float_mode.
22163 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22165 * config/aarch64/aarch64-simd.md
22166 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
22167 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
22168 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
22169 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
22170 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
22171 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
22172 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
22173 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
22174 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
22175 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
22177 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22180 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
22181 targ_options->x_arm_arch_string is non NULL.
22183 2018-01-11 Tamar Christina <tamar.christina@arm.com>
22185 * config/aarch64/aarch64.h
22186 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
22188 2018-01-11 Sudakshina Das <sudi.das@arm.com>
22191 * expmed.c (emit_store_flag_force): Swap if const op0
22192 and change VOIDmode to mode of op0.
22194 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
22196 PR rtl-optimization/83761
22197 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
22198 than bytes to mode_for_size.
22200 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22202 PR middle-end/83189
22203 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
22204 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
22207 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22209 PR middle-end/83575
22210 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
22211 when in layout mode.
22212 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
22213 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
22216 2018-01-10 Michael Collison <michael.collison@arm.com>
22218 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
22219 * config/aarch64/aarch64-option-extension.def: Add
22220 AARCH64_OPT_EXTENSION of 'fp16fml'.
22221 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22222 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
22223 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
22224 * config/aarch64/constraints.md (Ui7): New constraint.
22225 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
22226 (VFMLA_SEL_W): Ditto.
22229 (VFMLA16_LOW): New int iterator.
22230 (VFMLA16_HIGH): Ditto.
22231 (UNSPEC_FMLAL): New unspec.
22232 (UNSPEC_FMLSL): Ditto.
22233 (UNSPEC_FMLAL2): Ditto.
22234 (UNSPEC_FMLSL2): Ditto.
22235 (f16mac): New code attribute.
22236 * config/aarch64/aarch64-simd-builtins.def
22237 (aarch64_fmlal_lowv2sf): Ditto.
22238 (aarch64_fmlsl_lowv2sf): Ditto.
22239 (aarch64_fmlalq_lowv4sf): Ditto.
22240 (aarch64_fmlslq_lowv4sf): Ditto.
22241 (aarch64_fmlal_highv2sf): Ditto.
22242 (aarch64_fmlsl_highv2sf): Ditto.
22243 (aarch64_fmlalq_highv4sf): Ditto.
22244 (aarch64_fmlslq_highv4sf): Ditto.
22245 (aarch64_fmlal_lane_lowv2sf): Ditto.
22246 (aarch64_fmlsl_lane_lowv2sf): Ditto.
22247 (aarch64_fmlal_laneq_lowv2sf): Ditto.
22248 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
22249 (aarch64_fmlalq_lane_lowv4sf): Ditto.
22250 (aarch64_fmlsl_lane_lowv4sf): Ditto.
22251 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
22252 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
22253 (aarch64_fmlal_lane_highv2sf): Ditto.
22254 (aarch64_fmlsl_lane_highv2sf): Ditto.
22255 (aarch64_fmlal_laneq_highv2sf): Ditto.
22256 (aarch64_fmlsl_laneq_highv2sf): Ditto.
22257 (aarch64_fmlalq_lane_highv4sf): Ditto.
22258 (aarch64_fmlsl_lane_highv4sf): Ditto.
22259 (aarch64_fmlalq_laneq_highv4sf): Ditto.
22260 (aarch64_fmlsl_laneq_highv4sf): Ditto.
22261 * config/aarch64/aarch64-simd.md:
22262 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
22263 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22264 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
22265 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
22266 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
22267 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
22268 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
22269 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
22270 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
22271 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
22272 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
22273 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
22274 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
22275 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
22276 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
22277 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
22278 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
22279 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
22280 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
22281 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
22282 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
22283 (vfmlsl_low_u32): Ditto.
22284 (vfmlalq_low_u32): Ditto.
22285 (vfmlslq_low_u32): Ditto.
22286 (vfmlal_high_u32): Ditto.
22287 (vfmlsl_high_u32): Ditto.
22288 (vfmlalq_high_u32): Ditto.
22289 (vfmlslq_high_u32): Ditto.
22290 (vfmlal_lane_low_u32): Ditto.
22291 (vfmlsl_lane_low_u32): Ditto.
22292 (vfmlal_laneq_low_u32): Ditto.
22293 (vfmlsl_laneq_low_u32): Ditto.
22294 (vfmlalq_lane_low_u32): Ditto.
22295 (vfmlslq_lane_low_u32): Ditto.
22296 (vfmlalq_laneq_low_u32): Ditto.
22297 (vfmlslq_laneq_low_u32): Ditto.
22298 (vfmlal_lane_high_u32): Ditto.
22299 (vfmlsl_lane_high_u32): Ditto.
22300 (vfmlal_laneq_high_u32): Ditto.
22301 (vfmlsl_laneq_high_u32): Ditto.
22302 (vfmlalq_lane_high_u32): Ditto.
22303 (vfmlslq_lane_high_u32): Ditto.
22304 (vfmlalq_laneq_high_u32): Ditto.
22305 (vfmlslq_laneq_high_u32): Ditto.
22306 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
22307 (AARCH64_FL_FOR_ARCH8_4): New.
22308 (AARCH64_ISA_F16FML): New ISA flag.
22309 (TARGET_F16FML): New feature flag for fp16fml.
22310 (doc/invoke.texi): Document new fp16fml option.
22312 2018-01-10 Michael Collison <michael.collison@arm.com>
22314 * config/aarch64/aarch64-builtins.c:
22315 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
22316 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22317 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
22318 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
22319 (AARCH64_ISA_SHA3): New ISA flag.
22320 (TARGET_SHA3): New feature flag for sha3.
22321 * config/aarch64/iterators.md (sha512_op): New int attribute.
22322 (CRYPTO_SHA512): New int iterator.
22323 (UNSPEC_SHA512H): New unspec.
22324 (UNSPEC_SHA512H2): Ditto.
22325 (UNSPEC_SHA512SU0): Ditto.
22326 (UNSPEC_SHA512SU1): Ditto.
22327 * config/aarch64/aarch64-simd-builtins.def
22328 (aarch64_crypto_sha512hqv2di): New builtin.
22329 (aarch64_crypto_sha512h2qv2di): Ditto.
22330 (aarch64_crypto_sha512su0qv2di): Ditto.
22331 (aarch64_crypto_sha512su1qv2di): Ditto.
22332 (aarch64_eor3qv8hi): Ditto.
22333 (aarch64_rax1qv2di): Ditto.
22334 (aarch64_xarqv2di): Ditto.
22335 (aarch64_bcaxqv8hi): Ditto.
22336 * config/aarch64/aarch64-simd.md:
22337 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
22338 (aarch64_crypto_sha512su0qv2di): Ditto.
22339 (aarch64_crypto_sha512su1qv2di): Ditto.
22340 (aarch64_eor3qv8hi): Ditto.
22341 (aarch64_rax1qv2di): Ditto.
22342 (aarch64_xarqv2di): Ditto.
22343 (aarch64_bcaxqv8hi): Ditto.
22344 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
22345 (vsha512h2q_u64): Ditto.
22346 (vsha512su0q_u64): Ditto.
22347 (vsha512su1q_u64): Ditto.
22348 (veor3q_u16): Ditto.
22349 (vrax1q_u64): Ditto.
22350 (vxarq_u64): Ditto.
22351 (vbcaxq_u16): Ditto.
22352 * config/arm/types.md (crypto_sha512): New type attribute.
22353 (crypto_sha3): Ditto.
22354 (doc/invoke.texi): Document new sha3 option.
22356 2018-01-10 Michael Collison <michael.collison@arm.com>
22358 * config/aarch64/aarch64-builtins.c:
22359 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
22360 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22361 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
22362 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
22363 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
22364 (AARCH64_ISA_SM4): New ISA flag.
22365 (TARGET_SM4): New feature flag for sm4.
22366 * config/aarch64/aarch64-simd-builtins.def
22367 (aarch64_sm3ss1qv4si): Ditto.
22368 (aarch64_sm3tt1aq4si): Ditto.
22369 (aarch64_sm3tt1bq4si): Ditto.
22370 (aarch64_sm3tt2aq4si): Ditto.
22371 (aarch64_sm3tt2bq4si): Ditto.
22372 (aarch64_sm3partw1qv4si): Ditto.
22373 (aarch64_sm3partw2qv4si): Ditto.
22374 (aarch64_sm4eqv4si): Ditto.
22375 (aarch64_sm4ekeyqv4si): Ditto.
22376 * config/aarch64/aarch64-simd.md:
22377 (aarch64_sm3ss1qv4si): Ditto.
22378 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
22379 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
22380 (aarch64_sm4eqv4si): Ditto.
22381 (aarch64_sm4ekeyqv4si): Ditto.
22382 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
22383 (sm3part_op): Ditto.
22384 (CRYPTO_SM3TT): Ditto.
22385 (CRYPTO_SM3PART): Ditto.
22386 (UNSPEC_SM3SS1): New unspec.
22387 (UNSPEC_SM3TT1A): Ditto.
22388 (UNSPEC_SM3TT1B): Ditto.
22389 (UNSPEC_SM3TT2A): Ditto.
22390 (UNSPEC_SM3TT2B): Ditto.
22391 (UNSPEC_SM3PARTW1): Ditto.
22392 (UNSPEC_SM3PARTW2): Ditto.
22393 (UNSPEC_SM4E): Ditto.
22394 (UNSPEC_SM4EKEY): Ditto.
22395 * config/aarch64/constraints.md (Ui2): New constraint.
22396 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
22397 * config/arm/types.md (crypto_sm3): New type attribute.
22398 (crypto_sm4): Ditto.
22399 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
22400 (vsm3tt1aq_u32): Ditto.
22401 (vsm3tt1bq_u32): Ditto.
22402 (vsm3tt2aq_u32): Ditto.
22403 (vsm3tt2bq_u32): Ditto.
22404 (vsm3partw1q_u32): Ditto.
22405 (vsm3partw2q_u32): Ditto.
22406 (vsm4eq_u32): Ditto.
22407 (vsm4ekeyq_u32): Ditto.
22408 (doc/invoke.texi): Document new sm4 option.
22410 2018-01-10 Michael Collison <michael.collison@arm.com>
22412 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
22413 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
22414 (AARCH64_FL_FOR_ARCH8_4): New.
22415 (AARCH64_FL_V8_4): New flag.
22416 (doc/invoke.texi): Document new armv8.4-a option.
22418 2018-01-10 Michael Collison <michael.collison@arm.com>
22420 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
22421 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
22422 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
22423 * config/aarch64/aarch64-option-extension.def: Add
22424 AARCH64_OPT_EXTENSION of 'sha2'.
22425 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
22426 (crypto): Disable sha2 and aes if crypto disabled.
22427 (crypto): Enable aes and sha2 if enabled.
22428 (simd): Disable sha2 and aes if simd disabled.
22429 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
22431 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
22432 (TARGET_SHA2): New feature flag for sha2.
22433 (TARGET_AES): New feature flag for aes.
22434 * config/aarch64/aarch64-simd.md:
22435 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
22436 conditional on TARGET_AES.
22437 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
22438 (aarch64_crypto_sha1hsi): Make pattern conditional
22440 (aarch64_crypto_sha1hv4si): Ditto.
22441 (aarch64_be_crypto_sha1hv4si): Ditto.
22442 (aarch64_crypto_sha1su1v4si): Ditto.
22443 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
22444 (aarch64_crypto_sha1su0v4si): Ditto.
22445 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
22446 (aarch64_crypto_sha256su0v4si): Ditto.
22447 (aarch64_crypto_sha256su1v4si): Ditto.
22448 (doc/invoke.texi): Document new aes and sha2 options.
22450 2018-01-10 Martin Sebor <msebor@redhat.com>
22452 PR tree-optimization/83781
22453 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
22456 2018-01-11 Martin Sebor <msebor@gmail.com>
22457 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22459 PR tree-optimization/83501
22460 PR tree-optimization/81703
22462 * tree-ssa-strlen.c (get_string_cst): Rename...
22463 (get_string_len): ...to this. Handle global constants.
22464 (handle_char_store): Adjust.
22466 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
22467 Jim Wilson <jimw@sifive.com>
22469 * config/riscv/riscv-protos.h (riscv_output_return): New.
22470 * config/riscv/riscv.c (struct machine_function): New naked_p field.
22471 (riscv_attribute_table, riscv_output_return),
22472 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
22473 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
22474 (riscv_compute_frame_info): Only compute frame->mask if not a naked
22476 (riscv_expand_prologue): Add early return for naked function.
22477 (riscv_expand_epilogue): Likewise.
22478 (riscv_function_ok_for_sibcall): Return false for naked function.
22479 (riscv_set_current_function): New.
22480 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
22481 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
22482 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
22483 * doc/extend.texi (RISC-V Function Attributes): New.
22485 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
22487 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
22488 check for 128-bit long double before checking TCmode.
22489 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
22490 128-bit long doubles before checking TFmode or TCmode.
22491 (FLOAT128_IBM_P): Likewise.
22493 2018-01-10 Martin Sebor <msebor@redhat.com>
22495 PR tree-optimization/83671
22496 * builtins.c (c_strlen): Unconditionally return zero for the empty
22498 Use -Warray-bounds for warnings.
22499 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
22500 for non-constant array indices with COMPONENT_REF, arrays of
22501 arrays, and pointers to arrays.
22502 (gimple_fold_builtin_strlen): Determine and set length range for
22503 non-constant character arrays.
22505 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
22507 PR middle-end/81897
22508 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
22511 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
22513 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
22515 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22518 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
22519 VECTOR_MEM_ALTIVEC_OR_VSX_P.
22520 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
22521 indexed_or_indirect_operand predicate.
22522 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
22523 (*vsx_le_perm_load_v8hi): Likewise.
22524 (*vsx_le_perm_load_v16qi): Likewise.
22525 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
22526 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
22527 (*vsx_le_perm_store_v8hi): Likewise.
22528 (*vsx_le_perm_store_v16qi): Likewise.
22529 (eight unnamed splitters): Likewise.
22531 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
22533 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
22534 * config/rs6000/emmintrin.h: Likewise.
22535 * config/rs6000/mmintrin.h: Likewise.
22536 * config/rs6000/xmmintrin.h: Likewise.
22538 2018-01-10 David Malcolm <dmalcolm@redhat.com>
22541 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
22543 * tree.c (tree_nop_conversion): Return true for location wrapper
22545 (maybe_wrap_with_location): New function.
22546 (selftest::check_strip_nops): New function.
22547 (selftest::test_location_wrappers): New function.
22548 (selftest::tree_c_tests): Call it.
22549 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
22550 (maybe_wrap_with_location): New decl.
22551 (EXPR_LOCATION_WRAPPER_P): New macro.
22552 (location_wrapper_p): New inline function.
22553 (tree_strip_any_location_wrapper): New inline function.
22555 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
22558 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
22559 stack_realign_offset for the largest alignment of stack slot
22561 (ix86_find_max_used_stack_alignment): New function.
22562 (ix86_finalize_stack_frame_flags): Use it. Set
22563 max_used_stack_alignment if we don't realign stack.
22564 * config/i386/i386.h (machine_function): Add
22565 max_used_stack_alignment.
22567 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
22569 * config/arm/arm.opt (-mbranch-cost): New option.
22570 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
22573 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
22576 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
22577 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
22579 2018-01-10 Richard Biener <rguenther@suse.de>
22582 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
22583 early out so it also covers the case where we have a non-NULL
22586 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22588 PR tree-optimization/83753
22589 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
22590 for non-strided grouped accesses if the number of elements is 1.
22592 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
22595 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
22596 * i386.h (TARGET_USE_GATHER): Define.
22597 * x86-tune.def (X86_TUNE_USE_GATHER): New.
22599 2018-01-10 Martin Liska <mliska@suse.cz>
22602 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
22603 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
22605 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
22606 CLEANUP_NO_PARTITIONING is not set.
22608 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
22610 * doc/rtl.texi: Remove documentation of (const ...) wrappers
22611 for vectors, as a partial revert of r254296.
22612 * rtl.h (const_vec_p): Delete.
22613 (const_vec_duplicate_p): Don't test for vector CONSTs.
22614 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
22615 * expmed.c (make_tree): Likewise.
22618 * common.md (E, F): Use CONSTANT_P instead of checking for
22620 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
22621 checking for CONST_VECTOR.
22623 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22625 PR middle-end/83575
22626 * predict.c (force_edge_cold): Handle in more sane way edges
22627 with no prediction.
22629 2018-01-09 Carl Love <cel@us.ibm.com>
22631 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
22633 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
22634 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
22635 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
22636 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
22637 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
22638 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
22639 * config/rs6000/rs6000-protos.h: Add extern defition for
22640 rs6000_generate_float2_double_code.
22641 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
22643 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
22644 (float2_v2df): Add define_expand.
22646 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
22649 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
22650 op_mode in the force_to_mode call.
22652 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22654 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
22655 instead of checking each element individually.
22656 (aarch64_evpc_uzp): Likewise.
22657 (aarch64_evpc_zip): Likewise.
22658 (aarch64_evpc_ext): Likewise.
22659 (aarch64_evpc_rev): Likewise.
22660 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
22661 instead of checking each element individually. Return true without
22663 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
22664 whether all selected elements come from the same input, instead of
22665 checking each element individually. Remove calls to gen_rtx_REG,
22666 start_sequence and end_sequence and instead assert that no rtl is
22669 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22671 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
22672 order of HIGH and CONST checks.
22674 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
22676 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
22677 if the destination isn't an SSA_NAME.
22679 2018-01-09 Richard Biener <rguenther@suse.de>
22681 PR tree-optimization/83668
22682 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
22684 (canonicalize_loop_form): ... here, renamed from ...
22685 (canonicalize_loop_closed_ssa_form): ... this and amended to
22686 swap successor edges for loop exit blocks to make us use
22687 the RPO order we need for initial schedule generation.
22689 2018-01-09 Joseph Myers <joseph@codesourcery.com>
22691 PR tree-optimization/64811
22692 * match.pd: When optimizing comparisons with Inf, avoid
22693 introducing or losing exceptions from comparisons with NaN.
22695 2018-01-09 Martin Liska <mliska@suse.cz>
22698 * asan.c (shadow_mem_size): Add gcc_assert.
22700 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
22702 Don't save registers in main().
22705 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
22706 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
22707 * config/avr/avr.c (avr_set_current_function): Don't error if
22708 naked, OS_task or OS_main are specified at the same time.
22709 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
22711 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
22713 * common/config/avr/avr-common.c (avr_option_optimization_table):
22714 Switch on -mmain-is-OS_task for optimizing compilations.
22716 2018-01-09 Richard Biener <rguenther@suse.de>
22718 PR tree-optimization/83572
22719 * graphite.c: Include cfganal.h.
22720 (graphite_transform_loops): Connect infinite loops to exit
22721 and remove fake edges at the end.
22723 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22725 * ipa-inline.c (edge_badness): Revert accidental checkin.
22727 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
22730 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
22731 symbols; not inline clones.
22733 2018-01-09 Jakub Jelinek <jakub@redhat.com>
22736 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
22737 hard registers. Formatting fixes.
22739 PR preprocessor/83722
22740 * gcc.c (try_generate_repro): Pass
22741 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
22742 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
22745 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
22746 Kito Cheng <kito.cheng@gmail.com>
22748 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
22749 (riscv_leaf_function_p): Delete.
22750 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
22752 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22754 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
22756 (do_ifelse): New function.
22757 (do_isel): New function.
22758 (do_sub3): New function.
22759 (do_add3): New function.
22760 (do_load_mask_compare): New function.
22761 (do_overlap_load_compare): New function.
22762 (expand_compare_loop): New function.
22763 (expand_block_compare): Call expand_compare_loop() when appropriate.
22764 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
22765 option description.
22766 (-mblock-compare-inline-loop-limit): New option.
22768 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22771 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
22772 Reverse order of second and third operands in first alternative.
22773 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
22774 of first and second elements in UNSPEC_VPERMR vector.
22775 (altivec_expand_vec_perm_le): Likewise.
22777 2018-01-08 Jeff Law <law@redhat.com>
22779 PR rtl-optimizatin/81308
22780 * tree-switch-conversion.c (cfg_altered): New file scoped static.
22781 (process_switch): If group_case_labels makes a change, then set
22783 (pass_convert_switch::execute): If a switch is converted, then
22784 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
22786 PR rtl-optimization/81308
22787 * recog.c (split_all_insns): Conditionally cleanup the CFG after
22790 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
22792 PR target/83663 - Revert r255946
22793 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
22794 generation for cases where splatting a value is not useful.
22795 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
22796 across a vec_duplicate and a paradoxical subreg forming a vector
22797 mode to a vec_concat.
22799 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22801 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
22802 -march=armv8.3-a variants.
22803 * config/arm/t-multilib: Likewise.
22804 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
22806 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22808 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
22810 (cceq_ior_compare_complement): Give it a name so I can use it, and
22811 change boolean_or_operator predicate to boolean_operator so it can
22812 be used to generate a crand.
22813 (eqne): New code iterator.
22814 (bd/bd_neg): New code_attrs.
22815 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
22816 a single define_insn.
22817 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
22818 decrement (bdnzt/bdnzf/bdzt/bdzf).
22819 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
22820 with the new names of the branch decrement patterns, and added the
22821 names of the branch decrement conditional patterns.
22823 2018-01-08 Richard Biener <rguenther@suse.de>
22825 PR tree-optimization/83563
22826 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
22829 2018-01-08 Richard Biener <rguenther@suse.de>
22831 PR middle-end/83713
22832 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
22834 2018-01-08 Richard Biener <rguenther@suse.de>
22836 PR tree-optimization/83685
22837 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
22838 references to abnormals.
22840 2018-01-08 Richard Biener <rguenther@suse.de>
22843 * dwarf2out.c (output_indirect_strings): Handle empty
22844 skeleton_debug_str_hash.
22845 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
22847 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
22849 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
22850 (emit_store_direct): Likewise.
22851 (arc_trampoline_adjust_address): Likewise.
22852 (arc_asm_trampoline_template): New function.
22853 (arc_initialize_trampoline): Use asm_trampoline_template.
22854 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
22855 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
22856 * config/arc/arc.md (flush_icache): Delete pattern.
22858 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
22860 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
22861 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
22864 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
22867 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
22868 by not USED_FOR_TARGET.
22869 (make_pass_resolve_sw_modes): Likewise.
22871 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
22873 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
22876 2018-01-08 Richard Biener <rguenther@suse.de>
22878 PR middle-end/83580
22879 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
22881 2018-01-08 Richard Biener <rguenther@suse.de>
22883 PR middle-end/83517
22884 * match.pd ((t * 2) / 2) -> t): Add missing :c.
22886 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
22888 PR middle-end/81897
22889 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
22890 basic blocks with a small number of successors.
22891 (convert_control_dep_chain_into_preds): Improve handling of
22893 (dump_predicates): Split apart into...
22894 (dump_pred_chain): ...here...
22895 (dump_pred_info): ...and here.
22896 (can_one_predicate_be_invalidated_p): Add debugging printfs.
22897 (can_chain_union_be_invalidated_p): Improve check for invalidation
22899 (uninit_uses_cannot_happen): Avoid unnecessary if
22900 convert_control_dep_chain_into_preds yielded nothing.
22902 2018-01-06 Martin Sebor <msebor@redhat.com>
22904 PR tree-optimization/83640
22905 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
22906 subtracting negative offset from size.
22907 (builtin_access::overlap): Adjust offset bounds of the access to fall
22908 within the size of the object if possible.
22910 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
22912 PR rtl-optimization/83699
22913 * expmed.c (extract_bit_field_1): Restrict the vector usage of
22914 extract_bit_field_as_subreg to cases in which the extracted
22915 value is also a vector.
22917 * lra-constraints.c (process_alt_operands): Test for the equivalence
22918 substitutions when detecting a possible reload cycle.
22920 2018-01-06 Jakub Jelinek <jakub@redhat.com>
22923 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
22924 by default if flag_selective_schedling{,2}. Formatting fixes.
22926 PR rtl-optimization/83682
22927 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
22928 if it has non-VECTOR_MODE element mode.
22929 (vec_duplicate_p): Likewise.
22931 PR middle-end/83694
22932 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
22933 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
22935 2018-01-05 Jakub Jelinek <jakub@redhat.com>
22938 * config/i386/i386-builtin.def
22939 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
22940 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
22941 Require also OPTION_MASK_ISA_AVX512F in addition to
22942 OPTION_MASK_ISA_GFNI.
22943 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
22944 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
22945 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
22946 to OPTION_MASK_ISA_GFNI.
22947 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
22948 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
22949 OPTION_MASK_ISA_AVX512BW.
22950 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
22951 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
22952 addition to OPTION_MASK_ISA_GFNI.
22953 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
22954 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
22955 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
22956 to OPTION_MASK_ISA_GFNI.
22957 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
22958 a requirement for all ISAs rather than any of them with a few
22960 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
22962 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
22963 bitmasks to be enabled with 3 exceptions, instead of requiring any
22964 enabled ISA with lots of exceptions.
22965 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
22966 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
22967 Change avx512bw in isa attribute to avx512f.
22968 * config/i386/sgxintrin.h: Add license boilerplate.
22969 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
22970 to __AVX512F__ and __AVX512VL to __AVX512VL__.
22971 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
22972 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
22974 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
22975 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
22976 temporarily sse2 rather than sse if not enabled already.
22979 * config/i386/sse.md (VI248_VLBW): Rename to ...
22980 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
22981 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
22982 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
22983 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
22984 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
22985 mode iterator instead of VI248_VLBW.
22987 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
22989 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
22990 (record_modified): Skip clobbers; add debug output.
22991 (param_change_prob): Use sreal frequencies.
22993 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
22995 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
22996 punt for user-aligned variables.
22998 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23000 * tree-chrec.c (chrec_contains_symbols): Return true for
23003 2018-01-05 Sudakshina Das <sudi.das@arm.com>
23006 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
23007 of (x|y) == x for BICS pattern.
23009 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23011 PR tree-optimization/83605
23012 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
23013 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
23016 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
23018 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
23019 * config/epiphany/rtems.h: New file.
23021 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23022 Uros Bizjak <ubizjak@gmail.com>
23025 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
23026 QIreg_operand instead of register_operand predicate.
23027 * config/i386/i386.c (ix86_rop_should_change_byte_p,
23028 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
23029 comments instead of -fmitigate[-_]rop.
23031 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23034 * cgraphunit.c (symbol_table::compile): Switch to text_section
23035 before calling assembly_start debug hook.
23036 * run-rtl-passes.c (run_rtl_passes): Likewise.
23039 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23041 * tree-vrp.c (extract_range_from_binary_expr_1): Check
23042 range_int_cst_p rather than !symbolic_range_p before calling
23043 extract_range_from_multiplicative_op_1.
23045 2018-01-04 Jeff Law <law@redhat.com>
23047 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
23048 redundant test in assertion.
23050 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23052 * doc/rtl.texi: Document machine_mode wrapper classes.
23054 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23056 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
23057 using tree_to_uhwi.
23059 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23061 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
23062 the VEC_PERM_EXPR fold to fail.
23064 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23067 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
23068 to switched_sections.
23070 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23073 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
23074 test for d.testing.
23076 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
23079 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
23080 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
23082 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23085 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
23086 is BLKmode and bitpos not zero or mode change is needed.
23088 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
23091 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
23094 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
23097 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
23098 instead of MULT rtx. Update all corresponding splitters.
23099 (*saddl_se): Ditto.
23100 (*ssub<modesuffix>): Ditto.
23101 (*ssubl_se): Ditto.
23102 (*cmp_sadd_di): Update split patterns.
23103 (*cmp_sadd_si): Ditto.
23104 (*cmp_sadd_sidi): Ditto.
23105 (*cmp_ssub_di): Ditto.
23106 (*cmp_ssub_si): Ditto.
23107 (*cmp_ssub_sidi): Ditto.
23108 * config/alpha/predicates.md (const23_operand): New predicate.
23109 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
23110 Look for ASHIFT, not MULT inner operand.
23111 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
23113 2018-01-04 Martin Liska <mliska@suse.cz>
23115 PR gcov-profile/83669
23116 * gcov.c (output_intermediate_file): Add version to intermediate
23118 * doc/gcov.texi: Document new field 'version' in intermediate
23119 file format. Fix location of '-k' option of gcov command.
23121 2018-01-04 Martin Liska <mliska@suse.cz>
23124 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
23126 2018-01-04 Jakub Jelinek <jakub@redhat.com>
23128 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
23130 2018-01-03 Martin Sebor <msebor@redhat.com>
23132 PR tree-optimization/83655
23133 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
23134 checking calls with invalid arguments.
23136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23138 * tree-vect-stmts.c (vect_get_store_rhs): New function.
23139 (vectorizable_mask_load_store): Delete.
23140 (vectorizable_call): Return false for masked loads and stores.
23141 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
23142 instead of gimple_assign_rhs1.
23143 (vectorizable_load): Handle IFN_MASK_LOAD.
23144 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
23146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23148 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
23150 (vectorizable_mask_load_store): ...here.
23151 (vectorizable_load): ...and here.
23153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23155 * tree-vect-stmts.c (vect_build_all_ones_mask)
23156 (vect_build_zero_merge_argument): New functions, split out from...
23157 (vectorizable_load): ...here.
23159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23161 * tree-vect-stmts.c (vect_check_store_rhs): New function,
23163 (vectorizable_mask_load_store): ...here.
23164 (vectorizable_store): ...and here.
23166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23168 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
23170 (vectorizable_mask_load_store): ...here.
23172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23174 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
23175 (vect_model_store_cost): Take a vec_load_store_type instead of a
23177 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
23178 (vect_model_store_cost): Take a vec_load_store_type instead of a
23180 (vectorizable_mask_load_store): Update accordingly.
23181 (vectorizable_store): Likewise.
23182 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
23184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23186 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
23187 IFN_MASK_LOAD calls here rather than...
23188 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
23190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23191 Alan Hayward <alan.hayward@arm.com>
23192 David Sherwood <david.sherwood@arm.com>
23194 * expmed.c (extract_bit_field_1): For vector extracts,
23195 fall back to extract_bit_field_as_subreg if vec_extract
23198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23199 Alan Hayward <alan.hayward@arm.com>
23200 David Sherwood <david.sherwood@arm.com>
23202 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
23203 they are variable or constant sized.
23204 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
23205 slots for constant-sized data.
23207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23208 Alan Hayward <alan.hayward@arm.com>
23209 David Sherwood <david.sherwood@arm.com>
23211 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
23212 handling COND_EXPRs with boolean comparisons, try to find a better
23213 basis for the mask type than the boolean itself.
23215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23217 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
23218 is calculated and how it can be overridden.
23219 * genmodes.c (max_bitsize_mode_any_mode): New variable.
23220 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
23222 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
23225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23226 Alan Hayward <alan.hayward@arm.com>
23227 David Sherwood <david.sherwood@arm.com>
23229 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
23230 Remove the mode argument.
23231 (aarch64_simd_valid_immediate): Remove the mode and inverse
23233 * config/aarch64/iterators.md (bitsize): New iterator.
23234 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
23235 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
23236 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
23237 aarch64_simd_valid_immediate.
23238 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
23239 (aarch64_reg_or_bic_imm): Likewise.
23240 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
23241 with an insn_type enum and msl with a modifier_type enum.
23242 Replace element_width with a scalar_mode. Change the shift
23243 to unsigned int. Add constructors for scalar_float_mode and
23244 scalar_int_mode elements.
23245 (aarch64_vect_float_const_representable_p): Delete.
23246 (aarch64_can_const_movi_rtx_p)
23247 (aarch64_simd_scalar_immediate_valid_for_move)
23248 (aarch64_simd_make_constant): Update call to
23249 aarch64_simd_valid_immediate.
23250 (aarch64_advsimd_valid_immediate_hs): New function.
23251 (aarch64_advsimd_valid_immediate): Likewise.
23252 (aarch64_simd_valid_immediate): Remove mode and inverse
23253 arguments. Rewrite to use the above. Use const_vec_duplicate_p
23254 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
23255 and aarch64_float_const_representable_p on the result.
23256 (aarch64_output_simd_mov_immediate): Remove mode argument.
23257 Update call to aarch64_simd_valid_immediate and use of
23258 simd_immediate_info.
23259 (aarch64_output_scalar_simd_mov_immediate): Update call
23262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23263 Alan Hayward <alan.hayward@arm.com>
23264 David Sherwood <david.sherwood@arm.com>
23266 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
23267 (mode_nunits): Likewise CONST_MODE_NUNITS.
23268 * machmode.def (ADJUST_NUNITS): Document.
23269 * genmodes.c (mode_data::need_nunits_adj): New field.
23270 (blank_mode): Update accordingly.
23271 (adj_nunits): New variable.
23272 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
23274 (emit_mode_size_inline): Set need_bytesize_adj for all modes
23275 listed in adj_nunits.
23276 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
23277 listed in adj_nunits. Don't emit case statements for such modes.
23278 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
23279 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
23280 nothing if adj_nunits is nonnull.
23281 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
23282 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
23283 (emit_mode_fbit): Update use of print_maybe_const_decl.
23284 (emit_move_size): Likewise. Treat the array as non-const
23286 (emit_mode_adjustments): Handle adj_nunits.
23288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23290 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
23291 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
23292 (VECTOR_MODES): Use it.
23293 (make_vector_modes): Take the prefix as an argument.
23295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23296 Alan Hayward <alan.hayward@arm.com>
23297 David Sherwood <david.sherwood@arm.com>
23299 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
23300 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
23301 for MODE_VECTOR_BOOL.
23302 * machmode.def (VECTOR_BOOL_MODE): Document.
23303 * genmodes.c (VECTOR_BOOL_MODE): New macro.
23304 (make_vector_bool_mode): New function.
23305 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
23307 * lto-streamer-in.c (lto_input_mode_table): Likewise.
23308 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
23310 * stor-layout.c (int_mode_for_mode): Likewise.
23311 * tree.c (build_vector_type_for_mode): Likewise.
23312 * varasm.c (output_constant_pool_2): Likewise.
23313 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
23314 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
23315 for MODE_VECTOR_BOOL.
23316 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
23317 of mode class checks.
23318 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
23319 instead of a list of mode class checks.
23320 (expand_vector_scalar_condition): Likewise.
23321 (type_for_widest_vector_mode): Handle BImode as an inner mode.
23323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23324 Alan Hayward <alan.hayward@arm.com>
23325 David Sherwood <david.sherwood@arm.com>
23327 * machmode.h (mode_size): Change from unsigned short to
23329 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
23330 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23331 or if measurement_type is not polynomial.
23332 (fixed_size_mode::includes_p): Check for constant-sized modes.
23333 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
23334 return a poly_uint16 rather than an unsigned short.
23335 (emit_mode_size): Change the type of mode_size from unsigned short
23336 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
23337 (emit_mode_adjustments): Cope with polynomial vector sizes.
23338 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23340 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23342 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
23343 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
23344 * caller-save.c (setup_save_areas): Likewise.
23345 (replace_reg_with_saved_mem): Likewise.
23346 * calls.c (emit_library_call_value_1): Likewise.
23347 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
23348 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
23349 (gen_lowpart_for_combine): Likewise.
23350 * convert.c (convert_to_integer_1): Likewise.
23351 * cse.c (equiv_constant, cse_insn): Likewise.
23352 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
23353 (cselib_subst_to_values): Likewise.
23354 * dce.c (word_dce_process_block): Likewise.
23355 * df-problems.c (df_word_lr_mark_ref): Likewise.
23356 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
23357 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
23358 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
23359 (rtl_for_decl_location): Likewise.
23360 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
23361 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
23362 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
23363 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
23364 (expand_expr_real_1): Likewise.
23365 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
23366 (pad_below): Likewise.
23367 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23368 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
23369 * ira.c (get_subreg_tracking_sizes): Likewise.
23370 * ira-build.c (ira_create_allocno_objects): Likewise.
23371 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
23372 (ira_sort_regnos_for_alter_reg): Likewise.
23373 * ira-costs.c (record_operand_costs): Likewise.
23374 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
23375 (resolve_simple_move): Likewise.
23376 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
23377 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
23378 (lra_constraints): Likewise.
23379 (CONST_POOL_OK_P): Reject variable-sized modes.
23380 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
23381 (add_pseudo_to_slot, lra_spill): Likewise.
23382 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23383 * optabs-query.c (get_best_extraction_insn): Likewise.
23384 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23385 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
23386 (expand_mult_highpart, valid_multiword_target_p): Likewise.
23387 * recog.c (offsettable_address_addr_space_p): Likewise.
23388 * regcprop.c (maybe_mode_change): Likewise.
23389 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
23390 * regrename.c (build_def_use): Likewise.
23391 * regstat.c (dump_reg_info): Likewise.
23392 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
23393 (find_reloads, find_reloads_subreg_address): Likewise.
23394 * reload1.c (eliminate_regs_1): Likewise.
23395 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
23396 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
23397 (simplify_binary_operation_1, simplify_subreg): Likewise.
23398 * targhooks.c (default_function_arg_padding): Likewise.
23399 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
23400 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
23401 (verify_gimple_assign_ternary): Likewise.
23402 * tree-inline.c (estimate_move_cost): Likewise.
23403 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23404 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
23405 (get_address_cost_ainc): Likewise.
23406 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
23407 (vect_supportable_dr_alignment): Likewise.
23408 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23409 (vectorizable_reduction): Likewise.
23410 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
23411 (vectorizable_operation, vectorizable_load): Likewise.
23412 * tree.c (build_same_sized_truth_vector_type): Likewise.
23413 * valtrack.c (cleanup_auto_inc_dec): Likewise.
23414 * var-tracking.c (emit_note_insn_var_location): Likewise.
23415 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
23416 (ADDR_VEC_ALIGN): Likewise.
23418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23419 Alan Hayward <alan.hayward@arm.com>
23420 David Sherwood <david.sherwood@arm.com>
23422 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
23424 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
23425 or if measurement_type is polynomial.
23426 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
23427 * combine.c (make_extraction): Likewise.
23428 * dse.c (find_shift_sequence): Likewise.
23429 * dwarf2out.c (mem_loc_descriptor): Likewise.
23430 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
23431 (extract_bit_field, extract_low_bits): Likewise.
23432 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
23433 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
23434 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
23435 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
23436 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
23437 * reload.c (find_reloads): Likewise.
23438 * reload1.c (alter_reg): Likewise.
23439 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
23440 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
23441 * tree-if-conv.c (predicate_mem_writes): Likewise.
23442 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
23443 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
23444 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23445 * valtrack.c (dead_debug_insert_temp): Likewise.
23446 * varasm.c (mergeable_constant_section): Likewise.
23447 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
23449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23450 Alan Hayward <alan.hayward@arm.com>
23451 David Sherwood <david.sherwood@arm.com>
23453 * expr.c (expand_assignment): Cope with polynomial mode sizes
23454 when assigning to a CONCAT.
23456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23457 Alan Hayward <alan.hayward@arm.com>
23458 David Sherwood <david.sherwood@arm.com>
23460 * machmode.h (mode_precision): Change from unsigned short to
23462 (mode_to_precision): Return a poly_uint16 rather than an unsigned
23464 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
23465 or if measurement_type is not polynomial.
23466 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
23467 in which the mode is already known to be a scalar_int_mode.
23468 * genmodes.c (emit_mode_precision): Change the type of mode_precision
23469 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
23471 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23472 for GET_MODE_PRECISION.
23473 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23474 for GET_MODE_PRECISION.
23475 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
23477 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
23478 (expand_field_assignment, make_extraction): Likewise.
23479 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
23480 (get_last_value): Likewise.
23481 * convert.c (convert_to_integer_1): Likewise.
23482 * cse.c (cse_insn): Likewise.
23483 * expr.c (expand_expr_real_1): Likewise.
23484 * lra-constraints.c (simplify_operand_subreg): Likewise.
23485 * optabs-query.c (can_atomic_load_p): Likewise.
23486 * optabs.c (expand_atomic_load): Likewise.
23487 (expand_atomic_store): Likewise.
23488 * ree.c (combine_reaching_defs): Likewise.
23489 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
23490 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
23491 * tree.h (type_has_mode_precision_p): Likewise.
23492 * ubsan.c (instrument_si_overflow): Likewise.
23494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23495 Alan Hayward <alan.hayward@arm.com>
23496 David Sherwood <david.sherwood@arm.com>
23498 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
23499 polynomial numbers of units.
23500 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
23501 (valid_vector_subparts_p): New function.
23502 (build_vector_type): Remove temporary shim and take the number
23503 of units as a poly_uint64 rather than an int.
23504 (build_opaque_vector_type): Take the number of units as a
23505 poly_uint64 rather than an int.
23506 * tree.c (build_vector_from_ctor): Handle polynomial
23507 TYPE_VECTOR_SUBPARTS.
23508 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
23509 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
23510 (build_vector_from_val): If the number of units is variable,
23511 use build_vec_duplicate_cst for constant operands and
23512 VEC_DUPLICATE_EXPR otherwise.
23513 (make_vector_type): Remove temporary is_constant ().
23514 (build_vector_type, build_opaque_vector_type): Take the number of
23515 units as a poly_uint64 rather than an int.
23516 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
23518 * cfgexpand.c (expand_debug_expr): Likewise.
23519 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
23520 (store_constructor, expand_expr_real_1): Likewise.
23521 (const_scalar_mask_from_tree): Likewise.
23522 * fold-const-call.c (fold_const_reduction): Likewise.
23523 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
23524 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
23525 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
23526 (fold_relational_const): Likewise.
23527 (native_interpret_vector): Likewise. Change the size from an
23528 int to an unsigned int.
23529 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
23530 TYPE_VECTOR_SUBPARTS.
23531 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
23532 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
23533 duplicating a non-constant operand into a variable-length vector.
23534 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
23535 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
23536 * ipa-icf.c (sem_variable::equals): Likewise.
23537 * match.pd: Likewise.
23538 * omp-simd-clone.c (simd_clone_subparts): Likewise.
23539 * print-tree.c (print_node): Likewise.
23540 * stor-layout.c (layout_type): Likewise.
23541 * targhooks.c (default_builtin_vectorization_cost): Likewise.
23542 * tree-cfg.c (verify_gimple_comparison): Likewise.
23543 (verify_gimple_assign_binary): Likewise.
23544 (verify_gimple_assign_ternary): Likewise.
23545 (verify_gimple_assign_single): Likewise.
23546 * tree-pretty-print.c (dump_generic_node): Likewise.
23547 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23548 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
23549 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
23550 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
23551 (vect_shift_permute_load_chain): Likewise.
23552 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
23553 (expand_vector_condition, optimize_vector_constructor): Likewise.
23554 (lower_vec_perm, get_compute_type): Likewise.
23555 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
23556 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
23557 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
23558 (vect_recog_mask_conversion_pattern): Likewise.
23559 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
23560 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
23561 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23562 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
23563 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
23564 (vectorizable_shift, vectorizable_operation, vectorizable_store)
23565 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
23566 (supportable_widening_operation): Likewise.
23567 (supportable_narrowing_operation): Likewise.
23568 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
23570 * varasm.c (output_constant): Likewise.
23572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23573 Alan Hayward <alan.hayward@arm.com>
23574 David Sherwood <david.sherwood@arm.com>
23576 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
23577 so that both the length == 3 and length != 3 cases set up their
23578 own permute vectors. Add comments explaining why we know the
23579 number of elements is constant.
23580 (vect_permute_load_chain): Likewise.
23582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23583 Alan Hayward <alan.hayward@arm.com>
23584 David Sherwood <david.sherwood@arm.com>
23586 * machmode.h (mode_nunits): Change from unsigned char to
23588 (ONLY_FIXED_SIZE_MODES): New macro.
23589 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
23590 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
23591 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
23593 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
23594 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
23595 or if measurement_type is not polynomial.
23596 * genmodes.c (ZERO_COEFFS): New macro.
23597 (emit_mode_nunits_inline): Make mode_nunits_inline return a
23599 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
23600 Use ZERO_COEFFS when emitting initializers.
23601 * data-streamer.h (bp_pack_poly_value): New function.
23602 (bp_unpack_poly_value): Likewise.
23603 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
23604 for GET_MODE_NUNITS.
23605 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
23606 for GET_MODE_NUNITS.
23607 * tree.c (make_vector_type): Remove temporary shim and make
23608 the real function take the number of units as a poly_uint64
23609 rather than an int.
23610 (build_vector_type_for_mode): Handle polynomial nunits.
23611 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
23612 * emit-rtl.c (const_vec_series_p_1): Likewise.
23613 (gen_rtx_CONST_VECTOR): Likewise.
23614 * fold-const.c (test_vec_duplicate_folding): Likewise.
23615 * genrecog.c (validate_pattern): Likewise.
23616 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
23617 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
23618 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
23619 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
23620 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
23621 * rtlanal.c (subreg_get_info): Likewise.
23622 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23623 (vect_grouped_load_supported): Likewise.
23624 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
23625 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
23626 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
23627 (simplify_const_unary_operation, simplify_binary_operation_1)
23628 (simplify_const_binary_operation, simplify_ternary_operation)
23629 (test_vector_ops_duplicate, test_vector_ops): Likewise.
23630 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
23631 instead of CONST_VECTOR_NUNITS.
23632 * varasm.c (output_constant_pool_2): Likewise.
23633 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
23634 explicit-encoded elements in the XVEC for variable-length vectors.
23636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23638 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
23640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23641 Alan Hayward <alan.hayward@arm.com>
23642 David Sherwood <david.sherwood@arm.com>
23644 * coretypes.h (fixed_size_mode): Declare.
23645 (fixed_size_mode_pod): New typedef.
23646 * builtins.h (target_builtins::x_apply_args_mode)
23647 (target_builtins::x_apply_result_mode): Change type to
23648 fixed_size_mode_pod.
23649 * builtins.c (apply_args_size, apply_result_size, result_vector)
23650 (expand_builtin_apply_args_1, expand_builtin_apply)
23651 (expand_builtin_return): Update accordingly.
23653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23655 * cse.c (hash_rtx_cb): Hash only the encoded elements.
23656 * cselib.c (cselib_hash_rtx): Likewise.
23657 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
23658 CONST_VECTOR encoding.
23660 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23661 Jeff Law <law@redhat.com>
23664 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
23665 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
23666 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
23667 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
23670 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
23671 explicitly probe *sp in a noreturn function if there were any callee
23672 register saves or frame pointer is needed.
23674 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23677 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
23678 BLKmode for ternary, binary or unary expressions.
23681 * var-tracking.c (delete_vta_debug_insn): New inline function.
23682 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
23683 insns from get_insns () to NULL instead of each bb separately.
23684 Use delete_vta_debug_insn. No longer static.
23685 (vt_debug_insns_local, variable_tracking_main_1): Adjust
23686 delete_vta_debug_insns callers.
23687 * rtl.h (delete_vta_debug_insns): Declare.
23688 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
23689 instead of variable_tracking_main.
23691 2018-01-03 Martin Sebor <msebor@redhat.com>
23693 PR tree-optimization/83603
23694 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
23695 arguments past the endof the argument list in functions declared
23696 without a prototype.
23697 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
23698 Avoid checking when arguments are null.
23700 2018-01-03 Martin Sebor <msebor@redhat.com>
23703 * doc/extend.texi (attribute const): Fix a typo.
23704 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
23705 issuing -Wsuggest-attribute for void functions.
23707 2018-01-03 Martin Sebor <msebor@redhat.com>
23709 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
23710 offset_int::from instead of wide_int::to_shwi.
23711 (maybe_diag_overlap): Remove assertion.
23712 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
23713 * gimple-ssa-sprintf.c (format_directive): Same.
23714 (parse_directive): Same.
23715 (sprintf_dom_walker::compute_format_length): Same.
23716 (try_substitute_return_value): Same.
23718 2018-01-03 Jeff Law <law@redhat.com>
23720 PR middle-end/83654
23721 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
23722 non-constant residual for zero at runtime and avoid probing in
23723 that case. Reorganize code for trailing problem to mirror handling
23726 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23728 PR tree-optimization/83501
23729 * tree-ssa-strlen.c (get_string_cst): New.
23730 (handle_char_store): Call get_string_cst.
23732 2018-01-03 Martin Liska <mliska@suse.cz>
23734 PR tree-optimization/83593
23735 * tree-ssa-strlen.c: Include tree-cfg.h.
23736 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
23737 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
23738 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
23740 (strlen_dom_walker::before_dom_children): Call
23741 gimple_purge_dead_eh_edges. Dump tranformation with details
23743 (strlen_dom_walker::before_dom_children): Update call by adding
23744 new argument cleanup_eh.
23745 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
23747 2018-01-03 Martin Liska <mliska@suse.cz>
23750 * cif-code.def (VARIADIC_THUNK): New enum value.
23751 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
23754 2018-01-03 Jan Beulich <jbeulich@suse.com>
23756 * sse.md (mov<mode>_internal): Tighten condition for when to use
23757 vmovdqu<ssescalarsize> for TI and OI modes.
23759 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23761 Update copyright years.
23763 2018-01-03 Martin Liska <mliska@suse.cz>
23766 * ipa-visibility.c (function_and_variable_visibility): Skip
23767 functions with noipa attribure.
23769 2018-01-03 Jakub Jelinek <jakub@redhat.com>
23771 * gcc.c (process_command): Update copyright notice dates.
23772 * gcov-dump.c (print_version): Ditto.
23773 * gcov.c (print_version): Ditto.
23774 * gcov-tool.c (print_version): Ditto.
23775 * gengtype.c (create_file): Ditto.
23776 * doc/cpp.texi: Bump @copying's copyright year.
23777 * doc/cppinternals.texi: Ditto.
23778 * doc/gcc.texi: Ditto.
23779 * doc/gccint.texi: Ditto.
23780 * doc/gcov.texi: Ditto.
23781 * doc/install.texi: Ditto.
23782 * doc/invoke.texi: Ditto.
23784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23786 * vector-builder.h (vector_builder::m_full_nelts): Change from
23787 unsigned int to poly_uint64.
23788 (vector_builder::full_nelts): Update prototype accordingly.
23789 (vector_builder::new_vector): Likewise.
23790 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
23791 (vector_builder::operator ==): Likewise.
23792 (vector_builder::finalize): Likewise.
23793 * int-vector-builder.h (int_vector_builder::int_vector_builder):
23794 Take the number of elements as a poly_uint64 rather than an
23796 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
23797 from unsigned int to poly_uint64.
23798 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
23799 (vec_perm_indices::new_vector): Likewise.
23800 (vec_perm_indices::length): Likewise.
23801 (vec_perm_indices::nelts_per_input): Likewise.
23802 (vec_perm_indices::input_nelts): Likewise.
23803 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
23804 number of elements per input as a poly_uint64 rather than an
23805 unsigned int. Use the original encoding for variable-length
23806 vectors, rather than clamping each individual element.
23807 For the second and subsequent elements in each pattern,
23808 clamp the step and base before clamping their sum.
23809 (vec_perm_indices::series_p): Handle polynomial element counts.
23810 (vec_perm_indices::all_in_range_p): Likewise.
23811 (vec_perm_indices_to_tree): Likewise.
23812 (vec_perm_indices_to_rtx): Likewise.
23813 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
23814 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
23815 (tree_vector_builder::new_binary_operation): Handle polynomial
23816 element counts. Return false if we need to know the number
23817 of elements at compile time.
23818 * fold-const.c (fold_vec_perm): Punt if the number of elements
23819 isn't known at compile time.
23821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23823 * vec-perm-indices.h (vec_perm_builder): Change element type
23824 from HOST_WIDE_INT to poly_int64.
23825 (vec_perm_indices::element_type): Update accordingly.
23826 (vec_perm_indices::clamp): Handle polynomial element_types.
23827 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23828 (vec_perm_indices::all_in_range_p): Likewise.
23829 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
23831 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
23832 polynomial vec_perm_indices element types.
23833 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
23834 * fold-const.c (fold_vec_perm): Likewise.
23835 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
23836 * tree-vect-generic.c (lower_vec_perm): Likewise.
23837 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23838 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
23839 element type to HOST_WIDE_INT.
23841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23842 Alan Hayward <alan.hayward@arm.com>
23843 David Sherwood <david.sherwood@arm.com>
23845 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
23846 rather than an int. Use plus_constant.
23847 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
23848 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
23850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23851 Alan Hayward <alan.hayward@arm.com>
23852 David Sherwood <david.sherwood@arm.com>
23854 * calls.c (emit_call_1, expand_call): Change struct_value_size from
23855 a HOST_WIDE_INT to a poly_int64.
23857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23858 Alan Hayward <alan.hayward@arm.com>
23859 David Sherwood <david.sherwood@arm.com>
23861 * calls.c (load_register_parameters): Cope with polynomial
23862 mode sizes. Require a constant size for BLKmode parameters
23863 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
23864 forces a parameter to be padded at the lsb end in order to
23865 fill a complete number of words, require the parameter size
23866 to be ordered wrt UNITS_PER_WORD.
23868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23869 Alan Hayward <alan.hayward@arm.com>
23870 David Sherwood <david.sherwood@arm.com>
23872 * reload1.c (spill_stack_slot_width): Change element type
23873 from unsigned int to poly_uint64_pod.
23874 (alter_reg): Treat mode sizes as polynomial.
23876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23877 Alan Hayward <alan.hayward@arm.com>
23878 David Sherwood <david.sherwood@arm.com>
23880 * reload.c (complex_word_subreg_p): New function.
23881 (reload_inner_reg_of_subreg, push_reload): Use it.
23883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23884 Alan Hayward <alan.hayward@arm.com>
23885 David Sherwood <david.sherwood@arm.com>
23887 * lra-constraints.c (process_alt_operands): Reject matched
23888 operands whose sizes aren't ordered.
23889 (match_reload): Refer to this check here.
23891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23892 Alan Hayward <alan.hayward@arm.com>
23893 David Sherwood <david.sherwood@arm.com>
23895 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
23896 that the mode size is in the set {1, 2, 4, 8, 16}.
23898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23899 Alan Hayward <alan.hayward@arm.com>
23900 David Sherwood <david.sherwood@arm.com>
23902 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
23903 Use plus_constant instead of gen_rtx_PLUS.
23905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23906 Alan Hayward <alan.hayward@arm.com>
23907 David Sherwood <david.sherwood@arm.com>
23909 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
23910 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
23911 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
23912 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
23913 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
23914 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
23915 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
23916 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
23917 * config/i386/i386.c (ix86_push_rounding): ...this new function.
23918 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
23920 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
23921 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
23922 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
23923 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
23924 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
23925 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
23926 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
23927 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
23928 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
23929 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
23931 * expr.c (emit_move_resolve_push): Treat the input and result
23932 of PUSH_ROUNDING as a poly_int64.
23933 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
23934 (emit_push_insn): Likewise.
23935 * lra-eliminations.c (mark_not_eliminable): Likewise.
23936 * recog.c (push_operand): Likewise.
23937 * reload1.c (elimination_effects): Likewise.
23938 * rtlanal.c (nonzero_bits1): Likewise.
23939 * calls.c (store_one_arg): Likewise. Require the padding to be
23940 known at compile time.
23942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23943 Alan Hayward <alan.hayward@arm.com>
23944 David Sherwood <david.sherwood@arm.com>
23946 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
23947 Use plus_constant instead of gen_rtx_PLUS.
23949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23950 Alan Hayward <alan.hayward@arm.com>
23951 David Sherwood <david.sherwood@arm.com>
23953 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
23954 rather than an int.
23956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23957 Alan Hayward <alan.hayward@arm.com>
23958 David Sherwood <david.sherwood@arm.com>
23960 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
23961 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
23962 via stack temporaries. Treat the mode size as polynomial too.
23964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23965 Alan Hayward <alan.hayward@arm.com>
23966 David Sherwood <david.sherwood@arm.com>
23968 * expr.c (expand_expr_real_2): When handling conversions involving
23969 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
23970 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
23971 as a poly_uint64 too.
23973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23974 Alan Hayward <alan.hayward@arm.com>
23975 David Sherwood <david.sherwood@arm.com>
23977 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
23979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23980 Alan Hayward <alan.hayward@arm.com>
23981 David Sherwood <david.sherwood@arm.com>
23983 * combine.c (can_change_dest_mode): Handle polynomial
23984 REGMODE_NATURAL_SIZE.
23985 * expmed.c (store_bit_field_1): Likewise.
23986 * expr.c (store_constructor): Likewise.
23987 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
23988 and polynomial REGMODE_NATURAL_SIZE.
23989 (gen_lowpart_common): Likewise.
23990 * reginfo.c (record_subregs_of_mode): Likewise.
23991 * rtlanal.c (read_modify_subreg_p): Likewise.
23993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23994 Alan Hayward <alan.hayward@arm.com>
23995 David Sherwood <david.sherwood@arm.com>
23997 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
23998 numbers of elements.
24000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24001 Alan Hayward <alan.hayward@arm.com>
24002 David Sherwood <david.sherwood@arm.com>
24004 * match.pd: Cope with polynomial numbers of vector elements.
24006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24007 Alan Hayward <alan.hayward@arm.com>
24008 David Sherwood <david.sherwood@arm.com>
24010 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
24011 in a POINTER_PLUS_EXPR.
24013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24014 Alan Hayward <alan.hayward@arm.com>
24015 David Sherwood <david.sherwood@arm.com>
24017 * omp-simd-clone.c (simd_clone_subparts): New function.
24018 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
24019 (ipa_simd_modify_function_body): Likewise.
24021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24022 Alan Hayward <alan.hayward@arm.com>
24023 David Sherwood <david.sherwood@arm.com>
24025 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
24026 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
24027 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
24028 (expand_vector_condition, vector_element): Likewise.
24029 (subparts_gt): New function.
24030 (get_compute_type): Use subparts_gt.
24031 (count_type_subparts): Delete.
24032 (expand_vector_operations_1): Use subparts_gt instead of
24033 count_type_subparts.
24035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24036 Alan Hayward <alan.hayward@arm.com>
24037 David Sherwood <david.sherwood@arm.com>
24039 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
24040 (vect_compile_time_alias): ...this new function. Do the calculation
24041 on poly_ints rather than trees.
24042 (vect_prune_runtime_alias_test_list): Update call accordingly.
24044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24045 Alan Hayward <alan.hayward@arm.com>
24046 David Sherwood <david.sherwood@arm.com>
24048 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
24050 (vect_schedule_slp_instance): Likewise.
24052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24053 Alan Hayward <alan.hayward@arm.com>
24054 David Sherwood <david.sherwood@arm.com>
24056 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
24057 constant and extern definitions for variable-length vectors.
24058 (vect_get_constant_vectors): Note that the number of units
24059 is known to be constant.
24061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24062 Alan Hayward <alan.hayward@arm.com>
24063 David Sherwood <david.sherwood@arm.com>
24065 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
24066 of units as polynomial. Choose between WIDE and NARROW based
24069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24070 Alan Hayward <alan.hayward@arm.com>
24071 David Sherwood <david.sherwood@arm.com>
24073 * tree-vect-stmts.c (simd_clone_subparts): New function.
24074 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
24076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24077 Alan Hayward <alan.hayward@arm.com>
24078 David Sherwood <david.sherwood@arm.com>
24080 * tree-vect-stmts.c (vectorizable_call): Treat the number of
24081 vectors as polynomial. Use build_index_vector for
24082 IFN_GOMP_SIMD_LANE.
24084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24085 Alan Hayward <alan.hayward@arm.com>
24086 David Sherwood <david.sherwood@arm.com>
24088 * tree-vect-stmts.c (get_load_store_type): Treat the number of
24089 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
24090 for variable-length vectors.
24091 (vectorizable_mask_load_store): Treat the number of units as
24092 polynomial, asserting that it is constant if the condition has
24093 already been enforced.
24094 (vectorizable_store, vectorizable_load): Likewise.
24096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24097 Alan Hayward <alan.hayward@arm.com>
24098 David Sherwood <david.sherwood@arm.com>
24100 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
24101 of units as polynomial. Punt if we can't tell at compile time
24102 which vector contains the final result.
24104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24105 Alan Hayward <alan.hayward@arm.com>
24106 David Sherwood <david.sherwood@arm.com>
24108 * tree-vect-loop.c (vectorizable_induction): Treat the number
24109 of units as polynomial. Punt on SLP inductions. Use an integer
24110 VEC_SERIES_EXPR for variable-length integer reductions. Use a
24111 cast of such a series for variable-length floating-point
24114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24115 Alan Hayward <alan.hayward@arm.com>
24116 David Sherwood <david.sherwood@arm.com>
24118 * tree.h (build_index_vector): Declare.
24119 * tree.c (build_index_vector): New function.
24120 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
24121 of units as polynomial, forcibly converting it to a constant if
24122 vectorizable_reduction has already enforced the condition.
24123 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
24124 to create a {1,2,3,...} vector.
24125 (vectorizable_reduction): Treat the number of units as polynomial.
24126 Choose vectype_in based on the largest scalar element size rather
24127 than the smallest number of units. Enforce the restrictions
24130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24131 Alan Hayward <alan.hayward@arm.com>
24132 David Sherwood <david.sherwood@arm.com>
24134 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
24135 number of units as polynomial.
24137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24138 Alan Hayward <alan.hayward@arm.com>
24139 David Sherwood <david.sherwood@arm.com>
24141 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
24142 * target.def (autovectorize_vector_sizes): Return the vector sizes
24143 by pointer, using vector_sizes rather than a bitmask.
24144 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
24145 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
24146 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
24148 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
24149 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
24150 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
24151 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
24152 * omp-general.c (omp_max_vf): Likewise.
24153 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24154 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
24155 * tree-vect-loop.c (vect_analyze_loop): Likewise.
24156 * tree-vect-slp.c (vect_slp_bb): Likewise.
24157 * doc/tm.texi: Regenerate.
24158 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
24160 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
24161 the vector size as a poly_uint64 rather than an unsigned int.
24162 (current_vector_size): Change from an unsigned int to a poly_uint64.
24163 (get_vectype_for_scalar_type): Update accordingly.
24164 * tree.h (build_truth_vector_type): Take the size and number of
24165 units as a poly_uint64 rather than an unsigned int.
24166 (build_vector_type): Add a temporary overload that takes
24167 the number of units as a poly_uint64 rather than an unsigned int.
24168 * tree.c (make_vector_type): Likewise.
24169 (build_truth_vector_type): Take the number of units as a poly_uint64
24170 rather than an unsigned int.
24172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24173 Alan Hayward <alan.hayward@arm.com>
24174 David Sherwood <david.sherwood@arm.com>
24176 * target.def (get_mask_mode): Take the number of units and length
24177 as poly_uint64s rather than unsigned ints.
24178 * targhooks.h (default_get_mask_mode): Update accordingly.
24179 * targhooks.c (default_get_mask_mode): Likewise.
24180 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
24181 * doc/tm.texi: Regenerate.
24183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24184 Alan Hayward <alan.hayward@arm.com>
24185 David Sherwood <david.sherwood@arm.com>
24187 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
24188 * omp-general.c (omp_max_vf): Likewise.
24189 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
24190 (expand_omp_simd): Handle polynomial safelen.
24191 * omp-low.c (omplow_simd_context): Add a default constructor.
24192 (omplow_simd_context::max_vf): Change from int to poly_uint64.
24193 (lower_rec_simd_input_clauses): Update accordingly.
24194 (lower_rec_input_clauses): Likewise.
24196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24197 Alan Hayward <alan.hayward@arm.com>
24198 David Sherwood <david.sherwood@arm.com>
24200 * tree-vectorizer.h (vect_nunits_for_cost): New function.
24201 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
24202 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
24203 (vect_analyze_slp_cost): Likewise.
24204 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
24205 (vect_model_load_cost): Likewise.
24207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24208 Alan Hayward <alan.hayward@arm.com>
24209 David Sherwood <david.sherwood@arm.com>
24211 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
24212 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
24213 from an unsigned int * to a poly_uint64_pod *.
24214 (calculate_unrolling_factor): New function.
24215 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
24217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24218 Alan Hayward <alan.hayward@arm.com>
24219 David Sherwood <david.sherwood@arm.com>
24221 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
24222 from an unsigned int to a poly_uint64.
24223 (_loop_vec_info::slp_unrolling_factor): Likewise.
24224 (_loop_vec_info::vectorization_factor): Change from an int
24226 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
24227 (vect_get_num_vectors): New function.
24228 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
24229 (vect_get_num_copies): Use vect_get_num_vectors.
24230 (vect_analyze_data_ref_dependences): Change max_vf from an int *
24231 to an unsigned int *.
24232 (vect_analyze_data_refs): Change min_vf from an int * to a
24234 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24235 than an unsigned HOST_WIDE_INT.
24236 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
24237 (vect_analyze_data_ref_dependence): Change max_vf from an int *
24238 to an unsigned int *.
24239 (vect_analyze_data_ref_dependences): Likewise.
24240 (vect_compute_data_ref_alignment): Handle polynomial vf.
24241 (vect_enhance_data_refs_alignment): Likewise.
24242 (vect_prune_runtime_alias_test_list): Likewise.
24243 (vect_shift_permute_load_chain): Likewise.
24244 (vect_supportable_dr_alignment): Likewise.
24245 (dependence_distance_ge_vf): Take the vectorization factor as a
24246 poly_uint64 rather than an unsigned HOST_WIDE_INT.
24247 (vect_analyze_data_refs): Change min_vf from an int * to a
24249 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
24250 vfm1 as a poly_uint64 rather than an int. Make the same change
24251 for the returned bound_scalar.
24252 (vect_gen_vector_loop_niters): Handle polynomial vf.
24253 (vect_do_peeling): Likewise. Update call to
24254 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
24255 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
24257 * tree-vect-loop.c (vect_determine_vectorization_factor)
24258 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
24259 (vect_get_known_peeling_cost): Likewise.
24260 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
24261 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
24262 (vect_transform_loop): Likewise. Use the lowest possible VF when
24263 updating the upper bounds of the loop.
24264 (vect_min_worthwhile_factor): Make static. Return an unsigned int
24265 rather than an int.
24266 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
24267 polynomial unroll factors.
24268 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
24269 (vect_make_slp_decision): Likewise.
24270 (vect_supported_load_permutation_p): Likewise, and polynomial
24272 (vect_analyze_slp_cost): Handle polynomial vf.
24273 (vect_slp_analyze_node_operations): Likewise.
24274 (vect_slp_analyze_bb_1): Likewise.
24275 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
24276 than an unsigned HOST_WIDE_INT.
24277 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
24278 (vectorizable_load): Handle polynomial vf.
24279 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
24281 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
24283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24284 Alan Hayward <alan.hayward@arm.com>
24285 David Sherwood <david.sherwood@arm.com>
24287 * match.pd: Handle bit operations involving three constants
24288 and try to fold one pair.
24290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24292 * tree-vect-loop-manip.c: Include gimple-fold.h.
24293 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
24294 niters_maybe_zero parameters. Handle other cases besides a step of 1.
24295 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
24296 Add a path that uses a step of VF instead of 1, but disable it
24298 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
24299 and niters_no_overflow parameters. Update calls to
24300 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
24301 Create a new SSA name if the latter choses to use a ste other
24302 than zero, and return it via niters_vector_mult_vf_var.
24303 * tree-vect-loop.c (vect_transform_loop): Update calls to
24304 vect_do_peeling, vect_gen_vector_loop_niters and
24305 slpeel_make_loop_iterate_ntimes.
24306 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
24307 (vect_gen_vector_loop_niters): Update declarations after above changes.
24309 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
24311 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
24312 128-bit round to integer instructions.
24313 (ceil<mode>2): Likewise.
24314 (btrunc<mode>2): Likewise.
24315 (round<mode>2): Likewise.
24317 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24319 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
24320 unaligned VSX load/store on P8/P9.
24321 (expand_block_clear): Allow the use of unaligned VSX
24322 load/store on P8/P9.
24324 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24326 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
24328 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
24329 swap associated with both a load and a store.
24331 2018-01-02 Andrew Waterman <andrew@sifive.com>
24333 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
24334 * config/riscv/riscv.md (clear_cache): Use it.
24336 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
24338 * web.c: Remove out-of-date comment.
24340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24342 * expr.c (fixup_args_size_notes): Check that any existing
24343 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
24344 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
24345 (emit_single_push_insn): ...here.
24347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24349 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
24350 (const_vector_encoded_nelts): New function.
24351 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
24352 (const_vector_int_elt, const_vector_elt): Declare.
24353 * emit-rtl.c (const_vector_int_elt_1): New function.
24354 (const_vector_elt): Likewise.
24355 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
24356 of CONST_VECTOR_ELT.
24358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24360 * expr.c: Include rtx-vector-builder.h.
24361 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
24362 directly on the tree encoding.
24363 (const_vector_from_tree): Likewise.
24364 * optabs.c: Include rtx-vector-builder.h.
24365 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
24366 sequence of "u" values.
24367 * vec-perm-indices.c: Include rtx-vector-builder.h.
24368 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
24369 directly on the vec_perm_indices encoding.
24371 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24373 * doc/rtl.texi (const_vector): Describe new encoding scheme.
24374 * Makefile.in (OBJS): Add rtx-vector-builder.o.
24375 * rtx-vector-builder.h: New file.
24376 * rtx-vector-builder.c: Likewise.
24377 * rtl.h (rtx_def::u2): Add a const_vector field.
24378 (CONST_VECTOR_NPATTERNS): New macro.
24379 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
24380 (CONST_VECTOR_DUPLICATE_P): Likewise.
24381 (CONST_VECTOR_STEPPED_P): Likewise.
24382 (CONST_VECTOR_ENCODED_ELT): Likewise.
24383 (const_vec_duplicate_p): Check for a duplicated vector encoding.
24384 (unwrap_const_vec_duplicate): Likewise.
24385 (const_vec_series_p): Check for a non-duplicated vector encoding.
24386 Say that the function only returns true for integer vectors.
24387 * emit-rtl.c: Include rtx-vector-builder.h.
24388 (gen_const_vec_duplicate_1): Delete.
24389 (gen_const_vector): Call gen_const_vec_duplicate instead of
24390 gen_const_vec_duplicate_1.
24391 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
24392 (gen_const_vec_duplicate): Use rtx_vector_builder.
24393 (gen_const_vec_series): Likewise.
24394 (gen_rtx_CONST_VECTOR): Likewise.
24395 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
24396 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24397 Build a new vector rather than modifying a CONST_VECTOR in-place.
24398 (handle_special_swappables): Update call accordingly.
24399 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
24400 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
24401 Build a new vector rather than modifying a CONST_VECTOR in-place.
24402 (handle_special_swappables): Update call accordingly.
24404 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24406 * simplify-rtx.c (simplify_const_binary_operation): Use
24407 CONST_VECTOR_ELT instead of XVECEXP.
24409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24411 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
24412 the selector elements to be different from the data elements
24413 if the selector is a VECTOR_CST.
24414 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
24415 ssizetype for the selector.
24417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24419 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
24420 before testing each element individually.
24421 * tree-vect-generic.c (lower_vec_perm): Likewise.
24423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24425 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
24426 * selftest-run-tests.c (selftest::run_tests): Call it.
24427 * vector-builder.h (vector_builder::operator ==): New function.
24428 (vector_builder::operator !=): Likewise.
24429 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
24430 (vec_perm_indices::all_from_input_p): New function.
24431 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24432 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
24433 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
24434 instead of reading the VECTOR_CST directly. Detect whether both
24435 vector inputs are the same before constructing the vec_perm_indices,
24436 and update the number of inputs argument accordingly. Use the
24437 utility functions added above. Only construct sel2 if we need to.
24439 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24441 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
24442 the broadcast of the low byte.
24443 (expand_mult_highpart): Use an explicit encoding for the permutes.
24444 * optabs-query.c (can_mult_highpart_p): Likewise.
24445 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
24446 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24447 (vectorizable_bswap): Likewise.
24448 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
24449 explicit encoding for the power-of-2 permutes.
24450 (vect_permute_store_chain): Likewise.
24451 (vect_grouped_load_supported): Likewise.
24452 (vect_permute_load_chain): Likewise.
24454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24456 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
24457 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
24458 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
24459 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24460 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
24461 (vect_gen_perm_mask_any): Likewise.
24463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24465 * int-vector-builder.h: New file.
24466 * vec-perm-indices.h: Include int-vector-builder.h.
24467 (vec_perm_indices): Redefine as an int_vector_builder.
24468 (auto_vec_perm_indices): Delete.
24469 (vec_perm_builder): Redefine as a stand-alone class.
24470 (vec_perm_indices::vec_perm_indices): New function.
24471 (vec_perm_indices::clamp): Likewise.
24472 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
24473 (vec_perm_indices::new_vector): New function.
24474 (vec_perm_indices::new_expanded_vector): Update for new
24475 vec_perm_indices class.
24476 (vec_perm_indices::rotate_inputs): New function.
24477 (vec_perm_indices::all_in_range_p): Operate directly on the
24478 encoded form, without computing elided elements.
24479 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
24480 encoding. Update for new vec_perm_indices class.
24481 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
24482 the given vec_perm_builder.
24483 (expand_vec_perm_var): Update vec_perm_builder constructor.
24484 (expand_mult_highpart): Use vec_perm_builder instead of
24485 auto_vec_perm_indices.
24486 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
24487 vec_perm_indices instead of auto_vec_perm_indices. Use a single
24488 or double series encoding as appropriate.
24489 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
24490 vec_perm_indices instead of auto_vec_perm_indices.
24491 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24492 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24493 (vect_permute_store_chain): Likewise.
24494 (vect_grouped_load_supported): Likewise.
24495 (vect_permute_load_chain): Likewise.
24496 (vect_shift_permute_load_chain): Likewise.
24497 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24498 (vect_transform_slp_perm_load): Likewise.
24499 (vect_schedule_slp_instance): Likewise.
24500 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24501 (vectorizable_mask_load_store): Likewise.
24502 (vectorizable_bswap): Likewise.
24503 (vectorizable_store): Likewise.
24504 (vectorizable_load): Likewise.
24505 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
24506 vec_perm_indices instead of auto_vec_perm_indices. Use
24507 tree_to_vec_perm_builder to read the vector from a tree.
24508 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
24509 vec_perm_builder instead of a vec_perm_indices.
24510 (have_whole_vector_shift): Use vec_perm_builder and
24511 vec_perm_indices instead of auto_vec_perm_indices. Leave the
24512 truncation to calc_vec_perm_mask_for_shift.
24513 (vect_create_epilog_for_reduction): Likewise.
24514 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
24515 from auto_vec_perm_indices to vec_perm_indices.
24516 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24517 instead of changing individual elements.
24518 (aarch64_vectorize_vec_perm_const): Use new_vector to install
24519 the vector in d.perm.
24520 * config/arm/arm.c (expand_vec_perm_d::perm): Change
24521 from auto_vec_perm_indices to vec_perm_indices.
24522 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
24523 instead of changing individual elements.
24524 (arm_vectorize_vec_perm_const): Use new_vector to install
24525 the vector in d.perm.
24526 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
24527 Update vec_perm_builder constructor.
24528 (rs6000_expand_interleave): Likewise.
24529 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
24530 (rs6000_expand_interleave): Likewise.
24532 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24534 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
24535 to qimode could truncate the indices.
24536 * optabs.c (expand_vec_perm_var): Likewise.
24538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24540 * Makefile.in (OBJS): Add vec-perm-indices.o.
24541 * vec-perm-indices.h: New file.
24542 * vec-perm-indices.c: Likewise.
24543 * target.h (vec_perm_indices): Replace with a forward class
24545 (auto_vec_perm_indices): Move to vec-perm-indices.h.
24546 * optabs.h: Include vec-perm-indices.h.
24547 (expand_vec_perm): Delete.
24548 (selector_fits_mode_p, expand_vec_perm_var): Declare.
24549 (expand_vec_perm_const): Declare.
24550 * target.def (vec_perm_const_ok): Replace with...
24551 (vec_perm_const): ...this new hook.
24552 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
24553 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
24554 * doc/tm.texi: Regenerate.
24555 * optabs.def (vec_perm_const): Delete.
24556 * doc/md.texi (vec_perm_const): Likewise.
24557 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
24558 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
24559 expand_vec_perm for constant permutation vectors. Assert that
24560 the mode of variable permutation vectors is the integer equivalent
24561 of the mode that is being permuted.
24562 * optabs-query.h (selector_fits_mode_p): Declare.
24563 * optabs-query.c: Include vec-perm-indices.h.
24564 (selector_fits_mode_p): New function.
24565 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
24566 is defined, instead of checking whether the vec_perm_const_optab
24567 exists. Use targetm.vectorize.vec_perm_const instead of
24568 targetm.vectorize.vec_perm_const_ok. Check whether the indices
24569 fit in the vector mode before using a variable permute.
24570 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
24571 vec_perm_indices instead of an rtx.
24572 (expand_vec_perm): Replace with...
24573 (expand_vec_perm_const): ...this new function. Take the selector
24574 as a vec_perm_indices rather than an rtx. Also take the mode of
24575 the selector. Update call to shift_amt_for_vec_perm_mask.
24576 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
24577 Use vec_perm_indices::new_expanded_vector to expand the original
24578 selector into bytes. Check whether the indices fit in the vector
24579 mode before using a variable permute.
24580 (expand_vec_perm_var): Make global.
24581 (expand_mult_highpart): Use expand_vec_perm_const.
24582 * fold-const.c: Includes vec-perm-indices.h.
24583 * tree-ssa-forwprop.c: Likewise.
24584 * tree-vect-data-refs.c: Likewise.
24585 * tree-vect-generic.c: Likewise.
24586 * tree-vect-loop.c: Likewise.
24587 * tree-vect-slp.c: Likewise.
24588 * tree-vect-stmts.c: Likewise.
24589 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
24591 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
24592 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
24593 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
24594 (aarch64_vectorize_vec_perm_const): ...this new function.
24595 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24596 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24597 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
24598 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
24599 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24601 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
24603 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
24604 check for NEON modes.
24605 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
24606 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
24607 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
24608 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
24610 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
24611 the old VEC_PERM_CONST conditions.
24612 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
24613 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
24614 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
24615 (ia64_vectorize_vec_perm_const_ok): Merge into...
24616 (ia64_vectorize_vec_perm_const): ...this new function.
24617 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
24618 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
24619 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
24620 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
24621 * config/mips/mips.c (mips_expand_vec_perm_const)
24622 (mips_vectorize_vec_perm_const_ok): Merge into...
24623 (mips_vectorize_vec_perm_const): ...this new function.
24624 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
24625 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
24626 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
24627 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
24628 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
24629 (rs6000_expand_vec_perm_const): Delete.
24630 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
24632 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24633 (altivec_expand_vec_perm_const_le): Take each operand individually.
24634 Operate on constant selectors rather than rtxes.
24635 (altivec_expand_vec_perm_const): Likewise. Update call to
24636 altivec_expand_vec_perm_const_le.
24637 (rs6000_expand_vec_perm_const): Delete.
24638 (rs6000_vectorize_vec_perm_const_ok): Delete.
24639 (rs6000_vectorize_vec_perm_const): New function.
24640 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24641 an element count and rtx array.
24642 (rs6000_expand_extract_even): Update call accordingly.
24643 (rs6000_expand_interleave): Likewise.
24644 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
24645 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
24646 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
24647 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
24648 (rs6000_expand_vec_perm_const): Delete.
24649 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
24650 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24651 (altivec_expand_vec_perm_const_le): Take each operand individually.
24652 Operate on constant selectors rather than rtxes.
24653 (altivec_expand_vec_perm_const): Likewise. Update call to
24654 altivec_expand_vec_perm_const_le.
24655 (rs6000_expand_vec_perm_const): Delete.
24656 (rs6000_vectorize_vec_perm_const_ok): Delete.
24657 (rs6000_vectorize_vec_perm_const): New function. Remove stray
24658 reference to the SPE evmerge intructions.
24659 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
24660 an element count and rtx array.
24661 (rs6000_expand_extract_even): Update call accordingly.
24662 (rs6000_expand_interleave): Likewise.
24663 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
24664 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
24666 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
24668 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24670 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
24671 vector mode and that that mode matches the mode of the data
24673 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
24674 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
24675 directly using expand_vec_perm_1 when forcing selectors into
24677 (expand_vec_perm_var): New function, split out from expand_vec_perm.
24679 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24681 * optabs-query.h (can_vec_perm_p): Delete.
24682 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
24683 * optabs-query.c (can_vec_perm_p): Split into...
24684 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
24685 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
24686 particular selector is valid.
24687 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24688 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24689 (vect_grouped_load_supported): Likewise.
24690 (vect_shift_permute_load_chain): Likewise.
24691 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
24692 (vect_transform_slp_perm_load): Likewise.
24693 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24694 (vectorizable_bswap): Likewise.
24695 (vect_gen_perm_mask_checked): Likewise.
24696 * fold-const.c (fold_ternary_loc): Likewise. Don't take
24697 implementations of variable permutation vectors into account
24698 when deciding which selector to use.
24699 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
24700 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
24701 with a false third argument.
24702 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
24703 to test whether the constant selector is valid and can_vec_perm_var_p
24704 to test whether a variable selector is valid.
24706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24708 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
24709 * optabs-query.c (can_vec_perm_p): Likewise.
24710 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
24711 instead of vec_perm_indices.
24712 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
24713 (vect_gen_perm_mask_checked): Likewise,
24714 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
24715 (vect_gen_perm_mask_checked): Likewise,
24717 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
24719 * optabs-query.h (qimode_for_vec_perm): Declare.
24720 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
24721 (qimode_for_vec_perm): ...this new function.
24722 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
24724 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24726 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
24727 does not have a conditional at the top.
24729 2018-01-02 Richard Biener <rguenther@suse.de>
24731 * ipa-inline.c (big_speedup_p): Fix expression.
24733 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24736 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
24739 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
24743 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
24744 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
24745 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
24746 cond_taken_branch_cost 3->4.
24748 2018-01-01 Jakub Jelinek <jakub@redhat.com>
24750 PR tree-optimization/83581
24751 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
24752 TODO_cleanup_cfg if any changes have been made.
24754 PR middle-end/83608
24755 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
24756 convert_modes if target mode has the right side, but different mode
24759 PR middle-end/83609
24760 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
24761 last argument when extracting from CONCAT. If either from_real or
24762 from_imag is NULL, use expansion through memory. If result is not
24763 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
24764 the parts directly to inner mode, if even that fails, use expansion
24767 PR middle-end/83623
24768 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
24769 check for bswap in mode rather than HImode and use that in expand_unop
24772 Copyright (C) 2018 Free Software Foundation, Inc.
24774 Copying and distribution of this file, with or without modification,
24775 are permitted in any medium without royalty provided the copyright
24776 notice and this notice are preserved.