1 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4 bodies streamed in with -Q.
5 * dumpfile.c (dump_files): Add lto-stream-out dump file.
6 * dumpfile.h (tree_dump_index): Add lto_stream_out.
7 * gimple-streamer-out.c: Include gimple-pretty-print.h
8 (output_bb): Dump stmts streamed.
9 * lto-section-out.c: Include print-tree.h
10 (lto_begin_section): Dump sections created.
11 (lto_output_decl_index): Dump decl encoded.
12 * lto-streamer-out.c: Include print-tree.h
13 (create_output_block): Dump output block created.
14 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15 (output_function): Dump function output.
16 (output_constructor): Dump constructor streamed.
17 (write_global_stream): Output indexes encoded.
18 (produce_asm_for_decls): Dump streams encoded.
19 * lto-streamer.c (streamer_dump_file): New global var.
20 * lto-streamer.h (streamer_dump_file): Declare.
21 * passes.c (ipa_write_summaries): Initialize streamer dump.
22 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
25 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
28 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
29 offsets for register save directives. Emit a second batch of save
30 directives, if need be, when the function accesses prior frames.
32 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
34 * config/arc/fpu.md (fmasf4): Force operand to register.
37 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
39 * config/arc/arc-protos.h (arc_pad_return): Remove.
40 * config/arc/arc.c (machine_function): Remove force_short_suffix
42 (arc_print_operand): Adjust printing of '&'.
43 (arc_verify_short): Remove conditional printing of short suffix.
44 (arc_final_prescan_insn): Remove reference to size_reason.
45 (pad_return): New function.
46 (arc_reorg): Call pad_return.
47 (arc_pad_return): Remove.
48 (arc_init_machine_status): Remove reference to force_short_suffix.
49 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
50 (attr length): When attribute iscompact is true force to 2
51 regardless; in the case of maybe check if we want to force the
52 instruction to have 4 bytes length.
53 (nopv): Change it to generate 4 byte long nop as well.
54 (blockage): New pattern.
55 (simple_return): Remove call to arc_pad_return.
56 (p_return_i): Likewise.
58 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
60 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
62 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
64 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
67 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
69 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
72 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
75 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
76 operands[2] instead of operands[1].
79 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
81 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
82 case, check whether the outer register overlaps an unallocatable
83 register, not just whether it fits the required class.
85 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
87 * poly-int.h (can_div_trunc_p): Add new overload in which all values
89 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
90 (memrefs_conflict_p): Likewise.
91 (init_alias_analysis): Likewise.
92 * cfgexpand.c (expand_debug_expr): Likewise.
93 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
94 * cse.c (fold_rtx): Likewise.
95 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
96 * expr.c (emit_block_move_hints): Likewise.
97 (clear_storage_hints, push_block, emit_push_insn): Likewise.
98 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
99 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
100 (emit_group_store): Likewise.
101 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
102 to read the PRE/POST_MODIFY increment.
103 * calls.c (store_one_arg): Use strip_offset.
104 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
106 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
108 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
109 (simplify_binary_operation_1): Extend CONST_INT handling to
111 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
112 than a HOST_WIDE_INT.
113 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
115 (adjust_mems, add_stores): Update accodingly.
116 (vt_canonicalize_addr): Track polynomial offsets.
117 (emit_note_insn_var_location): Likewise.
118 (vt_add_function_parameter): Likewise.
119 (vt_initialize): Likewise.
121 2018-06-12 Jeff Law <law@redhat.com>
123 * config.gcc (alpha*-*-freebsd*): Remove.
124 * config/alpha/freebsd.h: Remove.
126 2018-06-12 David Malcolm <dmalcolm@redhat.com>
129 * spellcheck-tree.c (levenshtein_distance): Rename to...
130 (get_edit_distance): ...this, and update for underlying renaming.
131 * spellcheck-tree.h (levenshtein_distance): Rename to...
132 (get_edit_distance): ...this.
133 * spellcheck.c (levenshtein_distance): Rename to...
134 (get_edit_distance): ...this. Convert from Levenshtein distance
135 to Damerau-Levenshtein distance by supporting transpositions of
136 adjacent characters. Rename "v1" to "v_next" and "v0" to
138 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
139 (selftest::test_edit_distance_unit_test_oneway): ...this, and
140 update for underlying renaming.
141 (selftest::levenshtein_distance_unit_test): Rename to...
142 (selftest::test_get_edit_distance_unit): ...this, and update for
144 (selftest::test_find_closest_string): Add example from PR 69968
145 where transposition helps
146 (selftest::test_metric_conditions): Update for renaming.
147 (selftest::test_metric_conditions): Likewise.
148 (selftest::spellcheck_c_tests): Likewise.
149 * spellcheck.h (levenshtein_distance): Rename both overloads to...
150 (get_edit_distance): ...this.
151 (best_match::consider): Update for renaming.
153 2018-06-12 Martin Sebor <msebor@redhat.com>
155 PR tree-optimization/85259
156 * builtins.c (compute_objsize): Handle constant offsets.
157 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
158 true iff a warning has been issued.
159 * gimple.h (gimple_nonartificial_location): New function.
160 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
161 gimple_nonartificial_location and handle -Wno-system-headers.
162 (handle_builtin_stxncpy): Same.
164 2018-06-12 Martin Sebor <msebor@redhat.com>
167 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
169 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
171 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
172 BUILTIN_VEC_XST entries for pointer to double and long long.
174 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
177 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
179 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
182 2018-06-12 Martin Liska <mliska@suse.cz>
184 * doc/options.texi: Document IntegerRange.
186 2018-06-12 Martin Liska <mliska@suse.cz>
188 * config/i386/i386.opt: Make MPX-related options as Deprecated.
189 * opt-functions.awk: Handle Deprecated flag.
190 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
192 (read_cmdline_option): Report warning for a deprecated option.
193 * opts.h (struct cl_option): Add new field cl_deprecated.
194 (CL_ERR_DEPRECATED): New.
196 2018-06-12 Martin Liska <mliska@suse.cz>
198 * doc/options.texi: Document Deprecated option flag.
200 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
202 * config/arc/arc-arch.h (arc_extras): New enum.
203 (arc_cpu_t):Add field extra.
204 (arc_cpu_types): Consider the extras.
205 * config/arc/arc-cpus.def: Add extras info.
206 * config/arc/arc-opts.h (processor_type): Consider extra field.
207 * config/arc/arc.c (arc_override_options): Handle extra field.
209 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
211 * config/arc/arc-arch.h: Update ARC_OPTX macro.
212 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
214 * config/arc/arc.c (arc_init): Update pic warning.
215 (irq_range): Update irq range parsing warnings.
216 (arc_override_options): Update various warning messages.
217 (arc_handle_aux_attribute): Likewise.
219 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
221 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
223 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
225 * doc/sourcebuild.texi: Document usage of line number 0 in verify
226 compiler messages directives.
228 2018-06-12 Matthew Fortune <mfortune@gmail.com>
230 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
231 * config/mips/mips-tables.opt: Regenerate.
232 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
234 * doc/invoke.texi: Document -march=i6500.
236 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
238 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
239 (i6400_gpmul): Add cpu_unit.
240 (i6400_gpdiv): Likewise.
241 (i6400_msa_add_d): Update reservations.
242 (i6400_msa_int_add) Likewise.
243 (i6400_msa_short_logic3) Likewise.
244 (i6400_msa_short_logic2) Likewise.
245 (i6400_msa_short_logic) Likewise.
246 (i6400_msa_move) Likewise.
247 (i6400_msa_cmp) Likewise.
248 (i6400_msa_short_float2) Likewise.
249 (i6400_msa_div_d) Likewise.
250 (i6400_msa_long_logic1) Likewise.
251 (i6400_msa_long_logic2) Likewise.
252 (i6400_msa_mult) Likewise.
253 (i6400_msa_long_float2) Likewise.
254 (i6400_msa_long_float4) Likewise.
255 (i6400_msa_long_float5) Likewise.
256 (i6400_msa_long_float8) Likewise.
257 (i6400_fpu_fadd): Include frint type.
258 (i6400_fpu_store): New define_insn_reservation.
259 (i6400_fpu_load): Likewise.
260 (i6400_fpu_move): Likewise.
261 (i6400_fpu_fcmp): Likewise.
262 (i6400_fpu_fmadd): Likewise.
263 (i6400_int_mult): Include imul3nc type and update reservation.
264 (i6400_int_div): Include idiv3 type and update reservation.
265 (i6400_int_load): Update to check type not move_type.
266 (i6400_int_store): Likewise.
267 (i6400_int_prefetch): Set zero latency.
269 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
271 * gcc.c: Document new %@{...} sequence.
272 (LINK_COMMAND_SPEC): Use it for the -L switches.
273 (cpp_unique_options): Use it for the -I switches.
274 (at_file_argbuf): New global variable.
275 (in_at_file): Likewise.
276 (alloc_args): Create at_file_argbuf.
277 (clear_args): Truncate at_file_argbuf.
278 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
279 (open_at_file): New function.
280 (close_at_file): Likewise.
281 (create_at_file): Delete.
282 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
285 (validate_switches_from_spec): Deal with %@{...} sequence.
286 (validate_switches): Likewise.
287 (driver::finalize): Call clear_args.
289 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
291 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
293 2018-06-11 Martin Sebor <msebor@redhat.com>
295 * doc/invoke.texi (-Wall): List -Wc++17-compat.
296 (Wno-class-memaccess): Add @opindex.
297 (Wno-templates, Wno-multiple-inheritance): Same.
298 (Wno-virtual-inheritance, Wno-namespaces): Same.
299 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
300 (Wno-format-overflow, Wno-format-truncation): Same.
301 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
302 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
303 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
304 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
305 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
306 (Wno-misspelled-isr): Same.
308 2018-06-11 Martin Sebor <msebor@redhat.com>
310 * PR tree-optimization/86083
311 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
313 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
315 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
317 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
320 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
321 on the correct operand.
322 (*movdi_internal64): Ditto.
324 2018-06-11 Martin Liska <mliska@suse.cz>
326 PR tree-optimization/86089
327 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
329 2018-06-11 Julia Koval <julia.koval@intel.com>
331 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
332 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
333 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
335 2018-06-11 Olivier Hainque <hainque@adacore.com>
337 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
338 for Ada with strict dwarf2.
340 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
343 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
346 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
348 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
350 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
352 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
355 2018-06-08 Martin Liska <mliska@suse.cz>
357 * tree-cfg.h (debug_function): Fix argument type to match
360 2018-06-08 Martin Liska <mliska@suse.cz>
362 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
363 Remove usage of MPX-related (and removed) fields.
364 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
366 2018-06-08 David Malcolm <dmalcolm@redhat.com>
368 * cfg.c (debug): Use TDF_NONE rather than 0.
369 * cfghooks.c (debug): Likewise.
370 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
371 (struct dump_option_value_info): Convert to...
372 (struct kv_pair): ...this template type.
373 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
375 (optinfo_verbosity_options): Likewise.
376 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
378 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
379 than int for "optgroup_flags" param.
380 (dump_generic_expr_loc): Use dump_flags_t rather than int for
382 (dump_dec): Likewise.
383 (dump_finish): Use TDF_NONE rather than 0.
384 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
385 rather than int for "optgroup_flags" param. Use TDF_NONE rather
386 than 0. Update for change to option_ptr.
387 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
388 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
389 0. Update for changes to optinfo_verbosity_options and
391 (opt_info_switch_p): Convert optgroup_flags from int to
393 (dump_basic_block): Use dump_flags_t rather than int
394 for "dump_kind" param.
395 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
396 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
397 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
398 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
399 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
400 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
401 TDF_NONE): Convert from macros to...
402 (enum dump_flag): ...this new enum.
403 (dump_flags_t): Update to use enum.
404 (operator|, operator&, operator~, operator|=, operator&=):
405 Implement for dump_flags_t.
406 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
407 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
408 Convert from macros to...
409 (enum optgroup_flag): ...this new enum.
410 (optgroup_flags_t): New typedef.
411 (operator|, operator|=): Implement for optgroup_flags_t.
412 (struct dump_file_info): Convert field "alt_flags" to
413 dump_flags_t. Convert field "optgroup_flags" to
415 (dump_basic_block): Use dump_flags_t rather than int for param.
416 (dump_generic_expr_loc): Likewise.
417 (dump_dec): Likewise.
418 (dump_register): Convert param "optgroup_flags" to
420 (opt_info_enable_passes): Likewise.
421 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
423 * gimple-pretty-print.c (debug): Likewise.
424 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
425 (merged_store_group::apply_stores): Likewise.
426 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
427 * gimple.c (verify_gimple_pp): Likewise.
428 * graphite-poly.c (print_pbb_body): Likewise.
429 * passes.c (pass_manager::register_one_dump_file): Convert
430 local "optgroup_flags" to optgroup_flags_t.
431 * print-tree.c (print_node): Use TDF_NONE rather than 0.
433 (debug_body): Likewise.
434 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
436 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
438 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
439 (convert_mult_to_fma): Likewise.
440 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
441 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
442 * tree-vect-data-refs.c (dump_lower_bound): Convert param
443 "dump_kind" to dump_flags_t.
445 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
447 * config/rs6000/rs6000.c (min, max): Delete.
449 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
451 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
452 -mabi=spe and -mabi=no-spe.
454 2018-06-08 Martin Liska <mliska@suse.cz>
456 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
457 where we expect an existing summary.
459 2018-06-08 Martin Liska <mliska@suse.cz>
461 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
462 * ipa-inline.h (estimate_edge_growth): Likewise.
464 2018-06-08 Martin Liska <mliska@suse.cz>
466 * cgraph.c (function_version_hasher::hash): Use
467 cgraph_node::get_uid ().
468 (function_version_hasher::equal):
469 * cgraph.h (cgraph_node::get_uid): New method.
470 * ipa-inline.c (update_caller_keys): Use
471 cgraph_node::get_uid ().
472 (update_callee_keys): Likewise.
473 * ipa-utils.c (searchc): Likewise.
474 (ipa_reduced_postorder): Likewise.
475 * lto-cgraph.c (input_node): Likewise.
476 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
477 * symbol-summary.h (symtab_insertion): Likewise.
478 (symtab_removal): Likewise.
479 (symtab_duplication): Likewise.
480 * tree-pretty-print.c (dump_function_header): Likewise.
481 * tree-sra.c (convert_callers_for_node): Likewise.
483 2018-06-08 Martin Liska <mliska@suse.cz>
485 * cgraph.c (symbol_table::create_edge): Always assign a new
487 (symbol_table::free_edge): Do not recycle numbers.
488 * cgraph.h (cgraph_edge::get): New method.
489 * symbol-summary.h (symtab_removal): Use it.
490 (symtab_duplication): Likewise.
491 (call_summary::hashable_uid): Remove.
493 2018-06-08 Martin Liska <mliska@suse.cz>
495 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
496 (initialize_growth_caches): Remove.
497 (free_growth_caches): Likewise.
498 (do_estimate_edge_time): Use edge_growth_cache.
499 (do_estimate_edge_size): Likewise.
500 (do_estimate_edge_hints): Likewise.
501 * ipa-inline.c (reset_edge_caches): Likewise.
502 (recursive_inlining): Likewise.
503 (inline_small_functions): Likewise.
504 * ipa-inline.h (initialize_growth_caches): Remove.
505 (estimate_edge_size): Likewise.
506 (estimate_edge_time): Likewise.
507 (estimate_edge_hints): Likewise.
508 (reset_edge_growth_cache): Likewise.
509 * symbol-summary.h (call_summary::remove): New method.
511 2018-06-08 Martin Liska <mliska@suse.cz>
513 * ipa-cp.c (class edge_clone_summary): New summary.
514 (grow_edge_clone_vectors): Remove.
515 (ipcp_edge_duplication_hook): Remove.
516 (class edge_clone_summary_t): New call_summary class.
517 (ipcp_edge_removal_hook): Remove.
518 (edge_clone_summary_t::duplicate): New function.
519 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
520 (create_specialized_node): Likewise.
521 (ipcp_driver): Initialize edge_clone_summaries and do not
524 2018-06-08 Martin Liska <mliska@suse.cz>
526 * symbol-summary.h (get): New function.
527 (call_summary::m_initialize_when_cloning): New class member.
529 2018-06-08 Martin Liska <mliska@suse.cz>
531 * cgraph.c (cgraph_node::remove): Do not recycle uid.
532 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
533 (symbol_table::allocate_cgraph_symbol): Do not set uid.
534 * passes.c (uid_hash_t): Record removed_nodes by their uids.
535 (remove_cgraph_node_from_order): Use the removed_nodes set.
536 (do_per_function_toporder): Likwise.
537 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
538 instead of summary_uid.
539 (symtab_removal): Likewise.
540 (symtab_duplication): Likewise.
542 2018-06-08 Martin Liska <mliska@suse.cz>
544 * ipa-cp.c (ipcp_store_bits_results): Use
545 ipcp_transformation_sum.
546 (ipcp_store_vr_results): Likewise.
547 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
549 (ipcp_transformation_initialize): ... this.
550 (ipa_set_node_agg_value_chain):
551 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
552 (write_ipcp_transformation_info): Likewise.
553 (read_ipcp_transformation_info): Likewise.
554 (ipcp_update_bits): Likewise.
555 (ipcp_update_vr): Likewise.
556 (ipcp_transform_function): Likewise.
557 * ipa-prop.h: Rename ipcp_transformation_summary to
559 (class ipcp_transformation_t): New function summary.
560 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
561 (ipa_get_agg_replacements_for_node): Likewise.
563 2018-06-08 Martin Liska <mliska@suse.cz>
565 * ipa-pure-const.c (struct funct_state_d): Do it class instead
567 (class funct_state_summary_t): New function_summary class.
568 (has_function_state): Remove.
569 (get_function_state): Likewise.
570 (set_function_state): Likewise.
571 (add_new_function): Likewise.
572 (funct_state_summary_t::insert): New function.
573 (duplicate_node_data): Remove.
574 (remove_node_data): Remove.
575 (funct_state_summary_t::duplicate): New function.
576 (register_hooks): Create new funct_state_summaries.
577 (pure_const_generate_summary): Use it.
578 (pure_const_write_summary): Likewise.
579 (pure_const_read_summary): Likewise.
580 (propagate_pure_const): Likewise.
581 (propagate_nothrow): Likewise.
582 (dump_malloc_lattice): Likewise.
583 (propagate_malloc): Likewise.
584 (execute): Do not register hooks, just remove summary
586 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
589 2018-06-08 Martin Liska <mliska@suse.cz>
591 * ipa-reference.c (remove_node_data): Remove.
592 (duplicate_node_data): Likewise.
593 (class ipa_ref_var_info_summary_t): New class.
594 (class ipa_ref_opt_summary_t): Likewise.
595 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
596 (get_reference_optimization_summary): Use
597 ipa_ref_opt_sum_summaries.
598 (set_reference_vars_info): Remove.
599 (set_reference_optimization_summary): Likewise.
600 (ipa_init): Create summaries.
601 (init_function_info): Use function summary.
602 (ipa_ref_opt_summary_t::duplicate): New function.
603 (ipa_ref_opt_summary_t::remove): New function.
604 (get_read_write_all_from_node): Fix GNU coding style.
605 (propagate): Use function summary.
606 (write_node_summary_p): Fix GNU coding style.
607 (stream_out_bitmap): Likewise.
608 (ipa_reference_read_optimization_summary): Use function summary.
609 (ipa_reference_c_finalize): Do not release hooks.
611 2018-06-08 Martin Liska <mliska@suse.cz>
613 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
614 (analyze_function_body): Extract multiple calls of get_create.
615 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
616 * ipa-inline.c (recursive_inlining): Use ::get method.
617 * ipa-inline.h (estimate_edge_growth): Likewise.
619 2018-06-08 Martin Liska <mliska@suse.cz>
621 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
623 (hsa_function_summary::hsa_function_summary): Use the new enum
625 (hsa_gpu_implementation_p): Use hsa_summaries::get.
626 * hsa-gen.c (hsa_get_host_function): Likewise.
627 (get_brig_function_name): Likewise.
628 * ipa-hsa.c (process_hsa_functions): Likewise.
629 (ipa_hsa_write_summary): Likewise.
630 * symbol-summary.h (symtab_duplication): Use ::get function/
633 2018-06-08 Martin Liska <mliska@suse.cz>
635 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
637 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
638 (hsa_register_kernel): Likewise.
639 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
640 * hsa-gen.c (hsa_get_host_function): Likewise.
641 (get_brig_function_name): Likewise.
642 (generate_hsa): Likewise.
643 (pass_gen_hsail::execute): Likewise.
644 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
645 (devirtualization_time_bonus): Likewise.
646 (ipcp_propagate_stage): Likewise.
647 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
648 (edge_set_predicate): Likewise.
649 (evaluate_conditions_for_known_args): Likewise.
650 (evaluate_properties_for_edge): Likewise.
651 (ipa_fn_summary::reset): Likewise.
652 (ipa_fn_summary_t::duplicate): Likewise.
653 (dump_ipa_call_summary): Likewise.
654 (ipa_dump_fn_summary): Likewise.
655 (analyze_function_body): Likewise.
656 (compute_fn_summary): Likewise.
657 (estimate_edge_devirt_benefit): Likewise.
658 (estimate_edge_size_and_time): Likewise.
659 (estimate_calls_size_and_time): Likewise.
660 (estimate_node_size_and_time): Likewise.
661 (inline_update_callee_summaries): Likewise.
662 (remap_edge_change_prob): Likewise.
663 (remap_edge_summaries): Likewise.
664 (ipa_merge_fn_summary_after_inlining): Likewise.
665 (ipa_update_overall_fn_summary): Likewise.
666 (read_ipa_call_summary): Likewise.
667 (inline_read_section): Likewise.
668 (write_ipa_call_summary): Likewise.
669 (ipa_fn_summary_write): Likewise.
670 (ipa_free_fn_summary): Likewise.
671 * ipa-hsa.c (process_hsa_functions): Likewise.
672 (ipa_hsa_write_summary): Likewise.
673 (ipa_hsa_read_section): Likewise.
674 * ipa-icf.c (sem_function::merge): Likewise.
675 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
676 (do_estimate_edge_time): Likewise.
677 (estimate_size_after_inlining): Likewise.
678 (estimate_growth): Likewise.
679 (growth_likely_positive): Likewise.
680 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
681 (inline_call): Likewise.
682 * ipa-inline.c (caller_growth_limits): Likewise.
683 (can_inline_edge_p): Likewise.
684 (can_inline_edge_by_limits_p): Likewise.
685 (compute_uninlined_call_time): Likewise.
686 (compute_inlined_call_time): Likewise.
687 (want_inline_small_function_p): Likewise.
688 (edge_badness): Likewise.
689 (update_caller_keys): Likewise.
690 (update_callee_keys): Likewise.
691 (recursive_inlining): Likewise.
692 (inline_small_functions): Likewise.
693 (inline_to_all_callers_1): Likewise.
694 (dump_overall_stats): Likewise.
695 (early_inline_small_functions): Likewise.
696 (early_inliner): Likewise.
697 * ipa-inline.h (estimate_edge_growth): Likewise.
698 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
699 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
700 * ipa-prop.h (IPA_NODE_REF): Likewise.
701 (IPA_EDGE_REF): Likewise.
702 * ipa-pure-const.c (malloc_candidate_p): Likewise.
703 (propagate_malloc): Likewise.
704 * ipa-split.c (execute_split_functions): Likewise.
705 * symbol-summary.h: Rename get to get_create.
707 (get_create): Likewise.
708 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
710 2018-06-08 Martin Liska <mliska@suse.cz>
712 * symbol-summary.h (release): Move definition out of class
714 (symtab_removal): Likewise.
715 (symtab_duplication): Likewise.
717 2018-06-08 Martin Liska <mliska@suse.cz>
719 * symbol-summary.h (function_summary): Move constructor
720 implementation out of class declaration.
722 (symtab_insertion): Likewise.
723 (symtab_removal): Likewise.
724 (symtab_duplication): Likewise.
727 2018-06-08 Martin Liska <mliska@suse.cz>
729 * Makefile.in: Remove support for MPX (macros, related functions,
730 fields in cgraph_node, ...).
731 * builtin-types.def (BT_BND): Likewise.
732 (BT_FN_BND_CONST_PTR): Likewise.
733 (BT_FN_CONST_PTR_BND): Likewise.
734 (BT_FN_VOID_PTR_BND): Likewise.
735 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
736 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
737 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
738 (expand_builtin_mempcpy_with_bounds): Likewise.
739 (expand_builtin_memset_with_bounds): Likewise.
740 (expand_builtin_memset_args): Likewise.
741 (std_expand_builtin_va_start): Likewise.
742 (expand_builtin): Likewise.
743 (expand_builtin_with_bounds): Likewise.
744 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
745 (DEF_LIB_BUILTIN_CHKP): Likewise.
746 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
747 (DEF_CHKP_BUILTIN): Likewise.
748 (BUILT_IN_MEMCPY): Likewise.
749 (BUILT_IN_MEMMOVE): Likewise.
750 (BUILT_IN_MEMPCPY): Likewise.
751 (BUILT_IN_MEMSET): Likewise.
752 (BUILT_IN_STPCPY): Likewise.
753 (BUILT_IN_STRCAT): Likewise.
754 (BUILT_IN_STRCHR): Likewise.
755 (BUILT_IN_STRCPY): Likewise.
756 (BUILT_IN_STRLEN): Likewise.
757 (BUILT_IN_MEMCPY_CHK): Likewise.
758 (BUILT_IN_MEMMOVE_CHK): Likewise.
759 (BUILT_IN_MEMPCPY_CHK): Likewise.
760 (BUILT_IN_MEMSET_CHK): Likewise.
761 (BUILT_IN_STPCPY_CHK): Likewise.
762 (BUILT_IN_STRCAT_CHK): Likewise.
763 (BUILT_IN_STRCPY_CHK): Likewise.
764 * calls.c (store_bounds): Likewise.
765 (emit_call_1): Likewise.
766 (special_function_p): Likewise.
767 (maybe_warn_nonstring_arg): Likewise.
768 (initialize_argument_information): Likewise.
769 (finalize_must_preallocate): Likewise.
770 (compute_argument_addresses): Likewise.
771 (expand_call): Likewise.
772 * cfgexpand.c (expand_call_stmt): Likewise.
773 (expand_return): Likewise.
774 (expand_gimple_stmt_1): Likewise.
775 (pass_expand::execute): Likewise.
776 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
777 (cgraph_node::remove): Likewise.
778 (cgraph_node::dump): Likewise.
779 (cgraph_node::verify_node): Likewise.
780 * cgraph.h (chkp_function_instrumented_p): Likewise.
781 (symtab_node::get_alias_target): Likewise.
782 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
783 (cgraph_local_p): Likewise.
784 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
785 (cgraph_edge::rebuild_references): Likewise.
786 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
787 (walk_polymorphic_call_targets): Likewise.
788 (cgraph_node::expand_thunk): Likewise.
789 (symbol_table::output_weakrefs): Likewise.
790 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
791 (ix86_handle_option): Likewise.
792 * config/i386/constraints.md: Likewise.
793 * config/i386/i386-builtin-types.def (BND): Likewise.
797 * config/i386/i386-builtin.def (BDESC_END): Likewise.
798 (BDESC_FIRST): Likewise.
800 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
801 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
802 * config/i386/i386.c (enum reg_class): Likewise.
803 (ix86_target_string): Likewise.
804 (ix86_option_override_internal): Likewise.
805 (ix86_conditional_register_usage): Likewise.
806 (ix86_valid_target_attribute_inner_p): Likewise.
807 (ix86_set_indirect_branch_type): Likewise.
808 (ix86_set_current_function): Likewise.
809 (ix86_function_arg_regno_p): Likewise.
810 (init_cumulative_args): Likewise.
811 (ix86_function_arg_advance): Likewise.
812 (ix86_function_arg): Likewise.
813 (ix86_pass_by_reference): Likewise.
814 (ix86_function_value_regno_p): Likewise.
815 (ix86_function_value_1): Likewise.
816 (ix86_function_value_bounds): Likewise.
817 (ix86_return_in_memory): Likewise.
818 (ix86_setup_incoming_vararg_bounds): Likewise.
819 (ix86_va_start): Likewise.
820 (indirect_thunk_need_prefix): Likewise.
821 (print_reg): Likewise.
822 (ix86_print_operand): Likewise.
823 (ix86_expand_call): Likewise.
824 (ix86_output_function_return): Likewise.
825 (reg_encoded_number): Likewise.
826 (BDESC_VERIFYS): Likewise.
827 (ix86_init_mpx_builtins): Likewise.
828 (ix86_init_builtins): Likewise.
829 (ix86_emit_cmove): Likewise.
830 (ix86_emit_move_max): Likewise.
831 (ix86_expand_builtin): Likewise.
832 (ix86_builtin_mpx_function): Likewise.
833 (ix86_get_arg_address_for_bt): Likewise.
834 (ix86_load_bounds): Likewise.
835 (ix86_store_bounds): Likewise.
836 (ix86_load_returned_bounds): Likewise.
837 (ix86_store_returned_bounds): Likewise.
838 (ix86_class_likely_spilled_p): Likewise.
839 (ix86_hard_regno_mode_ok): Likewise.
840 (x86_order_regs_for_local_alloc): Likewise.
841 (ix86_mitigate_rop): Likewise.
842 (ix86_bnd_prefixed_insn_p): Likewise.
843 (ix86_mpx_bound_mode): Likewise.
844 (ix86_make_bounds_constant): Likewise.
845 (ix86_initialize_bounds): Likewise.
846 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
847 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
848 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
849 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
850 (TARGET_CHKP_BOUND_MODE): Likewise.
851 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
852 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
853 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
854 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
855 * config/i386/i386.h (TARGET_MPX): Likewise.
856 (TARGET_MPX_P): Likewise.
857 (VALID_BND_REG_MODE): Likewise.
858 (FIRST_BND_REG): Likewise.
859 (LAST_BND_REG): Likewise.
860 (enum reg_class): Likewise.
861 (BND_REG_P): Likewise.
862 (BND_REGNO_P): Likewise.
864 (ADJUST_INSN_LENGTH): Likewise.
865 * config/i386/i386.md: Likewise.
866 * config/i386/i386.opt: Likewise.
867 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
869 (LINK_MPX): Likewise.
870 (MPX_SPEC): Likewise.
871 (LIBMPX_SPEC): Likewise.
872 (LIBMPXWRAPPERS_SPEC): Likewise.
873 (CHKP_SPEC): Likewise.
874 * config/i386/predicates.md: Likewise.
875 * dbxout.c (dbxout_type): Likewise.
876 * doc/extend.texi: Likewise.
877 * doc/invoke.texi: Likewise.
878 * doc/md.texi: Likewise.
879 * doc/tm.texi: Likewise.
880 * doc/tm.texi.in: Likewise.
881 * dwarf2out.c (is_base_type): Likewise.
882 (gen_formal_types_die): Likewise.
883 (gen_subprogram_die): Likewise.
884 (gen_type_die_with_usage): Likewise.
885 (gen_decl_die): Likewise.
886 (dwarf2out_late_global_decl): Likewise.
887 * expr.c (expand_assignment): Likewise.
888 (emit_storent_insn): Likewise.
889 (store_expr_with_bounds): Likewise.
890 (store_expr): Likewise.
891 (expand_expr_real_1): Likewise.
892 * expr.h (store_expr_with_bounds): Likewise.
893 * function.c (use_register_for_decl): Likewise.
894 (struct bounds_parm_data): Likewise.
895 (assign_parms_augmented_arg_list): Likewise.
896 (assign_parm_find_entry_rtl): Likewise.
897 (assign_parm_is_stack_parm): Likewise.
898 (assign_parm_load_bounds): Likewise.
899 (assign_bounds): Likewise.
900 (assign_parms): Likewise.
901 (expand_function_start): Likewise.
902 * gcc.c (CHKP_SPEC): Likewise.
903 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
904 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
905 (wrestrict_dom_walker::check_call): Likewise.
906 * gimple.c (gimple_build_call_from_tree): Likewise.
907 * gimple.h (enum gf_mask): Likewise.
908 (gimple_call_with_bounds_p): Likewise.
909 (gimple_call_set_with_bounds): Likewise.
910 * gimplify.c (gimplify_init_constructor): Likewise.
911 * ipa-cp.c (initialize_node_lattices): Likewise.
912 (propagate_constants_across_call): Likewise.
913 (find_more_scalar_values_for_callers_subset): Likewise.
914 * ipa-hsa.c (process_hsa_functions): Likewise.
915 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
916 * ipa-icf.c (sem_function::merge): Likewise.
917 * ipa-inline.c (early_inliner): Likewise.
918 * ipa-pure-const.c (warn_function_noreturn): Likewise.
919 (warn_function_cold): Likewise.
920 (propagate_pure_const): Likewise.
921 * ipa-ref.h (enum GTY): Likewise.
922 * ipa-split.c (find_retbnd): Likewise.
923 (consider_split): Likewise.
924 (split_function): Likewise.
925 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
926 * ipa.c (walk_polymorphic_call_targets): Likewise.
927 (symbol_table::remove_unreachable_nodes): Likewise.
928 (process_references): Likewise.
929 (cgraph_build_static_cdtor_1): Likewise.
930 * lto-cgraph.c (lto_output_node): Likewise.
931 (output_refs): Likewise.
932 (compute_ltrans_boundary): Likewise.
933 (input_overwrite_node): Likewise.
934 (input_node): Likewise.
935 (input_cgraph_1): Likewise.
936 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
937 * passes.c (pass_manager::execute_early_local_passes): Likewise.
938 (class pass_chkp_instrumentation_passes): Likewise.
939 (make_pass_chkp_instrumentation_passes): Likewise.
940 * passes.def: Likewise.
941 * rtl.h (struct GTY): Likewise.
942 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
943 * stor-layout.c (layout_type): Likewise.
944 * symtab.c: Likewise.
945 * target.def: Likewise.
946 * targhooks.c (default_chkp_bound_type): Likewise.
947 (default_chkp_bound_mode): Likewise.
948 (default_builtin_chkp_function): Likewise.
949 (default_chkp_function_value_bounds): Likewise.
950 (default_chkp_make_bounds_constant): Likewise.
951 (default_chkp_initialize_bounds): Likewise.
952 * targhooks.h (default_chkp_bound_type): Likewise.
953 (default_chkp_bound_mode): Likewise.
954 (default_builtin_chkp_function): Likewise.
955 (default_chkp_function_value_bounds): Likewise.
956 (default_chkp_make_bounds_constant): Likewise.
957 (default_chkp_initialize_bounds): Likewise.
958 * toplev.c (compile_file): Likewise.
959 (process_options): Likewise.
960 * tree-core.h (DEF_BUILTIN): Likewise.
961 (DEF_BUILTIN_CHKP): Likewise.
962 * tree-inline.c (declare_return_variable): Likewise.
963 (remap_gimple_stmt): Likewise.
965 (initialize_inlined_parameters): Likewise.
966 (expand_call_inline): Likewise.
967 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
968 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
969 (make_pass_ipa_chkp_produce_thunks): Likewise.
970 (make_pass_chkp): Likewise.
971 (make_pass_chkp_opt): Likewise.
972 (make_pass_chkp_instrumentation_passes): Likewise.
973 * tree-pretty-print.c (dump_generic_node): Likewise.
974 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
975 * tree-ssa-dce.c (propagate_necessity): Likewise.
976 (eliminate_unnecessary_stmts): Likewise.
977 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
978 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
979 * tree-ssa-sccvn.h: Likewise.
980 * tree-ssa-strlen.c (get_string_length): Likewise.
981 (valid_builtin_call): Likewise.
982 (adjust_last_stmt): Likewise.
983 (handle_builtin_strchr): Likewise.
984 (handle_builtin_strcpy): Likewise.
985 (handle_builtin_stxncpy): Likewise.
986 (handle_builtin_memcpy): Likewise.
987 (handle_builtin_strcat): Likewise.
988 (strlen_check_and_optimize_stmt): Likewise.
989 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
990 * tree-streamer-in.c: Likewise.
991 * tree-streamer.c (record_common_node): Likewise.
992 * tree.c (tree_code_size): Likewise.
993 (wide_int_to_tree_1): Likewise.
994 (type_contains_placeholder_1): Likewise.
995 (build_common_tree_nodes): Likewise.
996 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
997 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
998 (POINTER_BOUNDS_P): Likewise.
999 (BOUNDED_TYPE_P): Likewise.
1000 (BOUNDED_P): Likewise.
1001 (CALL_WITH_BOUNDS_P): Likewise.
1002 (pointer_bounds_type_node): Likewise.
1003 * value-prof.c (gimple_ic): Likewise.
1004 * var-tracking.c (vt_add_function_parameters): Likewise.
1005 * varasm.c (make_decl_rtl): Likewise.
1006 (assemble_start_function): Likewise.
1007 (output_constant): Likewise.
1008 (maybe_assemble_visibility): Likewise.
1009 * varpool.c (ctor_for_folding): Likewise.
1010 * chkp-builtins.def: Remove.
1011 * ipa-chkp.c: Remove.
1012 * ipa-chkp.h: Remove.
1013 * rtl-chkp.c: Remove.
1014 * rtl-chkp.h: Remove.
1015 * tree-chkp-opt.c: Remove.
1016 * tree-chkp.c: Remove.
1017 * tree-chkp.h: Remove.
1019 2018-06-07 Carl Love <cel@us.ibm.com>
1021 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1022 vextract_fp_from_shortl): Add BE support.
1024 2018-06-07 Paul Koning <ni1d@arrl.net>
1026 * compare-elim.c (try_merge_compare): Don't merge compare if
1027 address contains a side effect.
1028 (try_eliminate_compare): Likewise.
1030 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1032 * config.gcc: Support "tremont".
1033 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1034 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1036 * config/i386/i386.c (m_TREMONT): Define.
1037 (processor_target_table): Add "tremont".
1038 (PTA_TREMONT): Define.
1039 (ix86_lea_outperforms): Add TARGET_TREMONT.
1040 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1041 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1042 and M_INTEL_GOLDMONT_PLUS.
1043 (fold_builtin_cpu): Add "tremont".
1044 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1045 (ix86_option_override_internal): Add "tremont".
1046 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1047 (processor_type): Add PROCESSOR_TREMONT.
1048 * config/i386/x86-tune.def: Add m_TREMONT.
1049 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1051 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1053 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1054 symbol defined for msp430i* devices to be lower case.
1056 gcc/testsuite/gcc.target/msp430/
1057 * msp430i-device-symbol.c: New test.
1058 * msp430f-device-symbol.c: New test.
1059 * msp430.h: New test header file.
1061 2018-06-07 Richard Biener <rguenther@suse.de>
1063 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1064 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1065 Properly wrap signed arithmetic if overflow wraps.
1067 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1069 PR tree-optimization/69615
1070 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1071 of a cast from a same precision integral SSA_NAME in a bb dominated
1072 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1073 cast to utype if rhs2 has already a compatible type.
1075 2018-06-07 Richard Biener <rguenther@suse.de>
1077 PR tree-optimization/85935
1078 * graphite-scop-detection.c (find_params_in_bb): Analyze
1079 condition operands with respect to the correct loop. Assert
1080 the analysis doesn't fail.
1082 2018-06-04 Carl Love <cel@us.ibm.com>
1084 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1085 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1086 as it is slightly cheaper.
1087 (first_match_or_eos_index_<mode>):
1088 Calculate index using natural element order.
1089 (first_match_index_<mode>):
1090 Calculate index using natural element order.
1091 (first_match_or_eos_index_<mode>):
1092 Calculate index using natural order.
1093 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1094 for BE and LE modes.
1095 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1096 P9V_BUILTIN_VCLZLSBB_V16QI.
1097 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1100 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1102 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1103 indentation and line wrap for many prototypes. Add missing
1104 @smallexample directives around block of prototypes for vec_xl and
1107 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1109 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1110 track if we pass or return IEEE 128-bit floating point.
1111 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1112 C++ mangling that is compatible with GCC 8.1.
1113 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1114 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1115 floating point types.
1116 (rs6000_function_arg_advance_1): Likewise.
1117 (rs6000_mangle_type): Optionally generate mangled names that match
1118 what GCC 8.1 generated for IEEE 128-bit floating point types.
1119 (rs6000_globalize_decl_name): If we have an external function that
1120 passes or returns IEEE 128-bit types, generate a weak reference
1121 from the mangled name used in GCC 8.1 to the current mangled
1123 (rs6000_init_builtins): Make __ibm128 use the long double type if
1124 long double is IBM extended double. Make __float128 use the long
1125 double type if long double is IEEE 128-bit.
1128 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1129 macro for __ibm128 built-in functions.
1130 (PACK_IF): Add __ibm128 pack/unpack functions.
1131 (UNPACK_IF): Likewise.
1132 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1133 enable long double built-in functions if long double is IEEE
1134 128-bit floating point.
1135 (rs6000_invalid_builtin): Update long double built-in function
1137 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1138 functions, adjust the built-in function to use the long double
1139 built-in function if __ibm128 and long double are the same type.
1140 * doc/extend.texi (PowerPC builtins): Update documention for
1141 __builtin_{,un}pack_longdouble. Add documentation for
1142 __builtin_{,un}pack_ibm128.
1144 2018-06-06 Jim Wilson <jimw@sifive.com>
1146 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1147 (struct machine_function): New field interrupt_mode.
1148 (riscv_handle_type_attribute): New function. Add forward declaration.
1149 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1150 (riscv_expand_epilogue): Check interrupt_mode field.
1151 (riscv_set_current_function): Check interrupt attribute args and
1152 set interrupt_mode field.
1153 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1154 (riscv_sret, riscv_uret): New.
1155 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1156 new arguments to interrupt attribute.
1158 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1161 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1162 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1164 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1166 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1167 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1168 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1169 several redundant entries.
1171 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1173 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1174 type from "rtx" to "rtx_insn *".
1175 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1176 for local "call_insn", removing cast.
1177 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1179 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1181 PR tree-optimization/86066
1182 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1183 for BIT_INSERT_EXPR stores.
1185 2018-06-06 Richard Biener <rguenther@suse.de>
1187 PR tree-optimization/86062
1188 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1189 component refs ontop
1190 of to be offsetted base.
1192 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1194 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1195 to be static and remove check on interrupt attribute name.
1197 gcc/testsuite/gcc.target/msp430/
1198 * function-attributes-4.c: New test.
1199 * static-interrupts.c: New test.
1201 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1203 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1204 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1206 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1209 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1211 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1212 Remove second argument, change how error is called.
1213 (aarch64_layout_arg): Remove second argument from
1214 aarch64_err_no_fpadvsimd call.
1215 (aarch64_init_cumulative_args): Ditto.
1216 (aarch64_gimplify_va_arg_expr): Ditto.
1217 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1219 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1221 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1222 (*simple_return_indirect_internal<mode>): Rename from
1223 simple_return_indirect_internal. Use W mode iterator.
1224 (rstorssp): New expander.
1225 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1226 (clrssbsy): New expander.
1227 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1229 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1231 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1233 (cmse_check_pointed_object): Likewise.
1235 2018-06-05 Martin Liska <mliska@suse.cz>
1237 PR gcov-profile/47618
1238 * doc/invoke.texi: Document how -fprofile-dir format
1241 2018-06-05 Richard Biener <rguenther@suse.de>
1243 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1244 removal pretend DOM info isn't available so we do not update
1245 it and only remove edges, not dominated blocks. Actually free
1246 DOM info in case we removed something. Remove unreachable blocks.
1247 (mfb_keep_latches): Work with either DOM info or marked backedges.
1248 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1249 first. Mark backedges if DOM info isn't available.
1250 (Re-)compute DOM info after cleanup_control_flow_pre.
1252 2018-06-05 Richard Biener <rguenther@suse.de>
1254 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1255 (locus_discrim_hasher::hash): Adjust.
1256 (locus_discrim_hasher::equal): Likewise.
1257 (next_discriminator_for_locus): Work on line directly.
1258 (same_line_p): Pass in expanded locus1 as well.
1259 (assign_discriminators): Avoid redundant location expansions.
1261 2018-06-05 Richard Biener <rguenther@suse.de>
1263 PR tree-optimization/86046
1264 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1265 if required after clearing TREE_ADDRESSABLE.
1267 2018-06-05 Richard Biener <rguenther@suse.de>
1269 PR tree-optimization/86047
1270 * tree-ssa-loop.c (for_each_index): Glob handling of all
1271 decls and constants and really handle all of them.
1273 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1276 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1277 qualifier_void_pointer and qualifier_const_void_pointer.
1278 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1279 (arm_init_builtins): Handle the above.
1280 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1281 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1282 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1285 2018-06-05 Martin Liska <mliska@suse.cz>
1287 * auto-profile.c (read_autofdo_file): Do not use
1288 gcov_ctr_summary struct.
1289 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1290 * coverage.c (struct counts_entry): Likewise.
1291 (read_counts_file): Read just single summary entry.
1292 (get_coverage_counts): Use gcov_summary struct.
1293 * coverage.h (get_coverage_counts): Likewise.
1294 * gcov-dump.c (dump_working_sets): Likewise.
1295 (tag_summary): Dump just single summary.
1296 * gcov-io.c (gcov_write_summary): Write just histogram
1298 (gcov_read_summary): Read just single summary.
1299 (compute_working_sets): Use gcov_summary struct.
1300 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1301 of GCOV_COUNTERS_SUMMABLE.
1302 (GCOV_COUNTERS_SUMMABLE): Remove.
1303 (GCOV_FIRST_VALUE_COUNTER): Replace with
1304 GCOV_COUNTER_V_INTERVAL.
1305 (struct gcov_ctr_summary): Remove.
1306 (struct gcov_summary): Directly use fields of former
1308 (compute_working_sets): Use gcov_summary struct.
1309 * gcov.c (read_count_file): Do not use ctrs fields.
1310 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1312 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1314 * profile.c: Likewise.
1315 * profile.h: Likewise.
1317 2018-06-05 Martin Liska <mliska@suse.cz>
1319 PR gcov-profile/84846
1320 * gcov.c (output_lines): Print working directory only
1321 in intermediate format.
1323 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1325 * config/s390/s390-builtin-types.def: Add void function type.
1326 * config/s390/s390-builtins.def: Use the function type for the
1329 2018-06-04 Jim Wilson <jimw@sifive.com>
1331 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1333 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1334 and maybe_eh_return. Change regno to unsigned int. Use new args to
1335 handle EH_RETURN_DATA_REGNO registers properly.
1336 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1337 (riscv_expand_epilogue): Update comment. Change argument name and
1338 type. Update code to use new name and type. Pass new args to
1339 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1341 * config/riscv/riscv.md (NORMAL_RETURN): New.
1342 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1343 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1344 (eh_return): Call gen_eh_return_internal and emit barrier.
1345 (eh_return_internal): Call riscv_expand_epilogue.
1347 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1349 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1350 bit_insertion field and declare can_be_merged_into method.
1351 (merged_store_group::can_be_merged_into): New method.
1352 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1353 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1354 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1356 2018-06-04 Richard Biener <rguenther@suse.de>
1358 PR tree-optimization/85955
1359 * builtins.c (fold_builtin_sincos): Convert pointers to
1360 destination to appropriate type before dereferencing.
1362 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1364 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1366 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1368 * expr.c (expand_expr_real_1): Force the operand into memory if
1369 its TYPE_MODE is BLKmode and if there is no integer mode for
1370 the number of bits being extracted.
1372 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1376 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1377 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1379 2018-06-04 Richard Biener <rguenther@suse.de>
1381 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1382 (cleanup_tree_cfg_noloop): ... single caller. Do
1383 start_recording_case_labels later.
1385 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1387 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1388 to _IMMINTRIN_H_INCLUDED.
1389 * config/i386/pconfigintrin.h: Ditto.
1390 * config/i386/waitpkgintrin.h: Ditto.
1391 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1392 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1393 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1394 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1395 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1396 waitpkgintrin.h and cldemoteintrin.h.
1398 2018-06-04 Richard Biener <rguenther@suse.de>
1400 PR tree-optimization/86038
1401 * tracer.c (find_best_successor): Check probability for
1402 being initialized, bail out if not.
1404 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1407 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1408 of bits to ignore when comparing architectures.
1410 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1412 PR tree-optimization/69615
1413 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1414 maximum or minimum of the type, try to merge it also as if
1415 range1 is + [-, x - 1] or + [x + 1, -].
1418 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1420 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1422 PR tree-optimization/86034
1423 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1424 the unsigned bitfield type in a bit insertion sequence if it does not
1425 have a larger precision than the bitfield size.
1426 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1428 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1430 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1432 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1434 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1435 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1436 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1437 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1439 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1441 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1442 Disable -fdelete-null-pointer-checks for ELF toolchain.
1444 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1445 Kito Cheng <kito.cheng@gmail.com>
1447 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1448 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1449 (nds32*-*-*): Add glibc and uclibc conditions.
1450 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1451 (TARGET_EXCEPT_UNWIND_INFO): Define.
1452 * config/nds32/elf.h: New file.
1453 * config/nds32/linux.h: New file.
1454 * config/nds32/nds32-elf.opt: New file.
1455 * config/nds32/nds32-linux.opt: New file.
1456 * config/nds32/nds32-fp-as-gp.c
1457 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1458 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1460 (nds32_asm_file_end): Ditto.
1461 (nds32_print_operand): Ditto.
1462 (nds32_insert_attributes): Ditto.
1463 (nds32_init_libfuncs): New function.
1464 (TARGET_HAVE_TLS): Define.
1465 (TARGET_INIT_LIBFUNCS): Define.
1466 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1468 (TARGET_ELF): Apply different mcmodel setting.
1469 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1470 been migrated into elf.h and linux.h files.
1471 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1472 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1473 (mcmodel): The content has been migrated into nds32-elf.opt and
1474 nds32-linux.opt files.
1475 * config/nds32/t-elf: New file.
1476 * config/nds32/t-linux: New file.
1478 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1479 Shiva Chen <shiva0217@gmail.com>
1481 * config/nds32/constants.md (unspec_volatile_element): Add
1482 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1483 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1485 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1486 (make_pass_nds32_fp_as_gp): Declare.
1487 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1489 (nds32_asm_function_end_prologue): Remove unused asm output.
1490 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1491 (nds32_asm_file_start): Output necessary fp_as_gp information.
1492 (nds32_option_override): Adjust register usage.
1493 (nds32_expand_prologue): Consider fp_as_gp situation.
1494 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1495 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1498 (simple_return): Ditto.
1499 (omit_fp_begin): Output special directive for fp_as_gp.
1500 (omit_fp_end): Output special directive for fp_as_gp.
1501 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1502 mforbid-fp-as-gp): New options.
1504 2018-06-01 Mark Wielaard <mark@klomp.org>
1506 * dwarf2out.c (dwarf2out_finish): Remove generation of
1507 DW_AT_loclists_base.
1509 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1511 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1512 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1513 (struct merged_store_group): Add bit_insertion field.
1514 (dump_char_array): Use standard hexadecimal format.
1515 (merged_store_group::merged_store_group): Set bit_insertion to false.
1516 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1517 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1518 also print the mask in the dump file.
1519 (pass_store_merging::gate): Minor tweak.
1520 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1521 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1522 stores with INTEGER_CST stores.
1523 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1524 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1525 and use it throughout. Generate bit insertion sequences if need be.
1526 (pass_store_merging::process_store): Remove redundant condition.
1527 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1529 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1531 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1532 the 128-bit floating point types. Fix function comment.
1534 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1536 * config/aarch64/aarch64-simd.md
1537 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1539 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1542 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1544 PR tree-optimization/85989
1545 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1547 (backprop::intersect_uses): Check it when deciding whether this
1548 is a backedge reference.
1549 (backprop::process_block): Add each phi to m_visited_phis
1550 after visiting it, then clear it at the end.
1552 2018-06-01 Richard Biener <rguenther@suse.de>
1554 * tree-vectorizer.h (vect_dr_stmt): New function.
1555 (vect_get_load_cost): Adjust.
1556 (vect_get_store_cost): Likewise.
1557 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1558 Use vect_dr_stmt instead of DR_SMTT.
1559 (vect_record_base_alignments): Likewise.
1560 (vect_calculate_target_alignment): Likewise.
1561 (vect_compute_data_ref_alignment): Likewise and make static.
1562 (vect_update_misalignment_for_peel): Likewise.
1563 (vect_verify_datarefs_alignment): Likewise.
1564 (vector_alignment_reachable_p): Likewise.
1565 (vect_get_data_access_cost): Likewise. Pass down
1566 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1567 (vect_get_peeling_costs_all_drs): Likewise.
1568 (vect_peeling_hash_get_lowest_cost): Likewise.
1569 (vect_enhance_data_refs_alignment): Likewise.
1570 (vect_find_same_alignment_drs): Likewise.
1571 (vect_analyze_data_refs_alignment): Likewise.
1572 (vect_analyze_group_access_1): Likewise.
1573 (vect_analyze_group_access): Likewise.
1574 (vect_analyze_data_ref_access): Likewise.
1575 (vect_analyze_data_ref_accesses): Likewise.
1576 (vect_vfa_segment_size): Likewise.
1577 (vect_small_gap_p): Likewise.
1578 (vectorizable_with_step_bound_p): Likewise.
1579 (vect_prune_runtime_alias_test_list): Likewise.
1580 (vect_analyze_data_refs): Likewise.
1581 (vect_supportable_dr_alignment): Likewise.
1582 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1583 (vect_gen_prolog_loop_niters): Likewise.
1584 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1585 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1587 (vect_recog_mask_conversion_pattern): Likewise.
1588 (vect_try_gather_scatter_pattern): Likewise.
1589 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1590 to vect_get_store_cost.
1591 (vect_get_store_cost): Get stmt_info instead of DR.
1592 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1593 (vect_get_load_cost): Get stmt_info instead of DR.
1595 2018-06-01 Richard Biener <rguenther@suse.de>
1598 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1599 vars wrapped in MEM_REFs.
1601 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1603 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1604 Fix subreg tests so that we only return a choice between
1605 GENERAL_REGS and FP_REGS if the original classes included both.
1607 2018-06-01 Richard Biener <rguenther@suse.de>
1610 * tree-ssa-structalias.c (get_function_part_constraint):
1611 Handle NULL fi->decl.
1612 (find_func_aliases_for_call): Properly handle indirect
1613 fi from direct call.
1614 (find_func_clobbers): Likewise.
1615 (ipa_pta_execute): Likewise.
1616 (create_variable_info_for): For functions that are ifunc_resolver
1617 resolve to a varinfo that contains the result of the resolver call.
1618 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1621 2018-05-31 Michael Collison <michael.collison@arm.com>
1623 * config/aarch64/aarch64.md:
1624 (*fix_to_zero_extenddfdi2): New pattern.
1625 * gcc.target/aarch64/fix_extend1.c: New testcase.
1627 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1631 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1632 and BUILT_IN_STRNCMP_EQ.
1633 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1634 BUILT_IN_STRNCMP_EQ.
1635 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1636 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1637 (gimple_fold_builtin): Likewise.
1638 * tree-ssa-strlen.c (compute_string_length): New function.
1639 (determine_min_obsize): New function.
1640 (handle_builtin_string_cmp): New function to handle calls to
1641 string compare functions.
1642 (strlen_optimize_stmt): Add handling to builtin string compare
1644 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1645 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1646 * tree.c (build_common_builtin_nodes): Add new defines of
1647 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1649 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1652 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1653 functions with naked attribute.
1655 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1657 * config/i386/sse.md (avx_vec_concat<mode>):
1658 Substitute concat_tg_mode mode attribute with xtg_mode.
1659 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1660 (concat_tg_mode): Remove mode attribute.
1662 2018-05-31 Martin Sebor <msebor@redhat.com>
1665 * calls.c (alloc_max_size): Correct a logic error/typo.
1666 Treat excessive arguments as infinite. Warn for invalid arguments.
1667 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1669 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1672 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1673 and movx for Haswell.
1675 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1676 Cesar Philippidis <cesar@codesourcery.com>
1679 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1680 when emitting error on private/firstprivate reductions.
1681 * omp-low.c (lower_omp_target): Avoid reference-type processing
1682 on pointers for firstprivate clause.
1684 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1686 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1689 * config/aarch64/aarch64-simd.md
1690 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1691 (aarch64_ld1_x3_<mode>): Likewise
1692 (aarch64_st1x2<VALLDIF:mode>): Likewise
1693 (aarch64_st1_x2_<mode>): Likewise
1694 (aarch64_st1x3<VALLDIF:mode>): Likewise
1695 (aarch64_st1_x3_<mode>): Likewise
1696 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1697 (vld1_s8_x3): Likewise.
1698 (vld1_u16_x3): Likewise.
1699 (vld1_s16_x3): Likewise.
1700 (vld1_u32_x3): Likewise.
1701 (vld1_s32_x3): Likewise.
1702 (vld1_u64_x3): Likewise.
1703 (vld1_s64_x3): Likewise.
1704 (vld1_f16_x3): Likewise.
1705 (vld1_f32_x3): Likewise.
1706 (vld1_f64_x3): Likewise.
1707 (vld1_p8_x3): Likewise.
1708 (vld1_p16_x3): Likewise.
1709 (vld1_p64_x3): Likewise.
1710 (vld1q_u8_x3): Likewise.
1711 (vld1q_s8_x3): Likewise.
1712 (vld1q_u16_x3): Likewise.
1713 (vld1q_s16_x3): Likewise.
1714 (vld1q_u32_x3): Likewise.
1715 (vld1q_s32_x3): Likewise.
1716 (vld1q_u64_x3): Likewise.
1717 (vld1q_s64_x3): Likewise.
1718 (vld1q_f16_x3): Likewise.
1719 (vld1q_f32_x3): Likewise.
1720 (vld1q_f64_x3): Likewise.
1721 (vld1q_p8_x3): Likewise.
1722 (vld1q_p16_x3): Likewise.
1723 (vld1q_p64_x3): Likewise.
1724 (vst1_s64_x2): Likewise.
1725 (vst1_u64_x2): Likewise.
1726 (vst1_f64_x2): Likewise.
1727 (vst1_s8_x2): Likewise.
1728 (vst1_p8_x2): Likewise.
1729 (vst1_s16_x2): Likewise.
1730 (vst1_p16_x2): Likewise.
1731 (vst1_s32_x2): Likewise.
1732 (vst1_u8_x2): Likewise.
1733 (vst1_u16_x2): Likewise.
1734 (vst1_u32_x2): Likewise.
1735 (vst1_f16_x2): Likewise.
1736 (vst1_f32_x2): Likewise.
1737 (vst1_p64_x2): Likewise.
1738 (vst1q_s8_x2): Likewise.
1739 (vst1q_p8_x2): Likewise.
1740 (vst1q_s16_x2): Likewise.
1741 (vst1q_p16_x2): Likewise.
1742 (vst1q_s32_x2): Likewise.
1743 (vst1q_s64_x2): Likewise.
1744 (vst1q_u8_x2): Likewise.
1745 (vst1q_u16_x2): Likewise.
1746 (vst1q_u32_x2): Likewise.
1747 (vst1q_u64_x2): Likewise.
1748 (vst1q_f16_x2): Likewise.
1749 (vst1q_f32_x2): Likewise.
1750 (vst1q_f64_x2): Likewise.
1751 (vst1q_p64_x2): Likewise.
1752 (vst1_s64_x3): Likewise.
1753 (vst1_u64_x3): Likewise.
1754 (vst1_f64_x3): Likewise.
1755 (vst1_s8_x3): Likewise.
1756 (vst1_p8_x3): Likewise.
1757 (vst1_s16_x3): Likewise.
1758 (vst1_p16_x3): Likewise.
1759 (vst1_s32_x3): Likewise.
1760 (vst1_u8_x3): Likewise.
1761 (vst1_u16_x3): Likewise.
1762 (vst1_u32_x3): Likewise.
1763 (vst1_f16_x3): Likewise.
1764 (vst1_f32_x3): Likewise.
1765 (vst1_p64_x3): Likewise.
1766 (vst1q_s8_x3): Likewise.
1767 (vst1q_p8_x3): Likewise.
1768 (vst1q_s16_x3): Likewise.
1769 (vst1q_p16_x3): Likewise.
1770 (vst1q_s32_x3): Likewise.
1771 (vst1q_s64_x3): Likewise.
1772 (vst1q_u8_x3): Likewise.
1773 (vst1q_u16_x3): Likewise.
1774 (vst1q_u32_x3): Likewise.
1775 (vst1q_u64_x3): Likewise.
1776 (vst1q_f16_x3): Likewise.
1777 (vst1q_f32_x3): Likewise.
1778 (vst1q_f64_x3): Likewise.
1779 (vst1q_p64_x3): Likewise.
1781 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1783 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1784 user_label_prefix to name.
1786 * tree-core.h: Update comment about the format of NAME string
1787 passed to handler in attribute_spec.
1789 * config/msp430/msp430.md: Remove erroneous subreg expression from
1790 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1791 zero_extend{q,h}isi2.
1793 2018-05-30 Borislav Petkov <bp@suse.de>
1795 * doc/extend.texi: Document some architecture specific
1796 constraints and sort entries.
1798 2018-05-30 Martin Sebor <msebor@redhat.com>
1801 * builtins.c (expand_builtin_stpcpy_1): New function.
1802 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1803 only if the former succeeds.
1805 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1807 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1810 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1812 * doc/invoke.texi (-flinker-output): Document
1814 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1816 * passes.c (ipa_write_summaries): Only modify statements if body
1818 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1819 incrementally linking.
1820 (ipa_passes): Likewise.
1821 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1822 pass down resolution info.
1823 * common.opt (flag_incremental_link): Update info.
1824 * gcc.c (plugin specs): Turn flinker-output=* to
1825 -plugin-opt=-linker-output-known
1826 * toplev.c (compile_file): Also cut compilation when doing incremental
1828 * flag-types. (enum lto_partition_model): Add
1829 LTO_LINKER_OUTPUT_NOLTOREL.
1830 (invoke.texi): Add -flinker-output docs.
1831 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1832 link same way as WPA; do not stream in dead initializers.
1834 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1835 darf2out_register_external_decl): Support incremental link.
1837 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1839 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1841 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1843 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1844 it down to simple_object_copy_lto_debug_sections.
1845 (run_gcc): Determine incremental LTO link time and configure
1846 lto1 into non-wpa mode, disable renaming of debug sections.
1848 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1850 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1851 descriptions of various incorrectly documented functions.
1853 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1856 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1857 address check not strict.
1859 2018-05-30 Richard Biener <rguenther@suse.de>
1861 PR tree-optimization/85964
1862 * tracer.c (better_p): Drop initialized count check, we only
1863 call the function with initialized counts now.
1864 (find_best_successor): Do find a best edge if one
1865 has uninitialized count.
1866 (find_best_predecessor): Likewise. Do BB frequency check only
1867 if count is initialized.
1869 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1871 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1872 (aarch64_ldrstr_offset_compare): New.
1873 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1874 load/store orderings.
1875 (aarch64_gen_adjusted_ldpstp): Likewise.
1877 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1879 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1880 Check for subset of GENERAL_REGS and FP_REGS.
1881 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1884 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1886 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1887 and wi::to_poly_offset. Add the current offset and then check
1888 whether the sum fits, rather than using an unchecked addition of
1889 a checked term. Check for a shwi rather than a uhwi.
1890 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1891 (store_constructor): Use poly_int_tree_p.
1892 (expand_expr_real_1): Likewise.
1893 * function.c (assign_temp): Likewise.
1894 * fold-const.c (const_binop): Use poly_int_tree_p and
1896 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1898 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1899 to_poly_offset for MEM offsets.
1900 * ipa-icf.c (sem_variable::equals): Likewise.
1901 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1902 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1903 wi::to_poly_offset for BIT_FIELD_REF offsets.
1904 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1906 * var-tracking.c (emit_note_insn_var_location): Use
1907 tree_to_poly_uint64.
1909 2018-05-29 Jim Wilson <jimw@sifive.com>
1911 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
1913 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
1916 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
1917 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
1919 (sse4_1_round<mode>2): Use nonimmediate_operand
1920 for operand 1 predicate.
1922 2018-05-29 Martin Sebor <msebor@redhat.com>
1923 Richard Biener <rguenther@suse.de>
1926 * calls.c (get_size_range): Call determine_value_range instead
1927 of get_value_range..
1928 * tree-vrp.h (determine_value_range): Declared new function.
1929 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
1931 2018-05-29 Richard Biener <rguenther@suse.de>
1933 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
1934 sure to use non-pattern stmts for get_earlier_stmt arguments.
1935 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
1936 called on pattern stmts.
1937 (get_later_stmt): Likewise.
1939 2018-05-29 Martin Liska <mliska@suse.cz>
1941 PR gcov-profile/85759
1942 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
1945 2018-05-29 Jakub Jelinek <jakub@redhat.com>
1947 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
1949 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
1953 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
1954 VEC_PACK_FLOAT_EXPR): New tree codes.
1955 * tree-pretty-print.c (op_code_prio): Handle
1956 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1957 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1958 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1959 * tree-inline.c (estimate_operator_cost): Likewise.
1960 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
1961 * fold-const.c (const_binop): Likewise.
1962 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
1963 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
1964 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
1965 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
1966 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
1967 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
1968 * expr.c (expand_expr_real_2): Likewise.
1969 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
1970 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
1971 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
1973 * optabs.c (expand_widen_pattern_expr): For
1974 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
1975 sign from result type rather than operand's type.
1976 (expand_binop_directly): For vec_packu_float_optab and
1977 vec_packs_float_optab allow result type to be different from operand's
1979 * optabs-tree.c (optab_for_tree_code): Handle
1980 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1981 VEC_PACK_FLOAT_EXPR. Formatting fixes.
1982 * tree-vect-generic.c (expand_vector_operations_1): Handle
1983 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
1984 VEC_PACK_FLOAT_EXPR.
1985 * tree-vect-stmts.c (supportable_widening_operation): Handle
1987 (supportable_narrowing_operation): Handle FLOAT_EXPR.
1988 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
1989 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
1990 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
1991 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
1993 (vec_pack<floatprefix>_float_<mode>): New expander.
1994 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
1996 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
1997 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
1998 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
1999 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2000 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2002 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2003 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2004 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2005 VEC_PACK_FLOAT_EXPR): Document.
2007 2018-05-29 Richard Biener <rguenther@suse.de>
2009 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2011 (stmt_vec_info_vec): Make pointer.
2012 (init_stmt_vec_info_vec): Remove.
2013 (free_stmt_vec_info_vec): Likewise.
2014 (set_stmt_vec_info_vec): New function.
2015 (free_stmt_vec_infos): Likewise.
2016 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2017 (set_vinfo_for_stmt): Likewise.
2018 (get_earlier_stmt): Likewise.
2019 (get_later_stmt): Likewise.
2020 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2021 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2022 (vec_info::~vec_info): Free stmt_vec_infos.
2023 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2024 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2025 (pass_slp_vectorize::execute): Likewise.
2026 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2027 (free_stmt_vec_info_vec): Likewise.
2028 (set_stmt_vec_info_vec): New function.
2029 (free_stmt_vec_infos): Likewise.
2030 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2031 the global stmt_vec_info_vec.
2032 * tree-parloops.c (gather_scalar_reductions): Use
2033 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2036 2018-05-29 Richard Biener <rguenther@suse.de>
2038 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2040 2018-05-29 Martin Liska <mliska@suse.cz>
2041 David Malcolm <dmalcolm@redhat.com>
2043 * vec.c (test_reverse): New.
2044 (vec_c_tests): Add new test.
2045 * vec.h (vl_ptr>::reverse): New function.
2047 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2049 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2051 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2054 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2056 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2058 2018-05-28 Richard Biener <rguenther@suse.de>
2060 PR tree-optimization/85933
2061 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2062 look at stmts marked as vectorizable.
2064 2018-05-28 Richard Biener <rguenther@suse.de>
2066 PR tree-optimization/85934
2067 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2068 vector boolean check before scalar optimization.
2070 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2072 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2075 2018-05-28 Mark Wielaard <mark@klomp.org>
2077 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2078 if it is an expression containing a minus sign.
2080 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2082 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2084 2018-05-27 Paul Koning <ni1d@arrl.net>
2086 * config/pdp11/pdp11.md (truncsihi2): Remove.
2088 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2089 Chung-Ju Wu <jasonwucj@gmail.com>
2091 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2093 (unaligned_store_dw): Ditto.
2094 * config/nds32/nds32-memory-manipulation.c
2095 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2096 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2097 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2098 (emit_setmem_word_loop): Rename to ...
2099 (emit_setmem_doubleword_loop): ... this.
2100 (nds32_gen_dup_4_byte_to_word_value): New function.
2101 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2102 (nds32_expand_setmem_loop): Refine implementation.
2103 (nds32_expand_setmem_loop_v3m): Ditto.
2104 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2107 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2109 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2111 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2113 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2114 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2115 machine->attr_no_prologue_p.
2116 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2117 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2118 (nds32_expand_epilogue): Consider attr_naked_p.
2119 (nds32_expand_epilogue_v3pop): Likewise.
2120 (nds32_can_use_return_insn): Likewise.
2121 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2122 attr_no_prologue_p fields.
2123 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2125 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2128 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2130 * config/i386/sse.md
2131 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2133 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2135 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2137 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2139 (*<floatsuffix>floatv2div2sf2): Rename to ...
2140 (*float<floatunssuffix>v2div2sf2): ... this.
2141 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2142 (float<floatunssuffix>v2div2sf2_mask): ... this.
2143 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2144 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2145 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2147 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2149 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2151 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2153 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2155 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2157 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2158 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2159 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2160 gen_ufix_truncv8dfv8si2.
2161 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2162 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2163 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2164 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2165 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2166 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2167 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2168 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2170 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2174 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2176 2018-05-25 Jim Wilson <jimw@sifive.com>
2178 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2179 * config/riscv/riscv.c (struct machine_function): Add
2180 interrupt_handler_p and attribute_checked_p fields.
2181 (riscv_attribute_table): Add interrupt.
2182 (riscv_interrupt_type_p): New.
2183 (riscv_save_reg_p): Save extra regs for interrupt handler.
2184 (riscv_use_save_libcall): Return false for interrupt handler.
2185 (riscv_first_stack_step): Add forward declaration.
2186 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2187 for interrupt handler with large frame. Use it for saved reg list.
2188 (riscv_expand_prologue): Move flag_stack_usage_info support to
2189 eliminate duplication.
2190 (riscv_expand_epilogue): Generate mret for interrupt handler.
2191 (riscv_epilogue_uses): New.
2192 (riscv_can_use_return_insn): Return false for interrupt handler.
2193 (riscv_function_ok_for_sibcall): Likewise.
2194 (riscv_set_current_function): Add interrupt handler support.
2195 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2196 * config/riscv/riscv.md (UNSPECV_MRET): New.
2198 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2200 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2202 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2204 PR tree-optimization/85712
2205 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2206 this candidate has already been replaced in-situ by a copy.
2208 2018-05-25 Jason Merrill <jason@redhat.com>
2210 PR c++/80485 - inline function non-zero address.
2211 * symtab.c (nonzero_address): Check DECL_COMDAT.
2213 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2216 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2217 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2218 extension of SImode operation. Use const123_operand predicate.
2219 (*saddsi_1): Remove.
2220 (*saddl_se_1): Ditto.
2222 (*ssubl_se_1): Ditto.
2223 * config/alpha/predicates.md (const123_operand): New predicate.
2224 * config/alpha/constraints.md (P): Use IN_RANGE.
2226 2018-05-25 Richard Biener <rguenther@suse.de>
2228 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2230 (ref_maybe_used_by_stmt_p): Likewise.
2231 (stmt_may_clobber_ref_p): Likewise.
2232 (stmt_may_clobber_ref_p_1): Likewise.
2233 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2235 (ref_maybe_used_by_stmt_p): Likewise.
2236 (stmt_may_clobber_ref_p): Likewise.
2237 (stmt_may_clobber_ref_p_1): Likewise.
2238 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2239 the alias oracle to disambiguate DRs with stmts DR analysis
2241 (vect_analyze_data_refs): Do not give up on not analyzable
2242 DRs for BB vectorization. Remove code truncating the dataref
2245 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2248 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2249 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2250 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2252 2018-05-25 Richard Biener <rguenther@suse.de>
2254 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2255 function, combining stmt data ref gathering and fatal analysis
2257 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2258 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2259 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2260 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2261 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2263 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2265 PR tree-optimization/85720
2266 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2267 SCC if all partitions are builtins.
2268 (version_loop_by_alias_check): New parameter. Generate cancelable
2269 runtime alias check if all partitions are builtins.
2270 (distribute_loop): Update call to above function.
2272 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2274 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2275 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2276 (parm_default_def_partition_arg): Ditto.
2277 (set_parm_default_def_partition): Ditto.
2278 (get_parm_default_def_partitions): Ditto and make it static.
2279 (get_undefined_value_partitions): Ditto and make it static.
2280 (remove_ssa_form): Refactor call to init_var_map here.
2281 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2282 computation for loop region.
2283 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2284 (register_default_def): Delete.
2285 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2286 (parm_default_def_partition_arg): Ditto.
2287 (set_parm_default_def_partition): Ditto.
2288 (get_parm_default_def_partitions): Ditto and make it static.
2289 (get_undefined_value_partitions): Ditto and make it static.
2290 (coalesce_with_default, coalesce_with_default): Update comment.
2291 (create_coalesce_list_for_region): New func factored out from
2292 create_outofssa_var_map.
2293 (populate_coalesce_list_for_outofssa): New func factored out from
2294 create_outofssa_var_map and coalesce_ssa_name.
2295 (create_outofssa_var_map): Delete.
2296 (coalesce_ssa_name): Refactor to support live range computation.
2297 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2298 (get_parm_default_def_partitions): Delete.
2299 (get_undefined_value_partitions): Ditto.
2300 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2301 computation for loop region.
2302 (new_tree_live_info, loe_visit_block): Ditto.
2303 (live_worklist, set_var_live_on_entry): Ditto.
2304 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2305 * tree-ssa-live.h (struct _var_map): New fields.
2306 (init_var_map): Change decl.
2307 (region_contains_p): New.
2309 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2311 * tree-ssa-live.h (live_merge_and_clear): Delete.
2313 2018-05-25 Richard Biener <rguenther@suse.de>
2316 * tree-dump.c (dequeue_and_dump): Remove access to removed
2317 operand 2 of a SWITCH_EXPR.
2319 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2321 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2322 multiplication and division.
2323 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2324 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2325 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2326 (cond_udiv_optab, cond_umod_optab): New optabs.
2327 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2328 (IFN_COND_RDIV): New internal functions.
2329 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2330 TRUNC_MOD_EXPR and RDIV_EXPR.
2331 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2332 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2334 (SVE_INT_BINARY): Include mult.
2335 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2336 (optab, sve_int_op): Handle mult.
2337 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2339 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2340 for SVE_INT_BINARY_SD.
2342 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2344 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2345 (optab, sve_int_op): Handle div and udiv.
2346 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2347 for SVE_INT_BINARY_SD.
2348 (*<optab><mode>3): New insn for the same.
2350 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2352 * tree-vect-patterns.c: Include predict.h.
2353 (vect_recog_divmod_pattern): Restrict check for division support
2354 to when optimizing for size.
2356 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2358 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2359 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2360 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2361 (gimple_match_op::set_op): Likewise.
2362 (gimple_resimplify4): Declare.
2363 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2364 (expr::gen_transform): Likewise.
2365 (decision_tree::gen): Generate a simplification routine for 4 operands.
2366 * gimple-match-head.c (gimple_simplify): Add an overload for
2367 4 operands. In the top-level function, handle up to 4 call
2368 arguments and call gimple_resimplify4.
2369 (gimple_resimplify4): New function.
2370 (build_call_internal): Pass a fourth operand.
2371 (maybe_push_to_seq): Likewise.
2372 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2373 Fold VEC_COND_EXPRs of an operation and a default value into
2374 an IFN_COND_* function if possible.
2375 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2377 (SVE_COND_FP_BINARY): Include them.
2378 (optab, sve_fp_op): Handle them.
2379 (SVE_INT_BINARY_REV): New code iterator.
2380 (SVE_COND_FP_BINARY_REV): New int iterator.
2381 (commutative): New int attribute.
2382 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2384 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2386 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2387 (*cond_<optab><mode>): New patterns for reversed operands.
2389 2018-05-25 Richard Biener <rguenther@suse.de>
2391 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2392 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2393 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2394 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2395 * tree-vect-data-refs.c (everywhere): Adjust users.
2396 * tree-vect-loop.c (everywhere): Likewise.
2397 * tree-vect-slp.c (everywhere): Likewise.
2398 * tree-vect-stmts.c (everywhere): Likewise.
2399 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2401 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2403 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2405 (gcc_cv_as_section_exclude): ... this.
2406 Try Solaris as #exclude syntax.
2407 * configure: Regenerate.
2408 * config.in: Regenerate.
2409 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2411 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2412 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2414 * varasm.c (default_elf_asm_named_section): Don't check if
2415 HAVE_GAS_SECTION_EXCLUDE is defined.
2417 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2419 * doc/md.texi: Update the documentation of the cond_* optabs
2420 to mention the new final operand. Fix GET_MODE_NUNITS call.
2421 Describe the scalar case too.
2422 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2423 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2425 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2426 (get_conditional_internal_fn): Update comment.
2427 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2428 accumulator value as a final argument to conditional functions.
2429 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2430 a define_expand and add an "else" operand. Assert for now that
2431 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2432 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2433 (*cond_<optab><mode>): New patterns.
2434 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2435 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2436 (UNSPEC_COND_EOR): Delete.
2437 (optab): Remove associated mappings.
2438 (SVE_INT_BINARY): New code iterator.
2439 (sve_int_op): Remove int attribute and add "minus" to the code
2441 (SVE_COND_INT_OP): Delete.
2442 (SVE_COND_FP_OP): Rename to...
2443 (SVE_COND_FP_BINARY): ...this.
2445 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2447 * optabs.c (can_reuse_operands_p): New function.
2448 (maybe_legitimize_operands): Try to reuse the results for
2451 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2453 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2454 Add {q} suffix to insn mnemonic.
2456 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2458 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2459 (msp430_warn_func_return): New.
2461 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2463 * fold-const.c (tree_nonzero_bits): New function.
2464 * fold-const.h (tree_nonzero_bits): Likewise.
2465 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2466 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2468 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2472 * varasm.c (assemble_alias): Check ifunc_resolver only on
2475 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2478 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2479 when memory input operand is handled.
2481 2018-05-24 Luis Machado <luis.machado@linaro.org>
2483 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2485 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2487 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2489 * match.pd: Delay FMA folds until after vectorization.
2491 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2494 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2495 address check not strict.
2497 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2499 * gimple-match.h (gimple_match_op): New class.
2500 (mprts_hook): Replace parameters with a gimple_match_op *.
2501 (maybe_build_generic_op): Likewise.
2502 (gimple_simplified_result_is_gimple_val): Replace parameters with
2503 a const gimple_match_op *.
2504 (gimple_simplify): Replace code_helper * and tree * parameters with
2505 a gimple_match_op * parameter.
2506 (gimple_resimplify1): Replace code_helper *, tree and tree *
2507 parameters with a gimple_match_op * parameter.
2508 (gimple_resimplify2): Likewise.
2509 (gimple_resimplify3): Likewise.
2510 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2511 parameters with a gimple_match_op * parameter.
2512 * gimple-match-head.c (gimple_simplify): Change prototypes of
2513 auto-generated functions to take a gimple_match_op * instead of
2514 separate code_helper * and tree * parameters. Make the same
2515 change in the top-level overload and update calls to the
2516 gimple_resimplify routines. Update calls to the auto-generated
2517 functions and to maybe_push_res_to_seq in the publicly-facing
2518 operation-specific gimple_simplify overloads.
2519 (gimple_match_op::MAX_NUM_OPS): Define.
2520 (gimple_resimplify1): Replace rcode and ops with a single res_op
2521 parameter. Update call to gimple_simplify.
2522 (gimple_resimplify2): Likewise.
2523 (gimple_resimplify3): Likewise.
2524 (mprts_hook): Replace parameters with a gimple_match_op *.
2525 (maybe_build_generic_op): Likewise.
2526 (build_call_internal): Replace type, nargs and ops with
2527 a gimple_match_op *.
2528 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2529 with a single gimple_match_op *. Update calls to mprts_hook,
2530 build_call_internal and gimple_simplified_result_is_gimple_val.
2531 Factor out code that is common to the tree_code and combined_fn cases.
2532 * genmatch.c (expr::gen_transform): Replace tem_code and
2533 tem_ops with a gimple_match_op called tem_op. Update calls
2534 to the gimple_resimplify functions and maybe_push_res_to_seq.
2535 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2536 res_ops. Update call to the gimple_resimplify functions.
2537 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2538 (decision_tree::gen): Make the functions take a gimple_match_op *
2539 called res_op instead of separate res_code and res_ops parameters.
2540 Update call accordingly.
2541 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2542 and ops with a single res_op parameter. Update calls to
2543 maybe_build_generic_op and maybe_push_res_to_seq.
2544 (fold_stmt_1): Update calls to gimple_simplify and
2545 replace_stmt_with_simplification.
2546 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2547 and gimple_simplified_result_is_gimple_val.
2548 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2550 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2551 with a gimple_match_op *.
2552 (vn_nary_build_or_lookup): Likewise. Update call to
2553 vn_nary_build_or_lookup_1.
2554 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2555 gimple_match_op *. Update calls to the gimple_resimplify routines
2556 and to gimple_simplified_result_is_gimple_val.
2557 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2558 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2559 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2560 (visit_nary_op): Likewise.
2561 (visit_reference_op_load): Likewise.
2563 2018-05-23 Luis Machado <luis.machado@linaro.org>
2565 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2566 modifier for printing the step amount.
2568 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2571 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2574 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2576 * doc/sourcebuild.texi (Endianness): New subsubsection.
2578 2018-05-23 Luis Machado <luis.machado@linaro.org>
2580 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2581 <prefetch_dynamic_strides>: New const bool field.
2582 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2583 prefetch_dynamic_strides.
2584 (exynosm1_prefetch_tune): Likewise.
2585 (thunderxt88_prefetch_tune): Likewise.
2586 (thunderx_prefetch_tune): Likewise.
2587 (thunderx2t99_prefetch_tune): Likewise.
2588 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2590 (aarch64_override_options_internal): Update to set
2591 PARAM_PREFETCH_DYNAMIC_STRIDES.
2592 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2593 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2594 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2595 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2596 prefetch-dynamic-strides setting.
2598 2018-05-23 Luis Machado <luis.machado@linaro.org>
2600 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2601 <minimum_stride>: New const int field.
2602 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2603 minimum_stride field defaulting to -1.
2604 (exynosm1_prefetch_tune): Likewise.
2605 (thunderxt88_prefetch_tune): Likewise.
2606 (thunderx_prefetch_tune): Likewise.
2607 (thunderx2t99_prefetch_tune): Likewise.
2608 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2609 <default_opt_level>: Set to 3.
2610 (aarch64_override_options_internal): Update to set
2611 PARAM_PREFETCH_MINIMUM_STRIDE.
2612 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2613 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2614 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2615 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2616 stride is constant and is below the minimum stride threshold.
2618 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2620 * config/arm/arm-cpus.in (mode26): Delete.
2621 (armv4): Delete mode26 reference.
2622 * config/arm/arm.c (arm_configure_build_target): Delete use of
2625 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2627 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2629 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2630 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2632 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2633 (floatunsdidf2): Ditto.
2635 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2637 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2638 (fixuns_trunc<mode>si2_avx512f): Ditto.
2639 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2640 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2641 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2643 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2645 PR rtl-optimization/79985
2646 * df-scan.c (df_insn_refs_collect): Remove special case for
2647 global registers and asm statements.
2649 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2651 * extend.texi (Global Register Variables): Rewrite the bullet list.
2652 Note that the register is available for allocation. Note that access
2653 via inline asm must use constraints. Add note about async-signal
2654 handlers. Remove paragraph about automagic register selection.
2656 2018-05-23 Richard Biener <rguenther@suse.de>
2658 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2659 of fixed offset from memset VN.
2661 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2663 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2665 (alloc_cand_and_find_basis): Initialize first_interp field.
2666 (slsr_process_mul): Modify first_interp field.
2667 (slsr_process_add): Likewise.
2668 (slsr_process_cast): Modify first_interp field for each new
2670 (slsr_process_copy): Likewise.
2671 (dump_candidate): Dump first_interp field.
2672 (replace_mult_candidate): Process all interpretations, not just
2674 (replace_rhs_if_not_dup): Likewise.
2675 (replace_one_candidate): Likewise.
2677 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2679 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2681 (aarch64_needs_frame_chain): New function.
2682 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2684 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2687 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2688 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2689 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2690 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2691 as true for strict-align.
2692 (aarch64_can_inline_p): Perform checks even when callee has no
2693 attributes to check for strict alignment.
2694 * doc/extend.texi (AArch64 Function Attributes): Document
2696 * doc/invoke.texi: (AArch64 Options): Likewise.
2698 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2700 PR tree-optimization/85853
2701 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2702 the handling of the root of the node to...
2703 (vect_slp_analyze_node_operations_1): ...this new function,
2704 and run the whole thing with the child nodes' def types
2705 set according to their SLP node's def type.
2707 2018-05-23 Richard Biener <rguenther@suse.de>
2710 * tree-data-ref.c (create_runtime_alias_checks): Defer
2711 and ignore overflow warnings.
2713 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2715 PR tree-optimization/85822
2716 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2719 2018-05-23 Richard Biener <rguenther@suse.de>
2721 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2722 memset constants via native_interpret_expr.
2724 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2727 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2729 (cgraph_node::create_alias): Likewise.
2730 (cgraph_node::get_availability): Check ifunc_resolver instead
2731 of looking up ifunc attribute.
2732 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2733 * varasm.c (do_assemble_alias): Likewise.
2734 (assemble_alias): Likewise.
2735 (default_binds_local_p_3): Likewise.
2736 * cgraph.h (cgraph_node): Add ifunc_resolver.
2737 (cgraph_node::only_called_directly_or_aliased_p): Return false
2739 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2741 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2742 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2743 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2744 instead of looking up ifunc attribute.
2746 2018-05-22 Luis Machado <luis.machado@linaro.org>
2748 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2750 2018-05-22 Martin Sebor <msebor@redhat.com>
2753 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2754 only when expasion succeeds.
2755 (expand_builtin_strcmp): Same.
2756 (expand_builtin_strncmp): Same.
2758 2018-05-22 Martin Sebor <msebor@redhat.com>
2760 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2762 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2763 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2765 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2766 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2768 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2769 Add check for legitimate_address.
2770 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2771 (aarch64_swap_ldrstr_operands): New.
2772 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2775 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2776 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2778 * config/aarch64/aarch64.md: New patterns to generate stp
2780 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2781 single words and double words.
2782 (load_pair_sw, load_pair_dw): Likewise.
2783 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2785 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2787 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2788 for different mode ldpstp and add peephole for merged zero stores.
2790 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2792 (aarch64_gen_store_pair): Rename calls to match new patterns.
2793 (aarch64_gen_load_pair): Rename calls to match new patterns.
2794 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2795 (load_pair<DREG:mode><DREG2:mode>): ... This.
2796 (store_pair<mode>): Rename to...
2797 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2798 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2800 (V_INT_EQUIV): Handle SImode.
2801 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2804 2018-05-22 Martin Sebor <msebor@redhat.com>
2807 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2808 or ajust the presumed bound on an operation to avoid unnecessary
2811 2018-05-22 Martin Sebor <msebor@redhat.com>
2813 PR tree-optimization/85826
2814 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2815 assuming that a DECL necesarily has a constant size.
2817 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2820 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2822 2018-05-22 Richard Biener <rguenther@suse.de>
2824 PR tree-optimization/85834
2825 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2826 non-constant and non-zero memset arguments.
2828 2018-05-22 Martin Liska <mliska@suse.cz>
2831 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2833 2018-05-22 Richard Biener <rguenther@suse.de>
2835 PR tree-optimization/85863
2836 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2837 comparisons when vectype is specified.
2838 (vectorizable_condition): Do not specify vectype for
2839 vect_is_simple_cond when SLP vectorizing.
2841 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2844 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2845 define __ibm128 as long double.
2846 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2847 as a distinct type when IEEE 128-bit support is enabled.
2848 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2849 128-bit types to use the correct functions.
2850 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2851 convert between 128-bit floating point types that have different
2852 modes but the same representation, instead of using gen_lowpart to
2854 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2856 (IFKF_reg): New attributes to give the register constraints for
2858 (extend<mode>tf2_internal): New insns to mark an explicit
2859 conversion between 128-bit floating point types that have a
2860 different mode but share the same representation.
2862 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2864 PR tree-optimization/85814
2865 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2866 a null return from get_strinfo when unsharing the next
2867 strinfo in the chain.
2869 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2872 * varasm.c (weak_finish): Clean up weak_decls.
2874 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2876 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2877 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2878 UNSPEC_UADALP values.
2879 * config/aarch64/iterators.md (ABAL): New int iterator.
2882 (sur): Add mappings for the above.
2883 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2885 (aarch64_<sur>abal<mode>_4): Likewise.
2886 (aarch64_<sur>adalp<mode>_3): Likewise.
2887 (<sur>sadv16qi): New define_expand.
2889 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2891 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2892 (*movdf_internal): Ditto.
2893 (*rcpsf2_sse): Ditto.
2894 (*rsqrtsf2_sse): Ditto.
2895 (*sqrt<mode>2_sse): Ditto.
2897 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2899 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2901 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2902 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2903 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2904 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2906 * config/aarch64/arm_neon.h: Likewise.
2907 * config/aarch64/iterators.md (VQ_I): New.
2909 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2911 * config.gcc: Add arc/t-multilib-linux to tmake_file for
2913 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
2916 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2918 * config/nds32/constraints.md (S): New constraint.
2919 * config/nds32/nds32.md (call_internal): Use constraint S.
2920 (call_value_internal): Likewise.
2921 (sibcall_internal): Likewise.
2922 (sibcall_value_internal): Likewise.
2924 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2925 Chung-Ju Wu <jasonwucj@gmail.com>
2927 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
2930 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
2931 Chung-Ju Wu <jasonwucj@gmail.com>
2933 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
2934 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
2935 (nds32_rtx_costs_impl): Simplify.
2936 (nds32_address_cost_impl): Simplify.
2937 (nds32_init_rtx_costs): New function.
2938 (nds32_rtx_costs_speed_prefer): Likewise.
2939 (nds32_rtx_costs_size_prefer): Likewise.
2940 (nds32_address_cost_speed_prefer): Likewise.
2941 (nds32_address_cost_speed_fwprop): Likewise.
2942 (nds32_address_cost_size_prefer): Likewise.
2943 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
2944 * config/nds32/nds32.c (nds32_option_override): Use
2945 nds32_init_rtx_costs function.
2947 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
2949 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
2950 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
2951 (TARGET_PIPELINE_N8): Likewise.
2952 (TARGET_PIPELINE_N10): Likewise.
2953 (TARGET_PIPELINE_N13): Likewise.
2954 (TARGET_PIPELINE_GRAYWOLF): Likewise.
2956 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
2958 * config/nds32/nds32-fpu.md: Update copyright year.
2960 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2962 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
2964 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2966 * config/nds32/nds32.c
2967 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
2968 * config/nds32/nds32.opt (minline-asm-r15): New option.
2970 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
2972 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
2974 * config/nds32/nds32.md (abssi2): New pattern.
2976 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
2978 * config/i386/i386.md (rex64namesuffix): New mode attribute.
2979 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
2980 Merge insn pattern from sse_cvtsi2ss<round_name> and
2981 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
2982 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
2983 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
2984 using SWI48 mode iterator.
2985 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
2986 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
2987 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
2988 pattern from sse_cvttss2si<round_saeonly_name>
2989 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
2990 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
2991 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
2992 using SWI48 mode iterator.
2993 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
2994 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
2995 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
2996 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
2997 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
2998 using SWI48 mode iterator.
2999 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3000 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3001 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3002 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3003 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3004 SWI48 mode iterator.
3005 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3006 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3007 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3008 pattern from sse_cvttsd2si<round_saeonly_name>
3009 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3011 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3013 * config/nds32/nds32-md-auxiliary.c
3014 (nds32_valid_smw_lwm_base_p): Refine.
3015 (nds32_output_smw_single_word): Refine.
3016 (nds32_output_smw_double_word): New.
3017 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3019 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3021 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3022 (nds32_output_stack_pop): Refine.
3023 (nds32_expand_unaligned_load): Refine.
3024 (nds32_expand_unaligned_store): Refine.
3026 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3027 Chung-Ju Wu <jasonwucj@gmail.com>
3029 * config/nds32/constants.md: Add TP_REGNUM constant.
3030 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3031 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3033 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3034 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3035 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3036 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3037 and PIC code generation.
3038 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3040 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3042 * config/nds32/nds32.md: Support TLS and PIC.
3043 * config/nds32/nds32.c: Support TLS and PIC.
3044 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3045 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3048 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3050 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3051 mode with E_ prefix.
3053 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3054 Chung-Ju Wu <jasonwucj@gmail.com>
3056 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3057 * config/nds32/nds32-md-auxiliary.c
3058 (symbolic_reference_mentioned_p): New.
3059 (nds32_legitimize_ict_address): New.
3060 (nds32_expand_ict_move): New.
3061 (nds32_indirect_call_referenced_p): New.
3062 (nds32_symbol_binds_local_p): Delete.
3063 (nds32_long_call_p): Modify.
3064 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3065 * config/nds32/nds32-protos.h
3066 (symbolic_reference_mentioned_p): Declare.
3067 (nds32_legitimize_ict_address): Declare.
3068 (nds32_expand_ict_move): Declare.
3069 (nds32_indirect_call_referenced_p): Declare.
3070 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3071 (nds32_relax_group): Use nds32_ict_const_p as condition.
3072 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3073 (nds32_asm_file_start): Output ict_model directive in asm code.
3074 (nds32_legitimate_address_p): Consider indirect call.
3075 (nds32_print_operand): Consider indirect call.
3076 (nds32_print_operand_address): Consider indirect call.
3077 (nds32_insert_attributes): Handle "indirect_call" attribute.
3078 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3079 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3080 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3081 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3082 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3083 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3084 (TARGET_ICT_MODEL_SMALL): Define.
3085 (TARGET_ICT_MODEL_LARGE): Define.
3086 * config/nds32/nds32.md (movsi): Consider ict model.
3087 (call, call_value): Consider ict model.
3088 (sibcall, sibcall_value): Consider ict model.
3089 * config/nds32/nds32.opt (mict-model): New option.
3090 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3093 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3094 Monk Chiang <sh.chiang04@gmail.com>
3095 Jim Wilson <jimw@sifive.com>
3097 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3098 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3099 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3100 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3101 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3102 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3103 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3104 compute save_libcall_adjustment properly.
3105 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3106 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3107 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3108 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3109 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3110 (ABI_SPEC): Handle mabi=ilp32e.
3111 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3112 (RVE): Add RVE mask.
3113 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3114 <-march>: Add rv32e as an example.
3116 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3119 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3120 (intra_create_variable_infos): Handle C++ constructors.
3122 2018-05-18 Martin Liska <mliska@suse.cz>
3124 * passes.def: Remove a redundant pass.
3126 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3129 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3131 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3133 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3135 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3137 (armv2, armv2a, armv3, armv3m): Delete architectures.
3138 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3139 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3140 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3142 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3143 (*mulsidi3adddi): Likewise.
3144 (mulsidi3): Likewise.
3145 (*mulsidi3_nov6): Likewise.
3146 (umulsidi3): Likewise.
3147 (umulsidi3_nov6): Likewise.
3148 (umaddsidi4): Likewise.
3149 (*umulsidi3adddi): Likewise.
3150 (smulsi3_highpart): Likewise.
3151 (*smulsi3_highpart_nov6): Likewise.
3152 (umulsi3_highpart): Likewise.
3153 (*umulsi3_highpart_nov6): Likewise.
3154 * config/arm/arm.h (arm_arch3m): Delete.
3155 * config/arm/arm.c (arm_arch3m): Delete.
3156 (arm_option_override_internal): Update armv3-related comment.
3157 (arm_configure_build_target): Delete use of isa_bit_mode32.
3158 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3159 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3160 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3162 (mulusa3): Likewise.
3163 * config/arm/arm-protos.h (arm_arch3m): Delete.
3164 * config/arm/arm-tables.opt: Regenerate.
3165 * config/arm/arm-tune.md: Likewise.
3166 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3167 deleted architectures.
3169 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3171 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3172 (armv5t, armv5te): New features.
3173 (ARMv5, ARMv5e): Delete fgroups.
3174 (ARMv5t, ARMv5te): Adjust for above changes.
3176 (armv5, armv5e): Delete arches.
3177 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3179 (*call_reg_arm): Likewise.
3180 (*call_value_reg_armv5): Likewise.
3181 (*call_value_reg_arm): Likewise.
3182 (*call_symbol): Likewise.
3183 (*call_value_symbol): Likewise.
3184 (*sibcall_insn): Likewise.
3185 (*sibcall_value_insn): Likewise.
3187 (prefetch): Likewise.
3188 (define_split and define_peephole2 dependent on arm_arch5):
3190 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3192 (TARGET_ARM_QBIT): Likewise.
3193 (TARGET_DSP_MULTIPLY): Likewise.
3194 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3195 (arm_arch5, arm_arch5e): Delete.
3196 (arm_arch5t, arm_arch5te): Declare.
3197 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3198 (arm_arch5t): Declare.
3199 (arm_option_reconfigure_globals): Update for the above.
3200 (arm_options_perform_arch_sanity_checks): Update comment, replace
3201 use of arm_arch5 with arm_arch5t.
3202 (use_return_insn): Likewise.
3203 (arm_emit_call_insn): Likewise.
3204 (output_return_instruction): Likewise.
3205 (arm_final_prescan_insn): Likewise.
3206 (arm_coproc_builtin_available): Likewise.
3207 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3208 arm_arch5e with arm_arch5t and arm_arch5te.
3209 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3210 (arm_arch5t, arm_arch5te): Declare.
3211 * config/arm/arm-tables.opt: Regenerate.
3212 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3213 * config/arm/t-multilib: Likewise.
3214 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3215 instead of arm_arch5.
3216 (*call_reg_thumb1): Likewise.
3217 (*call_value_reg_thumb1_v5): Likewise.
3218 (*call_value_reg_thumb1): Likewise.
3219 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3221 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3223 2018-05-18 Martin Liska <mliska@suse.cz>
3225 PR gcov-profile/84846
3226 * doc/gcov.texi: Document -t option of gcov tool.
3228 2018-05-18 Martin Liska <mliska@suse.cz>
3230 PR gcov-profile/84846
3231 * gcov.c (print_usage): Add new -t option.
3232 (process_args): Handle the option.
3233 (generate_results): Use stdout as output when requested by
3236 2018-05-18 Martin Liska <mliska@suse.cz>
3238 PR gcov-profile/84846
3239 * coverage.c (coverage_init): Write PWD to .gcno file.
3240 * doc/gcov.texi: Document how working directory is printed.
3241 * gcov-dump.c (dump_gcov_file): Print PWD.
3242 * gcov.c (output_intermediate_file): Likewise.
3243 (read_graph_file): Read PWD string.
3244 (output_lines): Print PWD.
3246 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3249 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3250 for retval and return false if all args to phi are zero.
3252 2018-05-18 Richard Biener <rguenther@suse.de>
3254 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3256 (evrp_dom_walker::before_dom_children): Call it.
3258 2018-05-18 Richard Biener <rguenther@suse.de>
3260 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3261 results when processing array refs with variable index.
3263 2018-05-18 Toon Moene <toon@moene.org>
3265 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3266 directly after that of -floop-interchange. Indicate that both
3267 options are enabled by default when specifying -O3.
3269 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3271 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3272 iterator. Delete separate integer-mode vec_set<mode> expander.
3273 (aarch64_simd_vec_setv2di): Delete.
3274 (vec_setv2di): Delete.
3275 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3276 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3277 the "w, r" alternative.
3279 2018-05-18 Martin Liska <mliska@suse.cz>
3281 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3282 * tree-pass.h (make_pass_lower_switch_O0): New function.
3283 * tree-switch-conversion.c (node_has_low_bound): Remove.
3284 (node_has_high_bound): Likewise.
3285 (node_is_bounded): Likewise.
3286 (class pass_lower_switch): Make it a template type and create
3288 (pass_lower_switch::execute): Add template argument.
3289 (make_pass_lower_switch): New function.
3290 (make_pass_lower_switch_O0): New function.
3291 (do_jump_if_equal): Remove.
3292 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3293 all the hard work to tree optimization passes.
3295 2018-05-18 Martin Liska <mliska@suse.cz>
3297 * dbgcnt.c (limit_low): Renamed from limit.
3298 (limit_high): New variable.
3299 (dbg_cnt_is_enabled): Check for upper limit.
3300 (dbg_cnt): Adjust dumping.
3301 (dbg_cnt_set_limit_by_index): Add new argument for high
3303 (dbg_cnt_set_limit_by_name): Likewise.
3304 (dbg_cnt_process_single_pair): Parse new format.
3305 (dbg_cnt_process_opt): Use strtok.
3306 (dbg_cnt_list_all_counters): Remove 'value' and add
3308 * doc/invoke.texi: Document changes.
3310 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3312 * doc/sourcebuild.texi (scalar_all_fma): Document.
3313 * tree.def (FMA_EXPR): Delete.
3314 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3315 * internal-fn.c (ternary_direct): New macro.
3316 (expand_ternary_optab_fn): Likewise.
3317 (direct_ternary_optab_supported_p): Likewise.
3318 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3319 * builtins.c (fold_builtin_fma): Delete.
3320 (fold_builtin_3): Don't call it.
3321 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3322 * expr.c (expand_expr_real_2): Likewise.
3323 * fold-const.c (operand_equal_p): Likewise.
3324 (fold_ternary_loc): Likewise.
3325 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3326 * gimple.c (DEFTREECODE): Likewise.
3327 * gimplify.c (gimplify_expr): Likewise.
3328 * optabs-tree.c (optab_for_tree_code): Likewise.
3329 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3330 * tree-eh.c (operation_could_trap_p): Likewise.
3331 (stmt_could_throw_1_p): Likewise.
3332 * tree-inline.c (estimate_operator_cost): Likewise.
3333 * tree-pretty-print.c (dump_generic_node): Likewise.
3334 (op_code_prio): Likewise.
3335 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3336 * tree-ssa-operands.c (get_expr_operands): Likewise.
3337 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3338 * fold-const-call.h (fold_fma): Delete.
3339 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3340 CFN_FNMA and CFN_FNMS.
3342 * genmatch.c (combined_fn): New enum.
3343 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3344 (commutative_op): New function.
3345 (commutate): Use it. Handle more than 2 operands.
3346 (dt_operand::gen_gimple_expr): Use commutative_op.
3347 (parser::parse_expr): Allow :c to be used with non-binary
3348 operators if the commutative operand is known.
3349 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3350 CFN_FMS, CFN_FNMA and CFN_FNMS.
3351 (backprop::process_assign_use): Remove FMA_EXPR handling.
3352 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3353 (gen_hsa_fma): New function.
3354 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3355 IFN_FNMA and IFN_FNMS.
3356 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3357 * gimple-fold.h (follow_all_ssa_edges): Declare.
3358 * gimple-fold.c (follow_all_ssa_edges): New function.
3359 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3360 gimple_build interface and use follow_all_ssa_edges to fold the result.
3361 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3362 instead of checking for optabs directly.
3363 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3364 rather than FMA_EXPRs.
3365 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3366 call to IFN_FMA instead of an FMA_EXPR.
3368 2018-05-17 Jim Wilson <jimw@sifive.com>
3370 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3371 sign extended promoted subreg, and the range does not have the sign bit
3372 set, then do a sign extend.
3374 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3375 test, check for sign extended subreg and/or constant operands, and
3376 do a sign extend in that case.
3378 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3380 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3381 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3383 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3384 Change logics_shift_reg to logics_shift_imm.
3385 (thunderx2t99_fp_loadpair_basic): Delete.
3386 (thunderx2t99_fp_storepair_basic): Delete.
3387 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3388 (thunderx2t99_asimd_polynomial): Delete.
3389 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3390 and neon_fp_mul_d_scalar_q.
3391 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3392 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3393 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3394 (thunderx2t99_asimd_lut): Add missing tbl types.
3395 (thunderx2t99_asimd_ext): Delete.
3396 (thunderx2t99_asimd_load1_1_mult): Delete.
3397 (thunderx2t99_asimd_load1_2_mult): Delete.
3398 (thunderx2t99_asimd_load1_ldp): New.
3399 (thunderx2t99_asimd_load1): New.
3400 (thunderx2t99_asimd_load2): Add missing *load2* types.
3401 (thunderx2t99_asimd_load3): New.
3402 (thunderx2t99_asimd_load4): New.
3403 (thunderx2t99_asimd_store1_1_mult): Delete.
3404 (thunderx2t99_asimd_store1_2_mult): Delete.
3405 (thunderx2t99_asimd_store2_mult): Delete.
3406 (thunderx2t99_asimd_store2_onelane): Delete.
3407 (thunderx2t99_asimd_store_stp): New.
3408 (thunderx2t99_asimd_store1): New.
3409 (thunderx2t99_asimd_store2): New.
3410 (thunderx2t99_asimd_store3): New.
3411 (thunderx2t99_asimd_store4): New.
3413 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3415 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3416 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3418 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3419 Segher Boessenkool <segher@kernel.crashing.org>
3422 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3425 2018-05-17 Richard Biener <rguenther@suse.de>
3427 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3428 for pruning loop and prune defs feeding only already visited PHIs.
3430 2018-05-17 Richard Biener <rguenther@suse.de>
3432 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3434 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3435 Richard Biener <rguenther@suse.de>
3437 PR tree-optimization/85793
3438 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3439 for VMAT_ELEMENTWISE.
3441 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3443 * internal-fn.h (lookup_internal_fn): Declare
3444 * internal-fn.c (lookup_internal_fn): New function.
3445 * gimple.c (gimple_build_call_from_tree): Handle calls to
3447 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3448 internal function names.
3449 * tree-pretty-print.c (dump_generic_node): Likewise.
3450 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3452 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3454 * gimple-fold.h (gimple_build): Make the function forms take
3455 combined_fn rather than built_in_function.
3456 (gimple_simplify): Likewise.
3457 * gimple-match-head.c (gimple_simplify): Likewise.
3458 * gimple-fold.c (gimple_build): Likewise.
3459 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3460 rather than gimple_build_call_internal.
3461 (get_initial_defs_for_reduction): Likewise.
3462 (vect_create_epilog_for_reduction): Likewise.
3463 (vectorizable_live_operation): Likewise.
3465 2018-05-17 Martin Liska <mliska@suse.cz>
3467 * gimple-ssa-sprintf.c (format_directive): Do not use
3468 space in between 'G_' and '('.
3470 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3473 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3474 even if the mask is not all ones.
3477 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3479 (ix86_gimple_fold_builtin): Likewise.
3482 * config/i386/i386.c: Include tree-vector-builder.h.
3483 (ix86_vector_shift_count): New function.
3484 (ix86_fold_builtin): Fold shift builtins by scalar count.
3485 (ix86_gimple_fold_builtin): Likewise.
3487 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3488 _mm512_setzero): New intrinsics.
3490 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3491 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3493 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3494 code generation for cases where splatting a value is not useful.
3495 * simplify-rtx.c (simplify_ternary_operation): Simplify
3496 vec_merge across a vec_duplicate and a paradoxical subreg forming
3497 a vector mode to a vec_concat.
3499 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3501 * config.gcc: Support "goldmont-plus".
3502 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3504 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3505 PROCESSOR_GOLDMONT_PLUS.
3506 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3507 (processor_target_table): Add "goldmont-plus".
3508 (PTA_GOLDMONT_PLUS): Define.
3509 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3510 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3511 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3512 (fold_builtin_cpu): Add "goldmont-plus".
3513 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3514 (ix86_option_override_internal): Add "goldmont-plus".
3515 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3516 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3517 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3518 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3520 2018-05-17 Richard Biener <rguenther@suse.de>
3522 PR tree-optimization/85757
3523 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3524 remove defs that only feed that PHI from further processing.
3526 2018-05-16 Jim Wilson <jimw@sifive.com>
3528 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3530 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3532 2018-05-16 Mark Wielaard <mark@klomp.org>
3534 * dwarf2out.c (count_index_strings): New function.
3535 (output_indirect_strings): Call count_index_strings and generate
3536 header for dwarf_version >= 5.
3538 2018-05-16 Mark Wielaard <mark@klomp.org>
3540 * dwarf2out.c (dwarf_FORM): New function.
3541 (set_indirect_string): Use dwarf_FORM.
3542 (reset_indirect_string): Likewise.
3543 (size_of_die): Likewise.
3544 (value_format): Likewise.
3545 (output_die): Likewise.
3546 (add_skeleton_AT_string): Likewise.
3547 (output_macinfo_op): Likewise.
3548 (index_string): Likewise.
3549 (output_index_string_offset): Likewise.
3550 (output_index_string): Likewise.
3551 (count_index_strings): Likewise.
3553 2018-05-16 Carl Love <cel@us.ibm.com>
3555 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3556 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3558 2018-05-16 Martin Jambor <mjambor@suse.cz>
3560 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3561 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3563 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3565 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3566 (fnma<mode>4): Likewise.
3567 (fms<mode>4): Likewise.
3568 (fnms<mode>4): Likewise.
3569 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3570 (aarch64_fnma<mode>4): Likewise.
3571 (aarch64_fms<mode>4): Likewise.
3572 (aarch64_fnms<mode>4): Likewise.
3573 (aarch64_fnmadd<mode>4): Likewise.
3575 2018-05-16 Jason Merrill <jason@redhat.com>
3577 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3579 2018-05-16 Richard Biener <rguenther@suse.de>
3581 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3582 (dump_stmt_cost): Declare.
3583 (add_stmt_cost): Dump cost we add.
3584 (add_stmt_costs): New function.
3585 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3587 (vect_analyze_stmt): Adjust prototype.
3588 (vectorizable_condition): Likewise.
3589 (vectorizable_live_operation): Likewise.
3590 (vectorizable_reduction): Likewise.
3591 (vectorizable_induction): Likewise.
3592 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3593 cost vector to pass to vectorizable_ and record afterwards.
3594 (vect_model_reduction_cost): Take cost vector argument and adjust.
3595 (vect_model_induction_cost): Likewise.
3596 (vectorizable_reduction): Likewise.
3597 (vectorizable_induction): Likewise.
3598 (vectorizable_live_operation): Likewise.
3599 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3600 SLP_TREE_NUMBER_OF_VEC_STMTS.
3601 (vect_analyze_slp_cost_1): Remove.
3602 (vect_analyze_slp_cost): Likewise.
3603 (vect_slp_analyze_node_operations): Take visited args and
3604 a target cost vector. Avoid processing already visited stmt sets.
3605 (vect_slp_analyze_operations): Use a local cost vector to gather
3606 costs and register those of non-discarded instances.
3607 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3608 (vect_schedule_slp_instance): Remove copying of
3609 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3611 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3612 adding cost. Record cost entry location.
3613 (vect_prologue_cost_for_slp_op): Function to compute cost of
3614 a constant or invariant generated for SLP vect in the prologue,
3615 split out from vect_analyze_slp_cost_1.
3616 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3617 (vect_model_promotion_demotion_cost): Likewise.
3618 (vect_model_store_cost): Likewise, make static.
3619 (vect_model_load_cost): Likewise.
3620 (vectorizable_bswap): Add cost vector arg and adjust.
3621 (vectorizable_call): Likewise.
3622 (vectorizable_simd_clone_call): Likewise.
3623 (vectorizable_conversion): Likewise.
3624 (vectorizable_assignment): Likewise.
3625 (vectorizable_shift): Likewise.
3626 (vectorizable_operation): Likewise.
3627 (vectorizable_store): Likewise.
3628 (vectorizable_load): Likewise.
3629 (vectorizable_condition): Likewise.
3630 (vectorizable_comparison): Likewise.
3631 (can_vectorize_live_stmts): Likewise.
3632 (vect_analyze_stmt): Likewise.
3633 (vect_transform_stmt): Adjust calls to vectorizable_*.
3634 * tree-vectorizer.c: Include gimple-pretty-print.h.
3635 (dump_stmt_cost): New function.
3637 2018-05-16 Richard Biener <rguenther@suse.de>
3639 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3640 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3641 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3642 (check_name): New callback.
3643 (dse_classify_store): Track cycles via a visited bitmap of PHI
3644 defs and simplify handling of in-loop and across loop dead stores
3645 and properly fail for loop-variant refs. Handle byte-tracking with
3646 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3649 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3651 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3652 (vect_get_mask_type_for_stmt): Likewise.
3653 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3655 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3656 to determine the statement's vector type and the vector type that
3657 should be used for calculating nunits. Deal with cases in which
3658 the type has to be deferred.
3659 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3660 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3661 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3662 (vect_determine_vf_for_stmt): New functions, split out from...
3663 (vect_determine_vectorization_factor): ...here.
3664 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3665 (vect_get_mask_type_for_stmt): New functions, split out from
3666 vect_determine_vectorization_factor.
3668 2018-05-16 Richard Biener <rguenther@suse.de>
3670 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3671 verify the [VEC_]COND_EXPR embedded comparison.
3673 2018-05-15 Martin Sebor <msebor@redhat.com>
3675 PR tree-optimization/85753
3676 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3677 RECORD_TYPE in addition to ARRAY_TYPE.
3679 2018-05-15 Martin Sebor <msebor@redhat.com>
3682 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3684 2018-05-15 Richard Biener <rguenther@suse.de>
3686 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3687 add by_clobber_p one. Change algorithm to collect all defs
3688 representing uses we need to walk and try reducing them to
3689 a single one before failing.
3690 (dse_dom_walker::dse_optimize_stmt): Adjust.
3692 2018-05-13 Mark Wielaard <mark@klomp.org>
3694 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3695 (size_of_loc_descr): Likewise.
3696 (output_loc_operands): Likewise.
3697 (output_loc_operands_raw): Likewise.
3698 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3699 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3700 (hash_loc_operands): Likewise.
3701 (compare_loc_operands): Likewise.
3703 2018-05-14 Mark Wielaard <mark@klomp.org>
3705 * dwarf2out.c (count_index_addrs): New function.
3706 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3708 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3710 PR tree-optimization/83648
3711 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3712 return value as malloc candidate.
3714 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3717 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3718 param as true in call to suggest_attribute.
3720 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3722 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3723 -mreadonly-in-sdata.
3725 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3727 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3729 (aarch64_crypto_aesd_fused): Likewise.
3731 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3733 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3734 (movsi_aarch64): Likewise.
3735 (load_pairsi): Likewise.
3736 (load_pairdi): Likewise.
3737 (store_pairsi): Likewise.
3738 (store_pairdi): Likewise.
3739 (load_pairsf): Likewise.
3740 (load_pairdf): Likewise.
3741 (store_pairsf): Likewise.
3742 (store_pairdf): Likewise.
3743 (zero_extend): Likewise.
3744 (trunc): Swap alternatives.
3745 (fcvt_target): Add '?' to prefer w over r.
3747 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3750 * config/i386/i386.md: Disallow non-commutative arithmetics in
3751 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3752 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3753 in the peephole2 before it.
3755 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3757 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3758 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3759 (ix86_handle_option): Handle -mcldemote.
3760 * config.gcc: New header.
3761 * config/i386/cldemoteintrin.h: New file.
3762 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3763 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3765 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3766 OPTION_MASK_ISA_CLDEMOTE.
3767 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3768 (ix86_valid_target_attribute_inner_p): Ditto.
3769 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3770 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3771 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3772 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3773 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3775 * config/i386/i386.opt: Add -mcldemote.
3776 * config/i386/x86intrin.h: New header.
3777 * doc/invoke.texi: Add -mcldemote.
3779 2018-05-14 Richard Biener <rguenther@suse.de>
3781 * doc/match-and-simplify.texi: Adjust :s documentation.
3783 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3785 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3786 intended memcpy size.
3787 (REORDER_45): Likewise.
3789 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3791 * sort.cc: New file.
3792 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3793 * vec.c (qsort_chk): Use gcc_qsort.
3794 * Makefile.in (OBJS-libcommon): Add sort.o.
3795 (build/sort.o): New target. Use it...
3796 (BUILD_RTL): ... here, and...
3797 (build/gencfn-macros): ... here, and...
3798 (build/genmatch): ... here.
3800 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3801 Chung-Ju Wu <jasonwucj@gmail.com>
3803 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3804 * config/nds32/nds32-graywolf.md: New file.
3805 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3806 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3808 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3809 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3810 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3811 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3812 * config/nds32/pipelines.md: Include n15 settings.
3814 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3815 Chung-Ju Wu <jasonwucj@gmail.com>
3817 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3818 * config/nds32/nds32-n13.md: New file.
3819 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3820 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3822 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3823 * config/nds32/nds32.md (pipeline_model): Add n13.
3824 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3825 * config/nds32/pipelines.md: Include n13 settings.
3827 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3828 Chung-Ju Wu <jasonwucj@gmail.com>
3830 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3831 * config/nds32/nds32-n10.md: New file.
3832 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3833 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3835 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3836 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3837 * config/nds32/nds32.md (pipeline_model): Add n10.
3838 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3839 * config/nds32/pipelines.md: Include n10 settings.
3841 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3842 Kito Cheng <kito.cheng@gmail.com>
3843 Chung-Ju Wu <jasonwucj@gmail.com>
3845 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3846 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3847 Add enum values for DSP extension instructions.
3848 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3850 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3851 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3853 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3854 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3855 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3856 * config/nds32/nds32-intrinsic.md: Likewise.
3857 * config/nds32/nds32_intrinsic.h: Likewise.
3858 * config/nds32/nds32-md-auxiliary.c: Likewise.
3859 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3860 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3861 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3862 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3863 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3864 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3865 TYPE_DMAC in switch statement.
3866 * config/nds32/nds32.c: New checking and implementation for DSP
3867 extension instructions.
3868 * config/nds32/nds32.h: Likewise.
3869 * config/nds32/nds32.md: Likewise.
3870 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3871 * config/nds32/predicates.md: Implement new predicates for DSP
3874 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3876 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3877 Reformat alternatives and attributes so it is easier to identify
3878 which constraints/attributes go with which instruction.
3879 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3880 (mov<mode>_softfloat32, FMOVE64): Likewise.
3881 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3882 (mov<mode>_softfloat64, FMOVE64): Likewise.
3884 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3886 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3888 (Basic PowerPC Built-in Functions): The new name of the
3889 subsection previously known as "PowerPC Built-in Functions".
3890 (Basic PowerPC Built-in Functions Available on all Configurations):
3892 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3893 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3894 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3895 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3897 2018-05-11 Martin Jambor <mjambor@suse.cz>
3900 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3903 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3906 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3908 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3910 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3911 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
3912 (ix86_handle_option): Handle -mwaitpkg.
3913 * config.gcc: New header.
3914 * config/i386/cpuid.h (bit_WAITPKG): New bit.
3915 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
3916 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
3918 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3919 OPTION_MASK_ISA_WAITPKG.
3920 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
3921 (ix86_option_override_internal): Add PTA_WAITPKG.
3922 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
3923 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
3924 IX86_BUILTIN_TPAUSE.
3925 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
3926 __builtin_ia32_umwait and __builtin_ia32_tpause.
3927 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
3928 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
3929 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
3930 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
3931 UNSPECV_TPAUSE): New.
3932 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
3933 * config/i386/i386.opt: Add -mwaitpkg.
3934 * config/i386/waitpkgintrin.h: New file.
3935 * config/i386/x86intrin.h: New header.
3936 * doc/invoke.texi: Add -mwaitpkg.
3938 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3941 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
3943 (cortex-m0): Use armv6s-m isa.
3944 (cortex-m0plus): Likewise.
3945 (cortex-m1): Likewise.
3946 (cortex-m0.small-multiply): Likewise.
3947 (cortex-m0plus.small-multiply): Likewise.
3948 (cortex-m1.small-multiply): Likewise.
3950 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
3951 Jakub Jelinek <jakub@redhat.com>
3953 PR tree-optimization/85692
3954 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
3955 source permute as well.
3957 2018-05-11 Martin Liska <mliska@suse.cz>
3960 * doc/extend.texi: Document LLVM style format for no_sanitize
3963 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
3965 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
3966 mode_supports_vsx_dform_quad to mode_supports_dq_form.
3967 (mode_supports_vsx_dform_quad): Likewise.
3968 (mode_supports_vmx_dform): Move these functions to be next to the
3969 other mode_supports functions.
3970 (mode_supports_dq_form): Likewise.
3971 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
3972 mode_supports_dq_form.
3973 (reg_offset_addressing_ok_p): Likewise.
3974 (offsettable_ok_by_alignment): Likewise.
3975 (rs6000_legitimate_offset_address_p): Likewise.
3976 (legitimate_lo_sum_address_p): Likewise.
3977 (rs6000_legitimize_address): Likewise.
3978 (rs6000_legitimize_reload_address): Likewise.
3979 (rs6000_secondary_reload_inner): Likewise.
3980 (rs6000_preferred_reload_class): Likewise.
3981 (rs6000_output_move_128bit): Likewise.
3983 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3985 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
3986 Generate SImode target register for null target.
3987 <case IX86_BUILTIN_XGETBV>: Ditto.
3988 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
3989 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
3991 2018-05-10 Carl Love <cel@us.ibm.com>
3993 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3994 dcbtt and dcbtstt if operands[2] is 0.
3996 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
3999 * config/i386/sse.md (usadv64qi): New expander.
4001 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4003 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4004 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4005 -maltivec=be support.
4006 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4007 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4008 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4009 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4010 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4011 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4012 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4013 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4014 altivec_vsumsws): Adjust.
4015 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4016 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4017 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4019 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4020 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4021 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4022 (altivec_lve<VI_char>x): Delete expand.
4023 (*altivec_lve<VI_char>x_internal): Rename to...
4024 (altivec_lve<VI_char>x): ... this.
4025 (altivec_lvxl_<mode>): Delete expand.
4026 (*altivec_lvxl_<mode>_internal): Rename to ...
4027 (altivec_lvxl_<mode>): ... this.
4028 (altivec_stvxl_<mode>): Delete expand.
4029 (*altivec_stvxl_<mode>_internal): Rename to ...
4030 (altivec_stvxl_<mode>): ... this.
4031 (altivec_stve<VI_char>x): Delete expand.
4032 (*altivec_stve<VI_char>x_internal): Rename to ...
4033 (altivec_stve<VI_char>x): ... this.
4034 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4035 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4036 reduc_plus_scal_<mode>): Adjust.
4037 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4039 (rs6000_cpu_cpp_builtins): Adjust.
4040 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4041 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4042 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4043 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4044 -maltivec=be support.
4045 (rs6000_split_vec_extract_var): Adjust.
4046 (rs6000_split_v4si_init): Adjust.
4047 (swap_selector_for_mode): Delete.
4048 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4049 altivec_expand_stvex_be): Delete.
4050 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4051 -maltivec=be support.
4052 (rs6000_gimple_fold_builtin): Ditto.
4053 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4055 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4056 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4057 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4058 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4059 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4060 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4061 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4062 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4063 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4064 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4065 anonymous split): Adjust.
4066 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4067 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4069 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4071 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4072 when --with-gxx-include-dir is also specified.
4073 * configure: Regenerate.
4075 2018-05-09 Jim Wilson <jimw@sifive.com>
4078 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4079 * config/riscv/t-withmultilib: New.
4080 * config/riscv/withmultilib.h: New.
4081 * doc/install.texi: Document RISC-V --with-multilib-list support.
4083 2018-05-09 Richard Biener <rguenther@suse.de>
4085 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4087 (vect_bb_vectorization_profitable_p): Adjust. Compute
4088 actual scalar cost using the cost vector and the add_stmt_cost
4091 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4093 PR rtl-optimization/85645
4094 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4095 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4097 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4099 PR rtl-optimization/85645
4100 * shrink-wrap.c (spread_components): Return a boolean saying if
4101 anything was changed.
4102 (try_shrink_wrapping_separate): Iterate spread_components until
4103 nothing changes anymore.
4105 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4107 PR rtl-optimization/85645
4108 * regrename.c (build_def_use): Also kill the chains that include the
4109 destination of a REG_CFA_REGISTER note.
4111 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4113 PR rtl-optimization/85645
4114 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4115 insn that has a REG_CFA_REGISTER note.
4117 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4119 * cfgexpand.c (expand_clobber): New function.
4120 (expand_gimple_stmt_1): Use it.
4121 * tree-vect-stmts.c (vect_clobber_variable): New function,
4123 (vectorizable_simd_clone_call): ...here.
4124 (vectorizable_store): Emit a clobber either side of an
4125 IFN_STORE_LANES sequence.
4126 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4128 2018-05-09 Tom de Vries <tom@codesourcery.com>
4131 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4132 (define_insn "trap_if_false"): Add exit after trap.
4134 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4136 PR rtl-optimization/85638
4137 * bb-reorder.c: Include common/common-target.h.
4138 (create_forwarder_block): New function extracted from...
4139 (fix_up_crossing_landing_pad): ...here. Rename into...
4140 (dw2_fix_up_crossing_landing_pad): ...this.
4141 (sjlj_fix_up_crossing_landing_pad): New function.
4142 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4143 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4144 from both partitions and exit the loop after one iteration.
4146 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4149 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4151 (Basic PowerPC Built-in Functions): The new name of the
4152 subsection previously known as "PowerPC Built-in Functions".
4153 (Basic PowerPC Built-in Functions Available on all Configurations):
4155 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4157 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4159 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4161 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4164 2018-05-08 Jim Wilson <jimw@sifive.com>
4166 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4167 (LD_EMUL_SUFFIX): New.
4168 (LINK_SPEC): Use it.
4170 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4172 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4174 (Basic PowerPC Built-in Functions): The new name of the
4175 subsection previously known as "PowerPC Built-in Functions".
4176 (Basic PowerPC Built-in Functions Available on all Configurations):
4178 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4180 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4182 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4184 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4187 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4190 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4191 after cmpelim optimization.
4193 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4195 * config.gcc: Support "goldmont".
4196 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4197 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4199 * config/i386/i386.c (m_GOLDMONT): Define.
4200 (processor_target_table): Add "goldmont".
4201 (PTA_GOLDMONT): Define.
4202 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4203 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4204 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4205 (fold_builtin_cpu): Add "goldmont".
4206 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4207 (ix86_option_override_internal): Add "goldmont".
4208 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4209 (processor_type): Add PROCESSOR_GOLDMONT.
4210 * config/i386/i386.md: Add CPU "glm".
4211 * config/i386/glm.md: New file.
4212 * config/i386/x86-tune.def: Add m_GOLDMONT.
4213 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4215 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4218 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4220 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4221 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4222 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4225 * config/i386/i386.c (ix86_fold_builtin): Handle
4226 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4229 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4230 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4232 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4235 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4236 (check_arch): Likewise.
4237 (check_fpu): Return the result rather than printing it.
4238 (end arch): Fix operator precedence.
4239 (end cpu): Likewise.
4240 (END): Print the result from check_fpu.
4242 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4243 Alan Hayward <alan.hayward@arm.com>
4244 David Sherwood <david.sherwood@arm.com>
4246 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4247 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4248 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4249 (*fcmuo<mode>_and): New patterns.
4251 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4253 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4254 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4255 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4256 (cmp_op, sve_imm_con): New code attributes.
4257 (SVE_COND_INT_CMP, imm_con): Delete.
4258 (cmp_op): Remove above unspecs from int attribute.
4259 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4261 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4262 comparison-specific unspecs.
4263 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4264 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4265 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4266 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4267 (*vec_fcm<cmp_op><mode>): Rename to...
4268 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4269 (*vec_fcmuo<mode>): Rename to...
4270 (*fcmuo<mode>): ...this and adjust likewise.
4271 (*pred_fcm<cmp_op><mode>): New pattern.
4272 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4273 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4275 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4277 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4278 (aarch64_emit_sve_predicated_cond): New function.
4279 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4280 (aarch64_emit_unspec_cond_or): Replace with...
4281 (aarch64_emit_sve_or_conds): ...this new function. Use
4282 aarch64_emit_sve_ptrue_op for the individual comparisons and
4283 aarch64_emit_binop to OR them together.
4284 (aarch64_emit_inverted_unspec_cond): Replace with...
4285 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4286 aarch64_emit_sve_ptrue_op for the comparison and
4287 aarch64_emit_unop to invert the result.
4288 (aarch64_expand_sve_vec_cmp_float): Update after the above
4289 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4291 2018-05-07 Nathan Sidwell <nathan@acm.org>
4293 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4294 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4295 (Backwards Compatibility): Likewise.
4297 2018-05-07 Luis Machado <luis.machado@linaro.org>
4301 2018-05-07 Luis Machado <luis.machado@linaro.org>
4303 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4304 <prefetch_dynamic_strides>: New const bool field.
4305 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4306 prefetch_dynamic_strides.
4307 (exynosm1_prefetch_tune): Likewise.
4308 (thunderxt88_prefetch_tune): Likewise.
4309 (thunderx_prefetch_tune): Likewise.
4310 (thunderx2t99_prefetch_tune): Likewise.
4311 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4313 (aarch64_override_options_internal): Update to set
4314 PARAM_PREFETCH_DYNAMIC_STRIDES.
4315 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4316 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4317 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4318 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4319 prefetch-dynamic-strides setting.
4321 2018-05-07 Luis Machado <luis.machado@linaro.org>
4323 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4324 <minimum_stride>: New const int field.
4325 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4326 minimum_stride field.
4327 (exynosm1_prefetch_tune): Likewise.
4328 (thunderxt88_prefetch_tune): Likewise.
4329 (thunderx_prefetch_tune): Likewise.
4330 (thunderx2t99_prefetch_tune): Likewise.
4331 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4332 (aarch64_override_options_internal): Update to set
4333 PARAM_PREFETCH_MINIMUM_STRIDE.
4334 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4335 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4336 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4337 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4338 stride is constant and is below the minimum stride threshold.
4340 2018-05-07 Luis Machado <luis.machado@linaro.org>
4342 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4345 2018-05-07 Luis Machado <luis.machado@linaro.org>
4347 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4348 <prefetch_dynamic_strides>: New const bool field.
4349 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4350 prefetch_dynamic_strides.
4351 (exynosm1_prefetch_tune): Likewise.
4352 (thunderxt88_prefetch_tune): Likewise.
4353 (thunderx_prefetch_tune): Likewise.
4354 (thunderx2t99_prefetch_tune): Likewise.
4355 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4357 (aarch64_override_options_internal): Update to set
4358 PARAM_PREFETCH_DYNAMIC_STRIDES.
4359 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4360 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4361 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4362 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4363 prefetch-dynamic-strides setting.
4365 2018-05-07 Luis Machado <luis.machado@linaro.org>
4367 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4368 <minimum_stride>: New const int field.
4369 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4370 minimum_stride field.
4371 (exynosm1_prefetch_tune): Likewise.
4372 (thunderxt88_prefetch_tune): Likewise.
4373 (thunderx_prefetch_tune): Likewise.
4374 (thunderx2t99_prefetch_tune): Likewise.
4375 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4376 (aarch64_override_options_internal): Update to set
4377 PARAM_PREFETCH_MINIMUM_STRIDE.
4378 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4379 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4380 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4381 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4382 stride is constant and is below the minimum stride threshold.
4384 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4387 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4388 the type is addressable. Don't force op into register if it has
4391 2018-05-05 Roland McGrath <mcgrathr@google.com>
4394 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4395 any section for which we don't know a specific type it should have,
4396 regardless of name. Previously this was done only for the exact
4397 names ".init_array", ".fini_array", and ".preinit_array".
4398 (default_elf_asm_named_section): Add comment about
4399 relationship with default_section_type_flags and SECTION_NOTYPE.
4400 (get_section): Don't consider it a type conflict if one side has
4401 SECTION_NOTYPE and the other doesn't, as long as neither has the
4402 SECTION_BSS et al used in the default_section_type_flags logic.
4404 2018-05-05 Tom de Vries <tom@codesourcery.com>
4407 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4408 (workaround_barsyncs): New function.
4409 (nvptx_reorg): Use workaround_barsyncs.
4410 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4411 (define_expand "nvptx_membar_cta"): New define_expand.
4412 (define_insn "*nvptx_membar_cta"): New insn.
4414 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4416 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4417 To improve optimization opportunities.
4418 * builtin-types.def: The new needed builtin types for the above.
4420 2018-05-04 Richard Biener <rguenther@suse.de>
4422 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4423 * gimple-ssa-store-merging.c
4424 (imm_store_chain_info::output_merged_store): Remove redundant create,
4425 release split_store vector contents on failure.
4426 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4427 scalar stmt vector on cache hit.
4429 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4431 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4433 * config.gcc (powerpc-xilinx-eabi*): Remove.
4434 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4436 (fusion_addis_mem_combo_load): Ditto.
4437 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4439 (rs6000_cpu_cpp_builtins): Ditto.
4440 * config/rs6000/rs6000-linux.c
4441 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4442 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4443 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4445 (rs6000_setup_reg_addr_masks): Ditto.
4446 (rs6000_init_hard_regno_mode_ok): Ditto.
4447 (rs6000_option_override_internal): Ditto.
4448 (legitimate_lo_sum_address_p): Ditto.
4449 (rs6000_legitimize_address): Ditto.
4450 (rs6000_legitimize_reload_address): Ditto.
4451 (rs6000_legitimate_address_p): Ditto.
4452 (abi_v4_pass_in_fpr): Ditto.
4453 (setup_incoming_varargs): Ditto.
4454 (rs6000_gimplify_va_arg): Ditto.
4455 (rs6000_split_multireg_move): Ditto.
4456 (rs6000_savres_strategy): Ditto.
4457 (rs6000_emit_prologue_components): Ditto.
4458 (rs6000_emit_epilogue_components): Ditto.
4459 (rs6000_emit_prologue): Ditto.
4460 (rs6000_emit_epilogue): Ditto.
4461 (rs6000_elf_file_end): Ditto.
4462 (rs6000_function_value): Ditto.
4463 (rs6000_libcall_value): Ditto.
4464 * config/rs6000/rs6000.h: Ditto.
4465 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4466 (TARGET_MINMAX): ... this. New.
4467 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4468 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4469 (*movsi_internal1_single): Delete.
4470 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4471 mfpu=, mxilinx-fpu): Delete.
4472 * config/rs6000/singlefp.h: Delete.
4473 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4474 * config/rs6000/t-rs6000: Ditto.
4475 * config/rs6000/t-xilinx: Delete.
4476 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4477 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4478 (VStype_simple): Delete.
4479 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4480 * config/rs6000/xfpu.h: Delete.
4481 * config/rs6000/xfpu.md: Delete.
4482 * config/rs6000/xilinx.h: Delete.
4483 * config/rs6000/xilinx.opt: Delete.
4484 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4485 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4487 2018-05-04 Tom de Vries <tom@codesourcery.com>
4490 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4493 2018-05-04 Richard Biener <rguenther@suse.de>
4496 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4497 (expand_complex_div_wide): Likewise.
4498 (expand_complex_operations_1): Likewise.
4499 (expand_complex_libcall): Preserve EH info of the original stmt.
4500 (tree_lower_complex): Handle removed blocks.
4501 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4502 on complex multiplication and division libcall builtins.
4504 2018-05-04 Richard Biener <rguenther@suse.de>
4507 * fold-const.c (negate_expr_p): Restrict negation of operand
4508 zero of a division to when we know that can happen without
4510 (fold_negate_expr_1): Likewise.
4512 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4515 * real.h (real_nextafter): Declare.
4516 * real.c (real_nextafter): New function.
4517 * fold-const-call.c (fold_const_nextafter): New function.
4518 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4519 CASE_CFN_NEXTTOWARD.
4520 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4521 even when arg1_mode is different from arg0_mode.
4523 2018-05-03 Nathan Sidwell <nathan@acm.org>
4525 * doc/extend.texi (Deprecated Features): Remove
4527 (Backwards Compatibility): Likewise.
4528 * doc/invoke.texi (C++ Language Options): Likewise.
4529 (C++ Dialect Options): Likewise.
4531 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4534 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4535 _mm512_mask_mullox_epi64): New intrinsics.
4537 2018-05-03 Tom de Vries <tom@codesourcery.com>
4540 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4541 dump files): Add offload-tree.
4543 2018-05-03 Richard Biener <rguenther@suse.de>
4545 PR tree-optimization/85615
4546 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4547 to loops not nested in BBs loop father to avoid creating multi-entry
4550 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4552 PR tree-optimization/70291
4553 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4554 arguments. Change return type to tree. Emit libcall as a new
4555 statement rather than replacing existing one when inplace_p is true.
4556 (expand_complex_multiplication_components): New function.
4557 (expand_complex_multiplication): Expand floating-point complex
4558 multiplication using the above.
4559 (expand_complex_division): Rename inner_type parameter to type.
4560 Update expand_complex_libcall call-site.
4561 (expand_complex_operations_1): Update expand_complex_multiplication
4562 and expand_complex_division call-sites.
4564 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4567 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4568 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4569 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4570 the highest significant bit of the shift count mask is clear. In
4571 check whether and[sq]i3 is needed verify that all significant bits
4572 of the shift count other than the highest are set.
4574 2018-05-02 Tom de Vries <tom@codesourcery.com>
4577 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4578 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4579 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4580 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4581 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4582 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4583 __builtin_goacc_parlevel_size.
4585 2018-05-02 Richard Biener <rguenther@suse.de>
4587 PR tree-optimization/85597
4588 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4589 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4592 2018-05-02 Tom de Vries <tom@codesourcery.com>
4595 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4596 dump files): Add ltrans-tree.
4598 2018-05-02 Tom de Vries <tom@codesourcery.com>
4601 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4602 dump files): Add wpa-ipa.
4604 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4606 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4607 powerpc*-*-linux*paired* target.
4608 * config/rs6000/750cl.h: Delete.
4609 * config/rs6000/paired.h: Delete.
4610 * config/rs6000/paired.md: Delete.
4611 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4613 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4614 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4615 comment. Remove paired float support.
4616 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4617 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4619 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4620 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4622 * config/rs6000/rs6000.c: Remove paired float support.
4623 (paired_expand_vector_init, paired_expand_vector_move,
4624 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4625 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4626 paired_expand_builtin, paired_expand_predicate_builtin,
4627 paired_init_builtins): Delete.
4628 * config/rs6000/rs6000.h: Remove paired float support.
4629 * config/rs6000/rs6000.md: Remove paired float support.
4630 (move_from_CR_ov_bit): Delete.
4631 * config/rs6000/rs6000.opt (mpaired): Delete.
4632 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4633 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4635 2018-05-02 Richard Biener <rguenther@suse.de>
4638 * gimplify.c (gimplify_save_expr): When in SSA form allow
4639 SAVE_EXPRs to compute to SSA vars.
4641 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4644 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4645 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4646 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4647 clobber operands[2], instead use a new pseudo. Formatting fixes.
4649 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4651 PR tree-optimization/85586
4652 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4653 exit early for statements in the same group if the accesses are
4656 2018-05-02 Tom de Vries <tom@codesourcery.com>
4659 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4662 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4664 PR tree-optimization/85143
4665 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4667 2018-05-01 Tom de Vries <tom@codesourcery.com>
4670 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4671 not found" error message.
4673 2018-05-01 Tom de Vries <tom@codesourcery.com>
4676 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4677 * vec.c (test_ordered_remove_if): New function.
4678 (vec_c_tests): Call test_ordered_remove_if.
4679 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4680 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4681 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4682 VEC_ORDERED_REMOVE_IF.
4684 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4686 PR tree-optimization/82665
4687 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4688 pointer subtraction where arguments come from a memchr call.
4690 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4692 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4693 --push-state --as-needed and --pop-state instead of --as-needed and
4694 --no-as-needed if ld supports it.
4695 * configure: Regenerated.
4698 * doc/install.texi2html: Replace _002d with - and _002a with * in
4699 generated html files using sed.
4701 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4704 * gcc-rich-location.c (blank_line_before_p): New function.
4705 (use_new_line): New function.
4706 (gcc_rich_location::add_fixit_insert_formatted): New function.
4707 * gcc-rich-location.h
4708 (gcc_rich_location::add_fixit_insert_formatted): New function.
4710 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4712 * selftest.c (assert_streq): Rename "expected" and "actual" to
4713 "val1" and "val2". Extend NULL-handling to cover both inputs
4714 symmetrically, while still requiring both to be non-NULL for a pass.
4715 * selftest.h (assert_streq): Rename "expected" and "actual" to
4717 (ASSERT_EQ): Likewise.
4718 (ASSERT_EQ_AT): Likewise.
4719 (ASSERT_KNOWN_EQ): Likewise.
4720 (ASSERT_KNOWN_EQ_AT): Likewise.
4721 (ASSERT_NE): Likewise.
4722 (ASSERT_MAYBE_NE): Likewise.
4723 (ASSERT_MAYBE_NE_AT): Likewise.
4724 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4725 the assertion to pass.
4726 (ASSERT_STREQ_AT): Likewise.
4728 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4730 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4731 interaction with -pie.
4733 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4735 * selftest.h: Fix alphabetization of per-source-file selftest
4738 2018-04-30 Jason Merrill <jason@redhat.com>
4740 PR c++/61982 - dead stores to destroyed objects.
4741 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4744 2018-04-30 Jason Merrill <jason@redhat.com>
4746 * tree.c (build_clobber): New.
4747 * tree.h: Declare it.
4748 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4750 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4752 * diagnostic-show-locus.c (layout::layout): Update for
4753 location_get_source_line returning a char_span.
4754 (struct char_span): Move to input.h.
4755 (struct correction): Update for fields in char_span becoming
4757 (struct source_line): Update for location_get_source_line
4758 returning a char_span.
4759 (layout::print_line): Likewise.
4760 * edit-context.c (edited_file::print_content): Likewise.
4761 (edited_file::print_diff_hunk): Likewise.
4762 (edited_file::print_run_of_changed_lines): Likewise.
4763 (edited_file::get_num_lines): Likewise.
4764 (edited_line::edited_line): Likewise.
4765 * final.c (asm_show_source): Likewise.
4766 * input.c (location_get_source_line): Convert return type
4767 from const char * to char_span, losing the final "line_len"
4769 (dump_location_info): Update for the above.
4770 (get_substring_ranges_for_loc): Likewise. Use a char_span
4771 when handling the literal within the line.
4772 (test_reading_source_line): Update for location_get_source_line
4773 returning a char_span.
4774 * input.h (class char_span): Move here from
4775 diagnostic-show-locus.c, converting from a struct to a class.
4776 Make data members private.
4777 (char_span::operator bool): New.
4778 (char_span::length): New.
4779 (char_span::get_buffer): New.
4780 (char_span::operator[]): New.
4781 (char_span::subspan): Make const.
4782 (char_span::xstrdup): New.
4783 (location_get_source_line): Convert return type from const char *
4784 to char_span, losing the final "line_size" param.
4786 2018-04-30 Jan Hubicka <jh@suse.cz>
4788 * lto-wrapper.c (ltrans_priorities): New static var.
4789 (cmp_priority): New.
4790 (run_gcc): Read priorities and if doing parallel build order
4791 the Makefile by them.
4793 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4795 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4797 2018-04-30 Richard Biener <rguenther@suse.de>
4799 * tree-cfg.c (verify_address): Remove base argument, add
4800 flag whether to check TREE_ADDRESSABLE and do that.
4801 (verify_expr): Remove.
4802 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4803 (verify_gimple_assign_single): Likewise.
4804 (verify_gimple_switch): Likewise.
4805 (verify_expr_location_1): Dereference tp once. Add (disabled)
4806 piece from verify_expr.
4807 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4809 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4811 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4813 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4815 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4816 (small_data_pattern): Likewise.
4817 (arc_rewrite_small_data): Likewise.
4818 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4819 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4820 (get_symbol_alignment): New function.
4821 (legitimate_small_data_address_p): Likewise.
4822 (legitimate_scaled_address): Update, call
4823 legitimate_small_data_address_p.
4824 (output_sdata): New static variable.
4825 (arc_print_operand): Update how we handle small data operands.
4826 (arc_print_operand_address): Likewise.
4827 (arc_legitimate_address_p): Update, use
4828 legitimate_small_data_address_p.
4829 (arc_rewrite_small_data_p): Remove.
4830 (arc_rewrite_small_data_1): Likewise.
4831 (arc_rewrite_small_data): Likewise.
4832 (small_data_pattern): Likewise.
4833 (compact_sda_memory_operand): Update to use
4834 legitimate_small_data_address_p and get_symbol_alignment.
4835 (prepare_move_operands): Don't rewite sdata pattern.
4836 (prepare_extend_operands): Remove.
4837 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4839 (zero_extendqisi2): Likewise.
4840 (zero_extendhisi2): Likewise.
4841 (extendqihi2): Likewise.
4842 (extendqisi2): Likewise.
4843 (extendhisi2): Likewise.
4847 * config/arc/constraints.md (Usd): Change it to memory constraint.
4849 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4851 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4852 as source of std instructions.
4853 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4854 allow 6-bit constants as source for store instructions.
4855 (movdi_insn): Update instruction pattern to allow 6-bit constants
4856 as source for store instructions.
4858 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4860 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4862 2018-04-30 Nathan Sidwell <nathan@acm.org>
4863 Sandra Loosemore <sandra@codesourcery.com>
4865 * dumpfile.c (dump_open): Allow '-' for stdout.
4866 * doc/invoke.texi (Developer Options): Document dump filename
4867 determination early. Document stdin/stdout selection.
4869 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4871 Microblaze Target: PIC data text relative
4873 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4874 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4876 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4877 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4878 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4879 New addressing mode for data-text relative position indepenedent code.
4880 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4881 'ADDRESS_SYMBOLIC_TXT_REL'.
4882 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4883 (microblaze_legitimate_pic_operand): Exclude function calls from
4884 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4885 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4887 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4888 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4889 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4890 for 'address + offset'.
4891 (microblaze_expand_prologue): Add new function prologue call for
4893 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4894 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4895 table in case of TARGET_PIC_DATA_TEXT_REL.
4896 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4897 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4898 Add new macros 'UNSPEC_TEXT',
4899 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4900 + exclude function calls from 'UNSPEC_PLT' in case of data text
4902 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4903 new target hook for generating address diff vector tables in case of
4905 * doc/tm.texi : Regenerate.
4906 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4907 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4908 of addr diff vector generation.
4909 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4910 target hook definition.
4911 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
4912 Add default function for generate_pic_addr_diff_vec -> flag_pic.
4913 * doc/invoke.texi (Add new pic option): Add new microblaze pic
4914 option for data text relative.
4916 2018-04-30 Richard Biener <rguenther@suse.de>
4918 * tree-chrec.h (evolution_function_is_constant_p): Remove
4920 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
4922 2018-04-30 Richard Biener <rguenther@suse.de>
4925 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
4927 2018-04-30 Richard Biener <rguenther@suse.de>
4929 PR tree-optimization/28364
4930 PR tree-optimization/85275
4931 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
4932 copying first exit test.
4934 2018-04-28 Mark Wielaard <mark@klomp.org>
4936 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
4938 (dwarf_AT): Handle DW_AT_addr_base.
4939 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
4941 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
4944 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
4945 (*ashl<dwi>3_doubleword_mask_1): Ditto.
4946 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
4947 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
4949 2018-04-28 Richard Biener <rguenther@suse.de>
4951 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
4952 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
4953 to reflect use. Only add interesting stmts.
4955 2018-04-27 Martin Jambor <mjambor@suse.cz>
4958 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
4959 the jump function allows for passing through aggregate values.
4961 2018-04-27 David Malcolm <dmalcolm@redhat.com>
4963 * input.h (in_system_header_at): Convert from macro to inline
4965 (from_macro_expansion_at): Likewise.
4966 (from_macro_definition_at): Likewise.
4968 2018-04-27 Jeff Law <law@redhat.com>
4970 * config.gcc: Mark tile* targets as deprecated/obsolete.
4972 2018-04-27 Richard Biener <rguenther@suse.de>
4974 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
4977 2018-04-27 Richard Biener <rguenther@suse.de>
4979 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
4981 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
4983 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
4984 with Yd constraint. Set "preferred_for_speed" attribute from
4985 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
4987 (*movdi_internal): Ditto.
4988 (movti_interunit splitters): Remove
4989 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
4990 (movdi_interunit splitters): Ditto.
4991 * config/i386/constraints.md (Ye): Remove.
4992 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
4994 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4997 * config/aarch64/constraints.md (Usg): Limit to 31.
5000 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5002 PR tree-optimization/85529
5003 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5004 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5005 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5006 zero extension or masking of the MSB bit.
5007 (optimize_range_tests): Add FIRST_BB argument, pass it through
5008 to optimize_range_tests_var_bound.
5009 (maybe_optimize_range_tests, reassociate_bb): Adjust
5010 optimize_range_tests callers.
5012 2018-04-26 Richard Biener <rguenther@suse.de>
5013 Jakub Jelinek <jakub@redhat.com>
5015 * cgraph.h (symbol_table): Just declare debug method here.
5016 * symtab.c (symbol_table::debug): Define.
5018 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5020 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5022 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5024 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5025 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5026 (*movdi_internal): Substitute Yi and Yj constraint with x
5027 and Ym and Yn constraint with y constraint. Update "isa"
5028 attribute and set "preferred_for_speed" attribute from
5029 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5030 (*movsi_internal): Ditto.
5031 (*movdf_internal): Ditto.
5032 (*movsf_internal): Ditto.
5033 (*zero_extendsidi2): Ditto.
5034 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5035 (sse2_loadld): Ditto.
5036 (*vec_extract<ssevecmodelower>_0): Ditto.
5037 (*vec_extractv4si_0_zext_sse4): Ditto.
5038 (vec_concatv2di): Ditto.
5039 (*vec_dup<mode>): Ditto.
5040 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5041 * config/i386/constraints.md (Yi): Remove.
5046 2018-04-26 Nathan Sidwell <nathan@acm.org>
5048 * dumpfile.c (dump_open): New.
5049 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5050 (dump_finish): Detect stdio/stderr by value not name.
5052 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5054 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5056 2018-04-26 Tom de Vries <tom@codesourcery.com>
5059 * config/nvptx/nvptx.c (verify_neutering_jumps)
5060 (verify_neutering_labels): New function
5061 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5063 2018-04-26 Tom de Vries <tom@codesourcery.com>
5066 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5067 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5070 2018-04-26 Richard Biener <rguenther@suse.de>
5071 Tom de Vries <tom@codesourcery.com>
5074 * lto-streamer-out.c (output_function): Fixup loops if required to match
5075 discovery done in the reader.
5077 2018-04-26 Richard Biener <rguenther@suse.de>
5079 PR tree-optimization/85116
5080 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5081 have a loop exit from the single latch predecessor. Remove
5082 case of header with just condition.
5083 (ch_base::copy_headers): Exclude infinite loops from any
5085 (pass_ch::execute): Record exits.
5087 2018-04-26 Richard Biener <rguenther@suse.de>
5089 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5090 prologue cost vector and pass it to vect_get_load_cost.
5091 (vect_get_peeling_costs_all_drs): Likewise.
5092 (vect_peeling_hash_get_lowest_cost): Likewise.
5093 (vect_enhance_data_refs_alignment): Likewise.
5095 2018-04-26 Richard Biener <rguenther@suse.de>
5098 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5099 checking of integer<->pointer conversions.
5100 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5101 sign-/zero-extending pointer types.
5102 (expand_omp_for_static_chunk): Likewise.
5104 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5105 Jean Lee <xiaoyur347@gmail.com>
5107 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5108 (TARGET_ASAN_SHADOW_OFFSET): Define.
5109 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5110 true for -fsanitize=address.
5112 2018-04-25 Mark Wielaard <mark@klomp.org>
5114 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5117 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5119 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5120 than "alu", remove explicit "memory" and "imm_disp" attributes.
5121 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5124 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5125 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5126 gen_lowpart_no_emit.
5128 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5131 * config/i386/i386.c (ix86_expand_builtin): Change memory
5132 operand to XI, extend p0 to Pmode.
5133 * config/i386/i386.md: Change unspec volatile and operand
5134 1 mode to XI, change operand 0 mode to P.
5136 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5138 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5139 GET_MODE_MASK before any checking.
5140 (nds32_can_use_bset_p): Likewise.
5141 (nds32_can_use_btgl_p): Likewise.
5143 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5145 * config/nds32/nds32-doubleword.md: New define_split pattern for
5146 illegal register number.
5148 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5150 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5152 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5154 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5156 2018-04-25 Richard Biener <rguenther@suse.de>
5158 * lto-streamer.h (LTO_major_version): Bump to 8.
5160 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5162 * BASE-VER: Set to 9.0.0.
5164 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5166 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5167 in __abskf2 and __powikf2.
5169 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5172 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5173 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5174 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5175 Use the above on operand 2. Reindent.
5176 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5178 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5181 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5183 * config/i386/i386.opt (mcet): Removed.
5184 * doc/install.texi: Remove -mcet documentation.
5185 * doc/invoke.texi: Likewise.
5187 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5190 * doc/install.texi: Remove -mcet from bootstrap-cet.
5192 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5195 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5196 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5200 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5201 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5202 containing a CONST_VECTOR.
5204 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5206 * doc/install.texi: Update newlib dependency for nvptx.
5208 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5211 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5212 instead of INTVAL when shifting x left.
5214 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5216 PR tree-optimization/85478
5217 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5218 vect_grouped_store_supported for single element vectors.
5220 2018-04-24 Richard Biener <rguenther@suse.de>
5223 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5224 load cost increase to the case of non-constant step.
5226 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5229 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5230 destination if any_malformed_asm.
5232 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5235 * expr.c (store_field): In the bitfield case, if the value comes from
5236 a function call and is returned in registers by means of a PARALLEL,
5237 do not change the mode of the temporary unless BLKmode and VOIDmode.
5239 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5241 PR rtl-optimization/85423
5242 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5243 dependencies to debug insns when the previous insn is non-debug.
5245 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5247 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5248 enums into a single definition.
5249 (fls): Fix predicates and printing.
5252 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5254 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5255 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5256 and short u6 immediate.
5257 (check_if_valid_sleep_operand): Remove.
5258 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5260 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5262 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5263 flag_always_save_lp condition.
5264 * config/nds32/nds32.opt (malways-save-lp): New option.
5266 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5268 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5269 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5270 * config/nds32/nds32.h
5271 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5272 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5274 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5276 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5277 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5279 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5280 Chung-Ju Wu <jasonwucj@gmail.com>
5282 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5284 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5285 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5287 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5289 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5291 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5293 * config/nds32/nds32-protos.h (nds32_data_alignment,
5294 nds32_local_alignment): Declare.
5295 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5296 nds32_local_alignment): New functions.
5297 (TARGET_CONSTANT_ALIGNMENT): Define.
5298 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5300 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5302 * config/nds32/nds32.c
5303 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5304 (TARGET_MODES_TIEABLE_P): Likewise.
5306 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5308 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5311 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5312 Chung-Ju Wu <jasonwucj@gmail.com>
5314 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5315 for unaligned access.
5316 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5318 * config/nds32/nds32-intrinsic.md: Likewise.
5319 * config/nds32/nds32_intrinsic.h: Likewise.
5320 * config/nds32/nds32.h (nds32_builtins): Likewise.
5321 * config/nds32/nds32.opt (munaligned-access): New option.
5322 * config/nds32/nds32.c (nds32_asm_file_start): Display
5323 flag_unaligned_access status.
5325 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5327 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5328 -mno-relax is present.
5329 * config/riscv/linux.h (LINK_SPEC): Ditto.
5331 2018-04-20 Martin Sebor <msebor@redhat.com>
5334 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5336 (gimple_fold_builtin_stxcpy_chk): Same.
5337 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5339 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5342 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5343 __powikf2 when long double is IEEE 128-bit.
5345 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5347 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5348 step to make sure stack always aligned.
5350 2018-04-20 Carl Love <cel@us.ibm.com>
5353 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5354 size check for arg0.
5356 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5357 Tom de Vries <tom@codesourcery.com>
5360 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5361 Emit insns for calls too.
5362 (nvptx_find_par): Always look for worker-level predecessor insn.
5363 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5365 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5366 (nvptx_process_pars): Propagate frames for calls.
5368 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5371 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5373 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5374 (ix86_handle_option): Don't handle OPT_mibt.
5375 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5377 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5379 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5380 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5381 (ix86_target_macros): Define __CET__ with flag_cf_protection
5382 for -fcf-protection.
5383 * config/i386/i386.c (isa2_opts): Remove -mibt.
5384 * config/i386/i386.h (TARGET_IBT): Removed.
5385 (TARGET_IBT_P): Likewise.
5386 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5387 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5388 * config/i386/i386.opt (mcet): Update help message.
5391 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5392 -mcet as an alias for -mshstk.
5394 2018-04-20 Richard Biener <rguenther@suse.de>
5397 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5398 complexity by forcing a single use of the multiply operand.
5400 2018-04-20 Martin Jambor <mjambor@suse.cz>
5403 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5404 recursion dependency to only apply to non-clones.
5406 2018-04-20 Martin Jambor <mjambor@suse.cz>
5409 * ipa-cp.c (create_specialized_node): Check that clones of
5410 self-recursive edges exist during IPA-CP.
5412 2018-04-19 Toon Moene <toon@moene.org>
5414 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5417 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5419 PR tree-optimization/85467
5420 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5421 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5422 VECTOR_CST element to type.
5424 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5427 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5428 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5429 (builtin_longjmp): Likewise.
5430 (save_stack_nonlocal): New pattern.
5431 (restore_stack_nonlocal): Likewise.
5433 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5436 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5437 Replace ASM_OUTPUT_LABEL with fprintf.
5439 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5442 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5443 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5444 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5445 define __IBT__ and __SHSTK__ for -fcf-protection.
5446 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5448 (ix86_trampoline_init): Likewise.
5449 (x86_output_mi_thunk): Likewise.
5450 (ix86_notrack_prefixed_insn_p): Likewise.
5451 (ix86_option_override_internal): Don't disallow -fcf-protection.
5452 * config/i386/i386.md (rdssp<mode>): Also enable for
5454 (incssp<mode>): Likewise.
5455 (nop_endbr): Likewise.
5456 * config/i386/i386.opt (mcet): Change help message to built-in
5460 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5461 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5462 enable CET built-in functions.
5464 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5466 * common/config/i386/i386-common.c
5467 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5468 OPTION_MASK_ISA_MOVDIRI_UNSET,
5469 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5470 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5471 * config.gcc (movdirintrin.h): New header.
5472 * config/i386/cpuid.h (bit_MOVDIRI,
5473 bit_MOVDIR64B): New bits.
5474 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5476 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5477 (VOID, PVOID, PCVOID)): New function types.
5478 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5479 __builtin_ia32_directstoreu_u64,
5480 __builtin_ia32_movdir64b): New builtins.
5481 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5482 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5484 (ix86_valid_target_attribute_inner_p): Ditto.
5485 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5486 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5487 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5488 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5489 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5490 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5491 (movdiri<mode>, movdir64b_<mode>): New.
5492 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5493 * config/i386/immintrin.h: Include movdirintrin.h.
5494 * config/i386/movdirintrin.h: New file.
5495 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5497 2018-04-19 Richard Biener <rguenther@suse.de>
5500 * cfg.c (clear_bb_flags): When loop state says we have
5501 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5503 2018-04-19 Richard Biener <rguenther@suse.de>
5505 PR tree-optimization/84737
5506 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5507 copying restrict info.
5508 (vect_setup_realignment): Use it.
5509 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5510 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5511 the first DR to all generated stores.
5512 (vectorizable_load): Likewise for loads.
5514 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5516 PR tree-optimization/85446
5517 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5518 the integral and pointer types to have the same precision.
5520 * doc/install.texi: Document --disable-cet being the default and
5523 2018-04-18 Martin Liska <mliska@suse.cz>
5525 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5528 2018-04-18 Martin Liska <mliska@suse.cz>
5531 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5534 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5535 arguments if they are comparable.
5537 2018-04-18 Martin Liska <mliska@suse.cz>
5540 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5543 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5546 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5549 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5550 ENDBR after calling __morestack.
5552 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5555 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5556 by using gcc_base_ver to generate a gcc_driver_version, and use
5557 it when generating GCC_DRIVER_NAME.
5558 * configure: Regenerate.
5560 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5563 * config.gcc: Obsolete powerpc*-*-*spe*.
5565 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5568 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5569 (stabstr_D): Change type of unum from unsigned int to
5570 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5573 2018-04-17 Jim Wilson <jimw@sifive.com>
5576 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5577 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5578 Set arg_pointer_offset after using pretend_args_size.
5580 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5582 PR rtl-optimization/85431
5583 * dse.c (record_store): Ignore zero width stores.
5586 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5587 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5588 __builtin_stack_restore rather than after it.
5589 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5590 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5591 argument instead of virtual_dynamic_stack_rtx.
5593 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5595 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5597 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5598 Add note to error message to explain internal mapping of overloaded
5599 built-in function name to non-overloaded built-in function name.
5600 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5603 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5606 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5607 where the inputs overlap with the output.
5609 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5611 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5612 (=v, v) alternative and explicit "memory" attribute.
5613 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5614 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5616 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5617 "sselog1" type instead of "sselog".
5618 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5619 "sselog". Remove explicit "memory" attribute.
5620 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5621 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5623 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5624 "sselog1" type instead of "sselog". Remove explicit "memory"
5626 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5627 use "sselog1" type instead of "sselog". Remove explicit "memory"
5629 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5630 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5632 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5633 "sselog1" type instead of "sselog". Remove explicit "memory"
5635 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5636 use "sselog1" type instead of "sselog". Remove explicit "memory"
5640 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5643 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5646 2018-04-17 Martin Jambor <mjambor@suse.cz>
5649 * ipa-cp.c (create_specialized_node): Call
5650 expand_all_artificial_thunks if necessary.
5652 2018-04-17 Martin Liska <mliska@suse.cz>
5655 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5656 in message, remote space in between '_G' and '('.
5658 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5661 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5662 avx512f_vmcmp<mode>3<round_saeonly_name>,
5663 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5664 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5665 avx512f_rndscale<mode><round_saeonly_name>,
5666 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5667 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5668 Use %<iptr>2 instead of %2 for -masm=intel.
5669 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5670 avx512f_vcvttss2usi<round_saeonly_name>,
5671 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5673 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5674 avx512f_vcvttsd2usi<round_saeonly_name>,
5675 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5676 Use %q1 instead of %1 for -masm=intel.
5677 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5678 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5679 of %3 for -masm=intel.
5680 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5682 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5684 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5686 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5687 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5689 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5691 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5692 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5693 %g1 and one with %0 and %1.
5694 (avx512er_vmrcp28<mode><round_saeonly_name>,
5695 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5697 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5698 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5699 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5700 of %0 and %{%4%} for -masm=intel.
5701 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5702 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5703 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5704 order of %0 and %{%5%}%{z%} for -masm=intel.
5706 2018-04-17 Jan Hubicka <jh@suse.cz>
5709 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5711 2018-04-17 Martin Liska <mliska@suse.cz>
5714 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5715 for target_clone error message. Make default implementation
5716 clone to be a local declaration.
5717 (separate_attrs): Add new argument and check for an empty
5719 (expand_target_clones): Handle it.
5720 (ipa_target_clone): Make redirection just for target_clones
5723 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5724 Tom de Vries <tom@codesourcery.com>
5727 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5728 tiled basic blocks without omp continue statements.
5730 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5733 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5734 vec_extract expression as having side effects to make sure it gets
5737 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5740 * config/i386/i386.c (get_builtin_code_for_version): Check
5743 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5746 * gcc/config.gcc: Support "skylake".
5747 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5749 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5750 (processor_target_table): Add "skylake".
5751 (ix86_option_override_internal): Add "skylake".
5752 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5753 PROCESSOR_CANNONLAKE.
5754 (get_builtin_code_for_version): Fix priority for
5755 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5756 PROCESSOR_SKYLAKE-AVX512.
5757 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5758 (processor_type): Add PROCESSOR_SKYLAKE.
5760 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5761 Jason Merrill <jason@redhat.com>
5764 * convert.c (convert_to_integer_1): Use direct recursion for
5765 enumeral types and types with a precision less than the number
5766 of bits in their mode.
5768 2018-04-16 Julia Koval <julia.koval@intel.com>
5771 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5772 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5774 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5777 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5778 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5779 and -mno-direct-move.
5781 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5784 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5785 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5786 Ensure negative shifts result in {0}.
5788 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5790 PR rtl-optimization/79916
5791 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5792 regs (if any) to define how to gnerate SD moves when LRA is in
5795 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5797 PR rtl-optimization/85393
5798 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5799 * except.c (expand_dw2_landing_pad_for_region): Make static.
5800 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5801 a label and unconditional jump to old_bb, rather than
5802 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5805 PR rtl-optimization/85376
5806 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5807 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5808 instead of a specific value.
5810 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5811 Bin Cheng <bin.cheng@arm.com>
5813 PR tree-optimization/82965
5814 PR tree-optimization/83991
5815 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5816 by_profile_only parameter.
5817 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5818 information if the loop was predicted to iterate too many times.
5819 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5821 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5824 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5827 2018-04-13 Martin Liska <mliska@suse.cz>
5828 Jakub Jelinek <jakub@redhat.com>
5831 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5832 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5833 * builtins.c (expand_builtin_memory_copy_args): Use
5834 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5835 handle dest_addr == pc_rtx.
5837 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5840 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5841 asked to not generate direct moves.
5842 (fix_trunc<mode>si2_stfiwx): Similar.
5843 (fix_trunc<mode>si2_internal): Similar.
5845 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5848 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5849 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5850 lookup if dest in some wider mode is known to be const0_rtx and
5851 if so, record permanent equivalence for it to be ZERO_EXTEND of
5852 the narrower mode destination.
5854 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5856 * lto-streamer-out.c (output_function): Revert 259346.
5857 * omp-expand.c (expand_oacc_for): Likewise.
5859 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5861 PR rtl-optimization/85354
5862 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5863 * sel-sched.c (sel_global_init): ... here.
5865 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5868 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5869 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5870 mode for PE-COFF targets.
5871 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5872 (i386_pe_asm_lto_end): Likewise.
5873 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5874 (TARGET_ASM_LTO_END): Likewise.
5875 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5876 (i386_pe_asm_lto_start): New function.
5877 (i386_pe_asm_lto_end): Likewise.
5879 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5880 Richard Biener <rguenther@suse.de>
5883 * lto-streamer-out.c (output_function): Fix CFG loop state before
5885 * omp-expand.c (expand_oacc_for): Handle calls to internal
5886 functions like regular functions.
5888 2018-04-12 Richard Biener <rguenther@suse.de>
5891 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5892 for the early LTO debug to properly generate references to it
5893 during DIE emission. Do not re-use that for the skeleton for
5895 (dwarf2out_early_finish): Likewise.
5897 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5900 * config/i386/sse.md
5901 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5902 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5903 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5904 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5905 and output is a reg, avoid creating invalid lowpart subreg, but
5906 instead split into a 512-bit move. Don't split if not AVX512VL,
5907 input is xmm16+ reg and output is a mem.
5908 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5909 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5910 xmm16+ reg and output is a mem.
5912 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5914 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
5915 also for flag_dwarf2_cfi_asm.
5917 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5919 PR rtl-optimization/85342
5920 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
5921 a bool scalar var inside of the loop instead. Don't try to update
5922 recog_data.operand after failed apply_change_group.
5924 2018-04-12 Tom de Vries <tom@codesourcery.com>
5927 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
5928 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
5929 array with flexible array member as array without given dimension.
5930 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
5931 argument for undefined param to true.
5933 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
5936 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
5937 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
5938 from PowerPC section.
5939 * config/rs6000/sysv4.opt (mcall-): Improve help text.
5940 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
5941 help text that is too long.
5942 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
5943 help text that is too long.
5944 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
5945 help text that is too long.
5947 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
5949 * config/alpha/alpha.md (stack_probe_internal): Rename
5950 from "probe_stack". Update all callers.
5952 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5954 PR rtl-optimization/84566
5955 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
5956 sched_macro_fuse_insns.
5958 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
5961 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
5962 (compute_block_dependences): ... from here.
5964 2018-04-11 Jakub Jelinek <jakub@redhat.com>
5966 PR tree-optimization/85331
5967 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
5968 from int to HOST_WIDE_INT.
5970 2018-04-11 Martin Jambor <mjambor@suse.cz>
5973 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
5974 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
5975 not the same as the source val.
5976 (cgraph_edge_brings_value_p): New parameter.
5977 (gather_edges_for_value): Pass destination value to
5978 cgraph_edge_brings_value_p.
5979 (perhaps_add_new_callers): Likewise.
5980 (get_info_about_necessary_edges): Likewise and exclude values brought
5981 only by self-recursive edges.
5982 (create_specialized_node): Redirect only clones of self-calling edges.
5983 (+self_recursive_pass_through_p): New function.
5984 (find_more_scalar_values_for_callers_subset): Use it.
5985 (find_aggregate_values_for_callers_subset): Likewise.
5986 (known_aggs_to_agg_replacement_list): Removed.
5987 (decide_whether_version_node): Re-calculate known constants for all
5988 remaining context clones.
5990 2018-04-11 Richard Biener <rguenther@suse.de>
5993 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
5994 from early DWARF output.
5995 (dwarf2out_early_finish): Output line info unconditionally into
5996 early DWARF and add reference to it.
5998 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6001 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6002 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6003 other than V2DFmode using iptr mode attribute.
6004 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6006 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6008 PR rtl-optimization/84659
6009 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6011 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6014 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6016 (output_loc_list): Pass address of a dummy size variable even in the
6017 locview handling loop.
6018 (index_location_lists): Add comment on why skip_loc_list_entry can't
6021 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6024 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6027 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6030 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6031 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6032 and -mstring-compare-inline-limit.
6034 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6037 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6038 for stack clash protection in a register whenever we need it to be in
6041 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6043 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6044 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6046 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6049 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6051 (mlong-double-): Ditto.
6052 * config/rs6000/sysv4.opt (msdata=): Ditto.
6053 (mtls-size=): Ditto.
6055 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6057 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6058 erroneous entries for
6059 "vector int vec_ldl (int, long int *)", and
6060 "vector unsigned int vec_ldl (int, unsigned long int *)".
6061 Add comments and entries for
6062 "vector bool char vec_ldl (int, bool char *)",
6063 "vector bool short vec_ldl (int, bool short *)",
6064 "vector bool int vec_ldl (int, bool int *)",
6065 "vector bool long long vec_ldl (int, bool long long *)",
6066 "vector pixel vec_ldl (int, pixel *)",
6067 "vector long long vec_ldl (int, long long *)",
6068 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6069 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6070 type tree bool_long_long_type_node and correct definition of
6071 bool_V2DI_type_node to make reference to this new type tree.
6072 (rs6000_mangle_type): Replace erroneous reference to
6073 bool_long_type_node with bool_long_long_type_node.
6074 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6075 comments to emphasize sign distinctions for char and int types and
6076 replace RS6000_BTI_bool_long constant with
6077 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6078 use of RS6000_BTI_pixel.
6079 (bool_long_type_node): Remove this macro definition.
6080 (bool_long_long_type_node): New macro definition
6082 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6084 PR rtl-optimization/85300
6085 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6086 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6087 simplify_unary_operation fails.
6089 2018-04-10 Martin Liska <mliska@suse.cz>
6091 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6092 cgraph_edge and ipa_ref.
6094 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6098 * config/i386/sse.md
6099 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6100 computation of the VEC_MERGE selector from mask.
6101 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6102 Fix decoding of the VEC_MERGE selector into mask.
6104 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6106 PR tree-optimization/85286
6107 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6109 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6111 * final.c (final_1): Set insn_last_address as well as
6112 insn_current_address.
6114 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6117 * explow.c (emit_stack_probe): Call validize_mem on memory location
6118 before passing it to gen_probe_stack. Create address operand and
6119 legitimize it for the probe_stack_address case.
6121 2018-04-09 Jan Hubicka <jh@suse.cz>
6124 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6125 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6126 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6128 (free_lang_data): Rebuild type inheritance graph.
6130 2018-04-09 Martin Sebor <msebor@redhat.com>
6132 * invoke.texi (-finline-small-functions): Mention other optimization
6134 (-findirect-inlining, -fpartial-inlining): Same.
6135 (-finline-functions-called-once): Same.
6136 (-freorder-blocks-and-partition): Same.
6138 2018-04-09 Jan Hubicka <jh@suse.cz>
6141 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6142 jumps; choose last target that matches the criteria (i.e.
6143 no partition changes for non-crossing jumps).
6144 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6145 support for redirecting crossing jumps to non-crossing.
6147 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6149 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6150 also for naked functions.
6152 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6154 * config/arc/arc.md (add_shift): New pattern.
6155 (add_shift2): Likewise.
6156 (sub_shift): Likewise.
6157 (sub_shift_cmp0_noout): Likewise.
6158 (compare_si_ashiftsi): Likewise.
6159 (xbfu_cmp0_noout): New combine pattern.
6160 (xbfu_cmp0"): Likewise.
6161 (movsi_set_cc_insn): Place the predicable variant first.
6162 (commutative_binary_cmp0_noout): Remove clobber.
6163 (commutative_binary_cmp0): New pattern.
6164 (noncommutative_binary_cmp0): Likewise.
6165 (noncommutative_binary_cmp0_noout): Likewise.
6166 (noncommutative_binary_comparison_result_used): Removed.
6167 (rsub_cmp0): New pattern.
6168 (rsub_cmp0_noout): Likewise.
6169 (extzvsi): Changed, keep only meaningful variants.
6170 (SQH, SEZ): New iterators.
6171 (SQH_postfix): New mode attribute.
6172 (SEZ_prefix): New code attribute.
6173 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6174 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6175 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6177 (noncommutative_operator): Check the availability of barrel
6180 2018-04-09 Richard Biener <rguenther@suse.de>
6182 PR tree-optimization/85284
6183 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6184 Only use the niter constraining form of simple_iv when the exit
6187 2018-04-09 Tom de Vries <tom@codesourcery.com>
6190 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6191 (define_expand "*memory_barrier"): New define_expand.
6192 (define_insn "memory_barrier"): New insn.
6194 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6196 PR rtl-optimization/80463
6197 PR rtl-optimization/83972
6198 PR rtl-optimization/83480
6200 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6201 correct producer for the insn.
6202 (tidy_control_flow): Fixup seqnos in case of debug insns.
6204 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6206 PR rtl-optimization/83913
6208 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6209 different sched-times when merging exprs.
6211 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6213 PR rtl-optimization/83962
6215 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6216 tidy_fallthru_edge and tidy_control_flow.
6218 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6220 PR rtl-optimization/83530
6222 * sel-sched.c (force_next_insn): New global variable.
6223 (remove_insn_for_debug): When force_next_insn is true, also leave only
6224 next insn in the ready list.
6225 (sel_sched_region): When the region wasn't scheduled, make another pass
6226 over it with force_next_insn set to 1.
6228 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6230 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6232 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6233 for interrupt control.
6234 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6235 functions for interrupt control.
6236 * config/nds32/nds32-intrinsic.md: Likewise.
6237 * config/nds32/nds32_intrinsic.h: Likewise.
6238 * config/nds32/nds32.h (nds32_builtins): Likewise.
6240 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6242 * config/nds32/nds32.c (nds32_init_machine_status,
6243 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6244 strict_aligned_p field.
6245 (nds32_expand_to_rtl_hook): New function.
6246 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6247 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6249 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6250 Chung-Ju Wu <jasonwucj@gmail.com>
6252 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6253 * config/nds32/nds32-n7.md: New file.
6254 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6255 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6257 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6258 * config/nds32/nds32.md (pipeline_model): Add n7.
6259 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6260 * config/nds32/pipelines.md: Include n7 settings.
6262 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6263 Chung-Ju Wu <jasonwucj@gmail.com>
6265 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6266 * config/nds32/nds32-e8.md: New file.
6267 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6268 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6270 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6271 * config/nds32/nds32.md (pipeline_model): Add e8.
6272 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6273 * config/nds32/pipelines.md: Include e8 settings.
6275 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6276 Chung-Ju Wu <jasonwucj@gmail.com>
6278 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6279 * config/nds32/nds32-n8.md: New file.
6280 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6281 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6283 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6284 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6285 * config/nds32/nds32.md (pipeline_model): Add n8.
6286 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6287 * config/nds32/pipelines.md: Include n8 settings.
6289 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6290 Chung-Ju Wu <jasonwucj@gmail.com>
6292 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6293 * config/nds32/nds32-n9-2r1w.md: New file.
6294 * config/nds32/nds32-n9-3r2w.md: New file.
6295 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6296 nds32_register_ports): New or modify for cpu n9.
6297 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6299 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6300 * config/nds32/nds32-utils.c: New file.
6301 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6302 TARGET_MUL_SLOW): Define.
6303 * config/nds32/nds32.md (pipeline_model): New attribute.
6304 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6305 New options that support cpu n9.
6306 * config/nds32/pipelines.md: Include n9 settings.
6307 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6309 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6311 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6312 information if necessary.
6313 (output_cond_branch_compare_zero): Likewise.
6314 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6315 (nds32_target_alignment): Refine for alignment.
6316 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6317 (FUNCTION_BOUNDARY): Modify.
6318 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6320 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6322 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6324 * config/nds32/constants.md (unspec_volatile_element): Add values for
6325 TLB operation and data prefetch.
6326 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6327 functions for TLB operation and data prefetch.
6328 * config/nds32/nds32-intrinsic.md: Likewise.
6329 * config/nds32/nds32_intrinsic.h: Likewise.
6330 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6331 (nds32_print_operand): Likewise.
6332 * config/nds32/nds32.h (nds32_builtins): Likewise.
6334 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6335 Andrew Pinski <pinsika@gcc.gnu.org>
6338 * match.pd: Use constant_boolean_node of correct type instead of
6339 boolean_true_node or boolean_false_node for simplifying
6340 pointer comparisons to zero.
6342 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6344 PR tree-optimization/80021
6345 * tree.c (verify_type_variant): Make error call in verify_variant_match
6346 translatable and remove final full stop.
6348 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6350 * config/nds32/constants.md (unspec_volatile_element): Add
6351 UNSPEC_VOLATILE_EH_RETURN.
6352 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6353 nds32_output_stack_pop): Support dwarf exception handling process.
6354 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6355 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6356 exception handling process.
6357 (nds32_compute_stack_frame): Likewise.
6358 (nds32_return_addr_rtx): Likewise.
6359 (nds32_initial_elimination_offset): Likewise.
6360 (nds32_expand_prologue): Likewise.
6361 (nds32_expand_epilogue): Likewise.
6362 (nds32_dynamic_chain_address): New function.
6363 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6365 (DYNAMIC_CHAIN_ADDRESS): Define.
6366 (EH_RETURN_DATA_REGNO): Define.
6367 (EH_RETURN_STACKADJ_RTX): Define.
6368 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6369 patterns for dwarf exception handling.
6371 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6373 * config/nds32/nds32.h: Clean up obsolete macros.
6375 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6377 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6378 Add enum values for particular instructions.
6379 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6380 particular intrinsic functions.
6381 * config/nds32/nds32-intrinsic.md: Likewise.
6382 * config/nds32/nds32_intrinsic.h: Likewise.
6383 * config/nds32/nds32.h (nds32_builtins): Likewise.
6384 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6385 (btst, ave): New patterns for particular instructions.
6387 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6389 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6390 Add enum values for atomic load/store and memory sync.
6391 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6393 * config/nds32/nds32-intrinsic.md: Likewise.
6394 * config/nds32/nds32_intrinsic.h: Likewise.
6395 * config/nds32/nds32.h (nds32_builtins): Likewise.
6397 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6399 PR tree-optimization/85257
6400 * fold-const.c (native_encode_vector): If not all elts could fit
6401 and off is -1, return 0 rather than offset.
6402 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6403 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6404 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6405 adjust buffer in native_interpret_expr call.
6407 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6409 * config/nds32/constants.md (unspec_volatile_element): Add cache
6410 control enum values.
6411 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6412 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6413 * config/nds32/nds32.c (nds32_cctl_names): New.
6414 (nds32_print_operand): Handle cache control register names.
6415 * config/nds32/nds32.h (nds32_builtins): New enum values.
6416 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6418 * config/nds32/nds32.md (type): Add mmu.
6419 * config/nds32/pipelines.md (simple_insn): Add mmu.
6421 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6423 * config/nds32/nds32.md (type): Remove call.
6424 * config/nds32/pipelines.md (simple_insn): Likewise.
6426 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6428 * config/nds32/constants.md (unspec_volatile_element): Add
6429 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6430 UNSPEC_VOLATILE_FMFCFG.
6431 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6432 description for fmfcfg and fmfcsr.
6433 (bdesc_1arg): Add fmtcsr.
6434 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6435 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6436 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6437 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6438 unspec_fmfcfg): New patterns.
6439 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6440 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6441 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6442 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6443 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6444 __nds32__fmfcfg): Define.
6446 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6448 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6449 intrinsic register names.
6450 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6451 intrinsic register enum values and macros.
6453 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6455 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6456 for load/store addressing form.
6457 (nds32_print_operand_address): Likewise.
6459 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6462 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6463 based on LABEL_REF. Remove useless assertion.
6464 (pic_address_needs_scratch): Fix formatting.
6465 (sparc_legitimize_pic_address): Minor tweaks.
6466 (sparc_delegitimize_address): Adjust assertion accordingly.
6467 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6468 into symbolic_operand.
6469 (movsi_high_pic_label_ref): Likewise.
6470 (movsi_lo_sum_pic_label_ref): Likewise.
6471 (movdi_pic_label_ref): Likewise.
6472 (movdi_high_pic_label_ref): Likewise.
6473 (movdi_lo_sum_pic_label_ref): Likewise.
6475 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6477 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6478 custom LIB_SPEC setup.
6480 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6481 Kito Cheng <kito.cheng@gmail.com>
6483 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6484 * config/riscv/freebsd.h: New.
6486 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6488 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6489 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6492 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6493 Kito Cheng <kito.cheng@gmail.com>
6495 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6496 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6497 * config/nds32/nds32-protos.h (nds32_output_call,
6498 nds32_output_return): Declare.
6499 * config/nds32/nds32.md: Refine all the call and return patterns.
6501 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6504 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6505 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6507 PR rtl-optimization/84872
6508 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6509 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6512 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6514 * expr.c (copy_blkmode_to_reg): Revert 254862.
6515 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6517 2018-04-06 Richard Biener <rguenther@suse.de>
6520 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6521 after seeing a component reference with an adjacent field. Treat
6522 refs to arrays at struct end of external decls similar to
6523 refs to unconstrained commons.
6525 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6528 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6529 look through SAVE_EXPRs with non-side-effects argument. Adjust
6531 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6534 2018-04-06 Richard Biener <rguenther@suse.de>
6537 * alias.c (find_base_term): New wrapper around find_base_term
6538 unwinding CSELIB_VAL_PTR changes.
6539 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6542 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6544 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6546 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6547 constant definitions.
6548 ("nop"): lr 0,0 -> nopr r0
6549 ("nop_lr0", "nop_lr1"): New insn definitions.
6551 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6553 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6554 NDS32_V3PUSH_AVAILABLE_P macro.
6556 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6557 Chung-Ju Wu <jasonwucj@gmail.com>
6559 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6560 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6561 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6562 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6563 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6564 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6565 * config/nds32/constraints.md: New constraints and checking for hard
6566 float configuration.
6567 * config/nds32/iterators.md: New mode iterator and attribute for hard
6568 float configuration.
6569 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6571 * config/nds32/nds32-fpu.md: New file.
6572 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6573 deal with hard float code generation.
6574 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6576 (abi_type, float_reg_number): New enum type.
6577 * config/nds32/nds32-predicates.c: New predicates for hard float.
6578 * config/nds32/nds32-protos.h: Declare functions for hard float.
6579 * config/nds32/nds32.c: Implementation for hard float configuration.
6580 * config/nds32/nds32.h: Definitions for hard float configuration.
6581 * config/nds32/nds32.md: Include hard float machine description and
6582 modify patterns for hard float configuration.
6583 * config/nds32/nds32.opt: New options for hard float configuration.
6584 * config/nds32/predicates.md: New predicates for hard float
6587 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6589 * common/config/nds32/nds32-common.c
6590 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6592 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6595 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6596 CONSTRUCTOR_ELT (ctor, ...)->value.
6598 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6601 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6603 2018-04-05 Tom de Vries <tom@codesourcery.com>
6606 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6609 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6610 Kito Cheng <kito.cheng@gmail.com>
6612 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6613 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6614 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6615 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6617 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6618 Kito Cheng <kito.cheng@gmail.com>
6620 * config/nds32/constraints.md (Ufe): New memory constraint.
6621 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6622 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6623 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6625 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6626 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6628 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6630 * config/nds32/nds32.md: Use optimize_size in the condition for
6631 alu-shift instructions.
6633 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6635 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6637 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6639 * config/nds32/nds32.md (negsi2): Refine pattern.
6641 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6642 Chung-Ju Wu <jasonwucj@gmail.com>
6644 * config/nds32/iterators.md (shift_rotate): New code iterator.
6645 (shift): New code attribute.
6646 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6647 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6648 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6649 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6650 bit-wise operations.
6651 (andsi3, *andsi3): Ditto.
6652 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6653 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6654 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6655 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6656 nds32_ior_operand, nds32_xor_operand): New predicates.
6658 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6660 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6661 (addsi3, subsi3): ... this.
6663 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6665 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6667 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6669 * config/nds32/nds32.md: Adjust indention.
6671 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6673 * config/nds32/nds32.md (feature): New attribute.
6675 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6677 * config/nds32/nds32.md (subtype): New attribute.
6679 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6682 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6683 expansion to perform a bitwise AND of the argument followed by a
6684 boolean negation of the result.
6686 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6688 PR rtl-optimization/84878
6689 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6690 the basic block. Assert the use reference is not artificial and that
6691 it has an associated insn.
6693 2018-04-04 Michael Matz <matz@suse.de>
6695 * builtins.c (compute_objsize): Pass correct operand
6696 to array_at_struct_end_p.
6698 2018-04-04 Richard Biener <rguenther@suse.de>
6701 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6702 from contexts for DINFO_LEVEL_TERSE and below.
6704 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6706 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6707 resiter_operand condition.
6708 * config/nds32/nds32.md (*move<mode>): Ditto.
6710 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6711 Monk Chiang <sh.chiang04@gmail.com>
6713 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6715 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6717 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6719 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6720 Kito Cheng <kito.cheng@gmail.com>
6722 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6723 nds32_cond_code_str, output_cond_branch,
6724 output_cond_branch_compare_zero, nds32_expand_cbranch,
6725 nds32_expand_cstore, nds32_expand_movcc,
6726 nds32_output_cbranchsi4_equality_zero,
6727 nds32_output_cbranchsi4_equality_reg,
6728 nds32_output_cbranchsi4_equality_reg_or_const_int,
6729 nds32_output_cbranchsi4_greater_less_zero: New functions.
6730 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6731 nds32_expand_cstore, nds32_expand_movcc,
6732 nds32_output_cbranchsi4_equality_zero,
6733 nds32_output_cbranchsi4_equality_reg,
6734 nds32_output_cbranchsi4_equality_reg_or_const_int,
6735 nds32_output_cbranchsi4_greater_less_zero): Declare.
6736 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6737 nds32_rimm11s_operand): New predicates.
6738 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6739 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6742 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6744 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6745 * config/nds32/nds32.md: Ditto.
6746 * config/nds32/pipelines.md: Ditto.
6748 2018-04-04 Richard Biener <rguenther@suse.de>
6750 PR tree-optimization/85168
6751 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6752 propagating abnormals.
6754 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6756 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6758 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6759 Kito Cheng <kito.cheng@gmail.com>
6761 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6762 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6763 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6764 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6765 * config/nds32/nds32.md (sibcall_internal): New.
6766 (sibcall_register): Remove.
6767 (sibcall_immediate): Remove.
6768 (sibcall_value_internal): New.
6769 (sibcall_value_register): Remove.
6770 (sibcall_value_immediate): Remove.
6771 * config/nds32/predicates.md (nds32_general_register_operand): New.
6772 (nds32_call_address_operand): New.
6774 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6776 PR rtl-optimization/85167
6777 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6778 bb_defs if *split_p, instead preinitialize it to NULL.
6780 PR tree-optimization/85156
6781 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6782 evaluating the argument multiple times.
6784 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6786 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6788 (_mm_cvtpd_ps): Likewise.
6789 (_mm_cvttpd_epi32): Likewise.
6790 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6791 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6792 vector, pixel, and bool following altivec.h include.
6794 2018-04-03 Martin Sebor <msebor@redhat.com>
6796 * doc/extend.texi (Common Function Attributes): Clarify.
6797 (const attribute): Likewise.
6798 (pure attribute): Likewise.
6800 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6803 * config/i386/i386.c (ix86_expand_vector_set): Use
6804 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6806 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6808 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6809 instructions when changing rounding bits to preserve precision bits
6810 in the x87 control word.
6812 2018-04-03 Martin Liska <mliska@suse.cz>
6814 PR tree-optimization/82491
6815 * rtl.h (strip_offset_and_add): Replace += suboffset with
6816 poly_uint64 () + suboffset.
6818 2018-03-29 Martin Liska <mliska@suse.cz>
6819 Martin Jambor <mjambor@suse.cz>
6822 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6823 param_type is not an integral or pointer type.
6825 2018-04-03 Richard Biener <rguenther@suse.de>
6827 * sese.h (recompute_all_dominators): Remove.
6829 2018-04-02 Martin Sebor <msebor@redhat.com>
6831 * doc/invoke.texi (-Wrestrict): Fix typos.
6833 2018-04-02 Jim Wilson <jimw@sifive.com>
6835 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6836 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6837 (<optab>di3, <optab>si3_extend): Likewise.
6838 (<optab>si3_mask, <optab>si3_mask_1): New.
6839 (<optab>di3_mask, <optab>di3_mask_1): New.
6840 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6841 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6842 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6844 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6846 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6849 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6851 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6852 (nds32_canonicalize_comparison): New function.
6854 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6855 Kito Cheng <kito.cheng@gmail.com>
6856 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6858 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6859 * config/nds32/constants.md (unspec_volatile_element): Add
6860 UNSPEC_VOLATILE_RELAX_GROUP.
6861 * config/nds32/nds32-relax-opt.c: New file.
6862 * config/nds32/nds32-predicates.c
6863 (nds32_symbol_load_store_p): New function.
6864 * config/nds32/nds32-protos.h
6865 (nds32_symbol_load_store_p): Declare function.
6866 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6867 * config/nds32/nds32.c
6868 (nds32_register_pass): New function to register pass.
6869 (nds32_register_passes): New function to register passes.
6870 * config/nds32/nds32.md (relax_group): New pattern.
6871 * config/nds32/nds32.opt (mrelax-hint): New option.
6872 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6874 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6876 * config/nds32/t-nds32: Modify files dependency.
6878 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6880 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6881 (PROFILE_HOOK): Define its implementation.
6883 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6885 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6886 type and 32-bit size.
6888 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6891 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6892 (V_128_256): New mode iterator.
6893 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6894 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6895 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6897 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6898 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6900 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6903 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6904 NaN inputs correctly.
6906 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6909 * config/rs6000/vsx.md (??r): New mode attribute.
6910 (*vsx_mov<mode>_64bit): Use it.
6911 (*vsx_mov<mode>_32bit): Likewise.
6913 2018-03-30 Martin Sebor <msebor@redhat.com>
6915 PR tree-optimization/84818
6916 * builtins.c (check_access): Use warning_n.
6918 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6921 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
6923 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
6926 2018-03-30 Julia Koval <julia.koval@intel.com>
6929 * x86-tune.def (movx, partial_reg_dependency): Enable for
6932 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
6935 * lra-constraints.c (process_alt_operands): Move setting
6936 this_alternative_matches below.
6938 2018-03-29 Martin Liska <mliska@suse.cz>
6941 * doc/invoke.texi: Document how LTO works with debug info.
6942 Describe auto-load support of binutils. Mention 'x86-64'
6943 as valid option value of -march option.
6945 2018-03-29 Jakub Jelinek <jakub@redhat.com>
6947 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
6950 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
6951 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
6952 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
6955 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6958 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
6959 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
6960 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
6961 for RS6000_BTM_POWERPC64.
6962 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
6963 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
6964 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
6969 2018-03-28 Carl Love <cel@us.ibm.com>
6972 2017-09-27 Carl Love <cel@us.ibm.com>
6974 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
6975 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
6976 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
6979 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6981 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
6982 instead of __vector bool.
6983 (_mm_max_pu8): Likewise.
6984 (_mm_min_pi16): Likewise.
6986 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
6989 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
6990 (DIVWEUO): Likewise.
6992 (DIVDEUO): Likewise.
6993 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
6994 DIVWEUO and DIVDEUO.
6995 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
6996 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
6997 (div_extend): Likewise.
6998 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7000 (__builtin_divweuo): Likewise.
7001 (__builtin_divdeo): Likewise.
7002 (__builtin_divdeuo): Likewise.
7004 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7007 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7008 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7010 PR tree-optimization/82004
7011 * gimple-match-head.c (optimize_pow_to_exp): New function.
7012 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7013 Don't fold to exp if optimize_pow_to_exp is false.
7015 2018-03-28 Martin Liska <mliska@suse.cz>
7018 * calls.c (initialize_argument_information): Fix trailing space.
7019 * common.opt: Fix typo and provide better explanation for
7020 -fsanitize-coverage option.
7021 * config/i386/i386.opt: Fix typo.
7023 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7024 Martin Liska <mliska@suse.cz>
7027 * gimplify.c (asan_poison_variable): Don't do the check for
7028 gimplify_omp_ctxp here.
7029 (gimplify_decl_expr): Do it here.
7030 (gimplify_target_expr): Likewise.
7032 2018-03-28 Martin Liska <mliska@suse.cz>
7035 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7036 chkp_type_bounds_count if MPX is not enabled.
7038 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7040 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7042 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7045 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7046 function to create the function decl for complex long double
7047 multiply and divide for -mabi=ieeelongdouble.
7048 (init_float128_ieee): Call it.
7050 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7053 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7054 -fcf-protection=branch -mibt.
7055 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7057 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7060 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7061 arm_word_relocations.
7063 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7066 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7067 extern array declarations.
7069 2018-03-27 Richard Biener <rguenther@suse.de>
7072 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7073 explicit single_use checks.
7075 2018-03-27 Richard Biener <rguenther@suse.de>
7077 PR tree-optimization/85082
7078 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7081 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7083 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7084 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7085 Turn on fasynchronous-unwind-tables and funwind-tables.
7087 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7090 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7091 (*bmi_blsr_<mode>_ccz): Ditto.
7093 2018-03-26 Tom de Vries <tom@codesourcery.com>
7095 PR tree-optimization/85063
7096 * omp-general.c (offloading_function_p): New function. Factor out
7098 * omp-offload.c (pass_omp_target_link::gate): ... here.
7099 * omp-general.h (offloading_function_p): Declare.
7100 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7101 with attribute omp declare target for offloading functions.
7103 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7105 PR tree-optimization/84005
7106 * tree-data-ref.h (get_base_for_alignment): Declare.
7107 * tree-data-ref.c (get_base_for_alignment_1): New function.
7108 (get_base_for_alignment): Likewise.
7109 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7110 get_base_for_alignment to find a suitable base object, instead
7111 of always using drb->base_address.
7113 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7116 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7117 known size by default.
7119 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7122 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7123 and non BLKmode operands.
7125 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7128 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7129 Clean up attributes.
7131 2018-03-23 Richard Biener <rguenther@suse.de>
7134 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7135 we are going to emit early debug for LTO.
7137 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7140 * function.c (match_asm_constraints_1): Don't optimize if input
7141 doesn't satisfy general_operand predicate for output's mode.
7144 * alias.c (write_dependence_p): Don't require for x_canonicalized
7145 non-VOIDmode if x has VOIDmode.
7148 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7149 just don't try to optimize it rather than assert it never happens.
7151 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7153 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7154 macro expansions for definition of ST_INTERNAL_<mode> and
7155 LD_INTERNAL_<mode> builtins.
7156 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7158 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7160 (altivec_expand_st_builtin): Likewise.
7161 (altivec_expand_builtin): Remove calls to deleted functions.
7162 (rs6000_address_for_altivec): Delete this function.
7163 * config/rs6000/vector.md: Remove expands for
7164 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7166 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7169 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7170 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7171 re-computing once computed.
7172 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7173 (arm_init_machine_status): Initialize
7174 machine->static_chain_stack_bytes.
7176 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7179 * doc/extend.texi: Add four new prototypes for vec_ld.
7180 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7181 definitions for more logical presentation.
7182 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7183 entries for V1TI variants of __builtin_altivec_ld builtin.
7184 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7185 handling of V1TI variant of LVX icode pattern.
7186 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7187 (rs6000_gimple_fold_builtin): Likewise.
7188 (altivec_init_builtins): Add code to define
7189 __builtin_altivec_lvx_v1ti function.
7191 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7194 * function.c (match_asm_constraints_1): Don't do the optimization
7195 if input isn't a REG, SUBREG, MEM or constant.
7197 2018-03-22 Tom de Vries <tom@codesourcery.com>
7199 PR tree-optimization/84956
7200 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7201 bb_has_abnormal_pred.
7203 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7206 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7207 DECL_INITIAL (decl) to decl at the end.
7208 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7211 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7213 * doc/extend.texi (__builtin_tgmath): Document when complex
7214 integer types are treated as _Complex _Float64.
7216 2018-03-21 Tom de Vries <tom@codesourcery.com>
7218 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7220 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7222 PR tree-optimization/84960
7223 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7224 if it is ENTRY block, move them into single succ of ENTRY in that case.
7226 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7228 PR tree-optimization/84811
7229 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7230 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7231 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7232 (known_subrange_p): Update accordingly. Cast each value involved
7233 in the size comparison, rather than casting the result of the
7236 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7238 PR tree-optimization/84982
7239 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7240 by flipping the least significant bit rather than all bits from
7241 bitpos to bitpos + bitsize - 1.
7243 2018-03-21 Nathan Sidwell <nathan@acm.org>
7245 * doc/extend.texi (Deprecated Features): Remove mention of
7246 long-deleted deprecations.
7248 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7251 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7252 * configure: Regenerate.
7254 2018-03-21 Tom de Vries <tom@codesourcery.com>
7256 PR tree-optimization/83126
7257 * tree-parloops.c (num_phis): New function.
7258 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7260 2018-03-21 Nathan Sidwell <nathan@acm.org>
7262 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7263 mention anon-struct/union members and trailing attributes.
7265 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7267 PR tree-optimization/84969
7268 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7269 builtin memset partitions if they set different rhs values.
7271 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7273 PR rtl-optimization/84989
7274 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7275 VEC_DUPLICATE with scalar result mode.
7277 2018-03-21 Martin Liska <mliska@suse.cz>
7280 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7281 not intended return statement.
7283 2018-03-21 Martin Liska <mliska@suse.cz>
7286 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7287 (chkp_find_bound_slots_1): Limit number of iterations.
7289 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7292 * Minor grammar fixes for x86 options.
7294 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7297 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7298 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7300 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7303 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7304 (altivec_lvx_<mode>_1op): Likewise.
7305 (altivec_stvx_<mode>_2op): Likewise.
7306 (altivec_stvx_<mode>_1op): Likewise.
7307 (altivec_lvx_<VM2:mode>): New define_expand.
7308 (altivec_stvx_<VM2:mode>): Likewise.
7309 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7310 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7311 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7312 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7313 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7314 (rs6000_gen_lvx): Likewise.
7315 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7316 (altivec_expand_stv_builtin): Likewise.
7317 (altivec_expand_builtin): Likewise.
7318 * config/rs6000/vector.md: Likewise.
7320 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7323 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7326 2018-03-20 Richard Biener <rguenther@suse.de>
7329 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7330 sign-conversions as zero, fall back to standard scalar_stmt
7333 2018-03-20 Martin Liska <mliska@suse.cz>
7336 * predict.c (rebuild_frequencies): Handle case when we have
7337 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7339 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7342 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7343 flag_section_anchors.
7344 * varasm.c (use_blocks_for_decl_p): Remove hack for
7345 dw2_force_const_mem.
7348 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7350 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7351 be created, use lowpart_subreg of operands[0] rather than operands[0]
7353 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7354 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7355 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7356 and n constraint instead of aarch64_shift_imm_di and Usd.
7357 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7358 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7360 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7363 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7364 to favor GPR over NEON registers.
7365 (<shift>di3_neon): Likewise.
7367 2018-03-20 Tom de Vries <tom@codesourcery.com>
7370 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7371 (nvptx_process_pars): Emit bar.sync asap and alap.
7373 2018-03-20 Tom de Vries <tom@codesourcery.com>
7376 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7377 seen_label if seen_label is already set.
7379 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7382 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7383 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7384 Use 1U instead of 1. Formatting fixes.
7387 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7388 instead of TREE_TYPE (s1) for the return value.
7390 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7392 PR tree-optimization/84946
7393 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7394 bitsize + bitsize in poly_uint64 rather than poly_int64.
7397 * dwarf2asm.c: Include fold-const.c.
7398 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7399 of decl rather than decl itself.
7401 PR rtl-optimization/84643
7402 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7404 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7407 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7408 calling assemble_variable.
7410 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7413 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7414 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7416 2018-03-19 Jim Wilson <jimw@sifive.com>
7419 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7420 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7421 (riscv_first_stack_step): Likewise.
7422 (riscv_option_override): Use STACK_BOUNDARY instead of
7424 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7426 (BIGGEST_ALIGNMENT): Set to 128.
7427 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7428 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7431 2018-03-19 Richard Biener <rguenther@suse.de>
7433 PR tree-optimization/84933
7434 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7435 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7437 2018-03-19 Richard Biener <rguenther@suse.de>
7439 PR tree-optimization/84859
7440 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7441 (cond_if_else_store_replacement): Perform sinking operation on
7442 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7443 Generalize what a BB with a single eligible store is.
7445 2018-03-19 Richard Biener <rguenther@suse.de>
7447 PR tree-optimization/84929
7448 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7449 chrec_is_positive against non-chrec arg.
7451 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7454 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7456 2018-03-18 Martin Liska <mliska@suse.cz>
7458 PR rtl-optimization/84635
7459 * regrename.c (build_def_use): Use matches_mode only when
7462 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7464 PR tree-optimization/84913
7465 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7466 vectorize chains of COND_EXPRs.
7468 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7470 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7472 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7474 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7476 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7478 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7480 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7481 Kito Cheng <kito.cheng@gmail.com>
7483 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7484 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7485 (nds32_adjust_reg_alloc_order): New function.
7486 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7488 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7490 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7491 nds32_print_operand, nds32_print_operand_address): Use
7492 HOST_WIDE_INT_PRINT_DEC instead.
7494 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7496 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7498 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7501 * config/i386/i386.c (initial_ix86_tune_features,
7502 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7504 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7505 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7506 rather than 1u << ix86_tune. Formatting fix.
7507 (ix86_option_override_internal): Change ix86_arch_mask from
7508 unsigned int to unsigned HOST_WIDE_INT, initialize to
7509 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7510 (ix86_function_specific_restore): Likewise.
7512 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7515 * postreload.c (reload_combine_recognize_pattern): Perform
7516 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7517 truncate_int_for_mode the result for the destination's mode.
7520 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7521 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7523 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7526 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7529 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7530 regno_allocno_class_array and sorted_pseudos.
7531 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7532 insns where regno is used.
7534 2018-03-16 Martin Liska <mliska@suse.cz>
7537 * multiple_target.c (create_dispatcher_calls): Redirect
7538 reference in the symbol table.
7540 2018-03-16 Martin Liska <mliska@suse.cz>
7543 * multiple_target.c (create_dispatcher_calls): Redirect also
7546 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7550 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7551 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7552 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7554 2018-03-16 Julia Koval <julia.koval@intel.com>
7556 * doc/invoke.texi (Skylake Server): Add CLWB.
7557 Cannonlake): Remove CLWB.
7559 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7561 PR tree-optimization/84841
7562 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7564 (FLOAT_ONE_CONST_TYPE): Define.
7565 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7566 (sort_by_operand_rank): Put entries with higher constant_type last
7567 rather than first to match comments.
7569 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7571 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7574 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7577 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7578 check and use address of target always.
7580 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7583 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7584 (indirect_thunk_bnd_needed): Likewise.
7585 (indirect_thunks_used): Likewise.
7586 (indirect_thunks_bnd_used): Likewise.
7587 (indirect_return_needed): New.
7588 (indirect_return_bnd_needed): Likewise.
7589 (output_indirect_thunk_function): Add a bool argument for
7591 (output_indirect_thunk_function): Don't generate alias for
7592 function return thunk.
7593 (ix86_code_end): Call output_indirect_thunk_function to generate
7594 function return thunks.
7595 (ix86_output_function_return): Set indirect_return_bnd_needed
7596 and indirect_return_needed instead of indirect_thunk_bnd_needed
7597 and indirect_thunk_needed.
7599 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7601 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7602 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7603 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7605 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7606 Paul Hua <paul.hua.gm@gmail.com>
7609 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7611 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7613 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7614 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7617 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7620 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7621 instead of GET_MODE_SIZE when comparing Units.
7623 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7626 * varasm.c (hash_section): Return an unchangeble hash value
7627 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7628 Return !aarch64_can_use_per_function_literal_pools_p ().
7630 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7633 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7634 rather than address of cmode as last argument to prepare_cmp_insn.
7636 2018-03-15 Julia Koval <julia.koval@intel.com>
7638 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7639 F_AVX512VNNI, F_AVX512BITALG): New.
7641 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7644 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7645 insn for floating-point loads and stores.
7647 2018-03-14 Carl Love <cel@us.ibm.com>
7649 * config/rs6000/rs6000-c.c: Add macro definitions for
7650 ALTIVEC_BUILTIN_VEC_PERMXOR.
7651 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7652 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7653 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7654 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7656 * config/doc/extend.texi: Add prototypes for vec_permxor.
7658 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7661 * diagnostic-show-locus.c (class layout_point): Convert m_line
7662 from int to linenum_type.
7663 (line_span::comparator): Use linenum "compare" function when
7664 comparing line numbers.
7665 (test_line_span): New function.
7666 (layout_range::contains_point): Convert param "row" from int to
7668 (layout_range::intersects_line_p): Likewise.
7669 (layout::will_show_line_p): Likewise.
7670 (layout::print_source_line): Likewise.
7671 (layout::should_print_annotation_line_p): Likewise.
7672 (layout::print_annotation_line): Likewise.
7673 (layout::print_leading_fixits): Likewise.
7674 (layout::annotation_line_showed_range_p): Likewise.
7675 (struct line_corrections): Likewise for field m_row.
7676 (line_corrections::line_corrections): Likewise for param "row".
7677 (layout::print_trailing_fixits): Likewise.
7678 (layout::get_state_at_point): Likewise.
7679 (layout::get_x_bound_for_row): Likewise.
7680 (layout::print_line): Likewise.
7681 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7682 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7683 * input.c (selftest::test_linenum_comparisons): New function.
7684 (selftest::input_c_tests): Call it.
7685 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7686 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7687 * selftest.h (ASSERT_GT): New macro.
7688 (ASSERT_GT_AT): New macro.
7689 (ASSERT_LT): New macro.
7690 (ASSERT_LT_AT): New macro.
7692 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7694 PR rtl-optimization/84780
7695 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7697 2018-03-14 Martin Liska <mliska@suse.cz>
7699 * tree.c (record_node_allocation_statistics): Use
7700 get_stats_node_kind.
7701 (get_stats_node_kind): New function extracted from
7702 record_node_allocation_statistics.
7703 (free_node): Use get_stats_node_kind.
7705 2018-03-14 Richard Biener <rguenther@suse.de>
7707 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7708 that the value-set of ANTIC_IN doesn't grow.
7711 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7713 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7714 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7716 2018-03-14 Julia Koval <julia.koval@intel.com>
7718 * config.gcc (icelake-client, icelake-server): New.
7720 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7721 (initial_ix86_arch_features): Ditto.
7722 (PTA_SKYLAKE): Add SGX.
7723 (PTA_ICELAKE): Remove.
7724 (PTA_ICELAKE_CLIENT): New.
7725 (PTA_ICELAKE_SERVER): New.
7726 (ix86_option_override_internal): Split up icelake on icelake client and
7728 (get_builtin_code_for_version): Ditto.
7729 (fold_builtin_cpu): Ditto.
7730 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7731 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7732 * config/i386/i386.h (processor_type): Ditto.
7733 * doc/invoke.texi: Ditto.
7735 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7738 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7739 INTEGER_CST offset, add it together with bitpos / 8 and
7740 sign extend based on POINTER_SIZE.
7744 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7747 * config/i386/constraints.md (Yc): New register constraint.
7748 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7749 Use Yc constraint for alternative 2 of operand 0. Remove
7750 preferred_for_speed attribute.
7752 2018-03-14 Richard Biener <rguenther@suse.de>
7754 PR tree-optimization/84830
7755 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7756 with the old one to avoid oscillations.
7758 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7761 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7763 (assign_by_spills): Return a flag of reload assignment failure.
7764 Do not process the reload assignment failures. Do not spill other
7765 reload pseudos if they has the same reg class. Update n if
7767 (lra_assign): Add a return arg. Set up from the result of
7768 assign_by_spills call.
7769 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7770 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7771 usage_insns if it is not NULL.
7772 (spill_hard_reg_in_range): New function.
7773 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7774 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7775 function prototypes.
7776 (lra_assign): Change prototype.
7777 * lra.c (lra): Add code to deal with fails by splitting hard reg
7780 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7782 * config/riscv/riscv.opt (mrelax): New option.
7783 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7784 "norelax" when riscv_mrelax is disabled.
7785 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7787 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7790 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7791 reassociation for int modes.
7793 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7795 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7796 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7798 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7799 * config/aarch64/aarch64-sve.md
7800 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7801 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7802 (*extend<mode><Vwide>2): Rename to...
7803 (aarch64_sve_extend<mode><Vwide>2): ...this.
7804 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7805 renaming the old pattern to...
7806 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7808 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7809 define_expand, renaming the old pattern to...
7810 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7811 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7812 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7813 account when deciding which SVE instruction the optab should use.
7814 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7816 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7818 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7819 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7820 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7821 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7822 (tlsdesc_small_<mode>): Turn a define_expand and use
7823 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7824 (tlsdesc_small_advsimd_<mode>): ...this.
7825 (tlsdesc_small_sve_<mode>): New pattern.
7827 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7829 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7830 (UNSPEC_UMUL_HIGHPART): New constants.
7831 (MUL_HIGHPART): New int iteraor.
7832 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7833 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7835 (*<su>mul<mode>3_highpart): New define_insn.
7837 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7840 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7843 2018-03-13 Martin Liska <mliska@suse.cz>
7846 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7848 (sem_item_optimizer::~sem_item_optimizer): Release it.
7849 (sem_item_optimizer::merge_classes): Register variable aliases.
7850 (sem_item_optimizer::fixup_pt_set): New function.
7851 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7852 * ipa-icf.h: Declare new variables and functions.
7854 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7857 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7858 integer_pow2p@2 and test integer_pow2p in condition.
7859 (A < 0 ? C : 0): Similarly for @1.
7862 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7863 characters starting at p contain '\0' character, don't look beyond
7867 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7868 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7871 * reg-stack.c (change_stack): Change update_end var from int to
7872 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7873 also call set_block_for_insn on the newly added insns and rescan.
7876 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7877 on the last operand.
7880 * tree.c (stabilize_reference_1): Return save_expr (e) for
7881 STATEMENT_LIST even if it doesn't have side-effects.
7883 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7885 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7887 2018-03-12 Renlin Li <renlin.li@arm.com>
7889 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7890 aarch64_output_scalar_simd_mov_immediate.
7892 2018-03-12 Martin Sebor <msebor@redhat.com>
7894 PR tree-optimization/83456
7895 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7896 for perfectly overlapping calls to memcpy.
7897 (gimple_fold_builtin_memory_chk): Same.
7898 (gimple_fold_builtin_strcpy): Handle no-warning.
7899 (gimple_fold_builtin_stxcpy_chk): Same.
7900 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7902 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7904 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7905 parameter. Use it for SFmode.
7906 (rs6000_function_arg_advance_1): Adjust.
7907 (rs6000_function_arg): Adjust.
7908 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7910 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7912 PR rtl-optimization/84169
7913 PR rtl-optimization/84780
7914 * combine.c (can_combine_p): Check for a 2-insn combination whether
7915 the destination register is used between the two insns, too.
7917 2018-03-12 Richard Biener <rguenther@suse.de>
7919 PR tree-optimization/84803
7920 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
7921 for refs DR analysis didn't process.
7923 2018-03-12 Richard Biener <rguenther@suse.de>
7925 PR tree-optimization/84777
7926 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
7927 force-vectorize loops ignore whether we are optimizing for size.
7929 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
7931 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
7932 (TARGET_MD_ASM_ADJUST): Define.
7934 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
7935 Kito Cheng <kito.cheng@gmail.com>
7936 Chung-Ju Wu <jasonwucj@gmail.com>
7938 * config/nds32/nds32.c (nds32_compute_stack_frame,
7939 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
7940 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
7941 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
7942 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
7943 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
7944 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
7945 * config/nds32/nds32.md (prologue, epilogue): Use macro
7946 NDS32_V3PUSH_AVAILABLE_P to do checking.
7948 2018-03-11 Jakub Jelinek <jakub@redhat.com>
7951 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
7952 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
7953 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
7954 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
7955 addition of most attributes on !orig_type_die or the attribute not
7956 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
7958 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7959 Chung-Ju Wu <jasonwucj@gmail.com>
7961 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
7963 * config/nds32/nds32.opt (mvh): New option.
7965 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7966 Chung-Ju Wu <jasonwucj@gmail.com>
7968 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
7970 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
7971 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
7974 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
7975 Chung-Ju Wu <jasonwucj@gmail.com>
7977 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
7979 * config/nds32/nds32-multiple.md (strlensi): New pattern.
7980 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
7982 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
7983 Kito Cheng <kito.cheng@gmail.com>
7984 Chung-Ju Wu <jasonwucj@gmail.com>
7986 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
7987 UNSPEC_FFMISM and UNSPEC_FLMISM.
7988 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
7989 for ffb, ffmism and flmism.
7990 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
7991 (unspec_ffmism): Ditto.
7992 (unspec_flmism): Ditto.
7993 (nds32_expand_builtin_impl): Check if string extension is available.
7994 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
7995 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
7997 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8000 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8003 * lra-assigns.c (assign_by_spills): Return a flag of reload
8004 assignment failure. Do not process the reload assignment
8005 failures. Do not spill other reload pseudos if they has the same
8007 (lra_assign): Add a return arg. Set up from the result of
8008 assign_by_spills call.
8009 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8010 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8011 usage_insns if it is not NULL.
8012 (spill_hard_reg_in_range): New function.
8013 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8014 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8015 function prototypes.
8016 (lra_assign): Change prototype.
8017 * lra.c (lra): Add code to deal with fails by splitting hard reg
8020 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8023 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8025 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8028 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8029 (dw_val_node): Add val_symbolic_view.
8030 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8031 (symview_upper_bound): New.
8032 (new_line_info_table): Initialize symviews_since_reset.
8033 (dwarf2out_source_line): Count symviews_since_reset and set
8034 symview_upper_bound.
8035 (dw_val_equal_p): Handle symview.
8036 (add_AT_symview): New.
8037 (print_dw_val): Handle symview.
8038 (attr_checksum, attr_checksum_ordered): Likewise.
8039 (same_dw_val_p, size_of_die): Likewise.
8040 (value_format, output_die): Likewise.
8041 (add_high_low_attributes): Use add_AT_symview for entry_view.
8042 (dwarf2out_finish): Reset symview_upper_bound, clear
8045 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8048 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8049 Add strict argument and use it.
8050 (rs6000_split_multireg_move): Update for new strict argument.
8051 (mem_operand_gpr): Disallow all non-offsettable addresses.
8052 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8054 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8057 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8058 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8059 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8062 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8063 decl, use remap_type if we want to use the type.
8065 2018-03-09 Martin Sebor <msebor@redhat.com>
8067 PR tree-optimization/84526
8068 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8070 (builtin_access::generic_overlap): Be prepared to handle non-array
8073 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8075 PR rtl-optimization/84682
8076 * lra-constraints.c (process_address_1): Check is_address flag
8077 for address constraints.
8078 (process_alt_operands): Likewise.
8079 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8080 preprocess_constraints.
8081 * recog.h (preprocess_constraints): Add oploc parameter.
8083 * recog.c (preprocess_constraints): Test address_operand for
8084 CT_ADDRESS constraints.
8086 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8089 * lra-assigns.c (assign_by_spills): Return a flag of reload
8090 assignment failure. Do not process the reload assignment
8091 failures. Do not spill other reload pseudos if they has the same
8093 (lra_assign): Add a return arg. Set up from the result of
8094 assign_by_spills call.
8095 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8096 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8097 usage_insns if it is not NULL.
8098 (spill_hard_reg_in_range): New function.
8099 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8100 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8101 function prototypes.
8102 (lra_assign): Change prototype.
8103 * lra.c (lra): Add code to deal with fails by splitting hard reg
8106 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8109 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8110 Accept complain bool parameter. Only emit errors if it is true.
8111 (arm_parse_cpu_option_name): Likewise.
8112 (arm_target_thumb_only): Adjust callers of the above.
8113 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8114 prototype to take a default true bool parameter.
8115 (arm_parse_arch_option_name): Likewise.
8117 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8118 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8122 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8123 * configure: Regenerate.
8124 * configure.ac ("linker --version-script option"): New.
8125 ("linker soname option"): New.
8127 2018-03-09 Richard Biener <rguenther@suse.de>
8129 PR tree-optimization/84775
8130 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8131 immediate uses of predicate stmts and mark them modified.
8134 PR tree-optimization/84178
8135 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8137 (version_loop_for_if_conversion): Delay update_ssa call.
8138 (tree_if_conversion): Delay update_ssa until after predicate
8141 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8144 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8145 when the function accesses prior frames.
8147 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8150 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8151 gen_llsym, otherwise call maybe_gen_llsym.
8154 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8155 has ',' character inside of it.
8157 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8160 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8161 as clobbering CC_REGNUM.
8163 2018-03-08 Richard Biener <rguenther@suse.de>
8166 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8167 (follow_copies_to_constant): Do not follow SSA names registered
8170 2018-03-08 Richard Biener <rguenther@suse.de>
8172 PR tree-optimization/84178
8173 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8175 (version_loop_for_if_conversion): Delay update_ssa call.
8176 (tree_if_conversion): Delay update_ssa until after predicate
8179 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8181 PR tree-optimization/84178
8182 * tree-if-conv.c (release_bb_predicate): Remove the
8183 the assertion that the stmts have NULL use_ops.
8184 Discard the statements, asserting that they haven't
8185 yet been added to a BB.
8187 2018-03-08 Richard Biener <rguenther@suse.de>
8189 PR tree-optimization/84746
8190 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8191 (phi_translate): Pass in destination ANTIC_OUT set.
8192 (phi_translate_1): Likewise. For a simplified result lookup
8193 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8194 (phi_translate_set): Adjust.
8195 (do_pre_regular_insertion): Likewise.
8196 (do_pre_partial_partial_insertion): Likewise.
8198 2018-03-08 Martin Liska <mliska@suse.cz>
8200 PR gcov-profile/84735
8201 * doc/gcov.texi: Document usage of profile files.
8202 * gcov-io.h: Document changes in the format.
8204 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8208 * dwarf2out.c (struct dw_line_info_table): Update comments for
8210 (FORCE_RESET_NEXT_VIEW): New.
8211 (FORCE_RESETTING_VIEW_P): New.
8212 (RESETTING_VIEW_P): Check for -1 too.
8213 (ZERO_VIEW_P): Likewise.
8214 (new_line_info_table): Force-reset next view.
8215 (dwarf2out_begin_function): Likewise.
8216 (dwarf2out_source_line): Simplify zero_view_p initialization.
8217 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8218 view directly. Omit view when omitting .loc at line 0.
8220 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8222 PR tree-optimization/84740
8223 * tree-switch-conversion.c (process_switch): Call build_constructors
8224 only if info.phi_count is non-zero.
8226 PR tree-optimization/84739
8227 * tree-tailcall.c (find_tail_calls): Check call arguments against
8228 DECL_ARGUMENTS (current_function_decl) rather than
8229 DECL_ARGUMENTS (func) when checking for tail recursion.
8231 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8233 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8234 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8235 Volker Reichelt's entry and add entries for people that perform
8236 GCC fuzzy testing and report numerous bugs.
8238 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8241 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8242 readonly data in sdata, if that is disabled.
8243 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8244 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8245 -mreadonly-in-sdata option.
8247 2018-03-07 Martin Sebor <msebor@redhat.com>
8249 PR tree-optimization/84468
8250 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8251 basic block when looking for nul assignment.
8253 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8256 * except.h (output_function_exception_table): Adjust prototype.
8257 * except.c (output_function_exception_table): Remove FNNAME parameter
8258 and add SECTION parameter. Ouput one part of the table at a time.
8259 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8260 the first part of the exception table and emit unwind directives.
8261 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8262 (i386_pe_seh_cold_init): Likewise.
8263 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8264 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8265 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8266 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8267 * config/i386/winnt.c: Include except.h.
8268 (struct seh_frame_state): Add reg_offset, after_prologue and
8269 in_cold_section fields.
8270 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8271 (i386_pe_seh_cold_init): New function.
8272 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8273 to seh->in_cold_section.
8274 (seh_emit_push): Record the offset of the push.
8275 (seh_emit_save): Record the offet of the save.
8276 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8277 Test seh->after_prologue to disregard the epilogue.
8278 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8279 (i386_pe_end_cold_function): New function.
8281 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8284 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8285 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8288 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8290 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8291 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8294 * multiple_target.c: Include tree-inline.h and intl.h.
8295 (expand_target_clones): Diagnose and fail if node->definition and
8296 !tree_versionable_function_p (node->decl).
8298 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8300 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8302 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8303 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8304 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8306 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8309 * combine.c (try_combine): Use reg_or_subregno instead of handling
8310 just paradoxical SUBREGs and REGs.
8312 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8314 * config/arc/arc.c (arc_finalize_pic): Remove function.
8315 (arc_must_save_register): We use single base PIC register, remove
8316 checks to save/restore the PIC register.
8317 (arc_expand_prologue): Likewise.
8318 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8320 (arc_verify_short): Likewise.
8321 (arc_attr_type): Likewise.
8322 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8323 (walk_stores): Likewise.
8324 (arc_address_cost): Make it static.
8325 (arc_verify_short): Likewise.
8326 (branch_dest): Likewise.
8327 (arc_attr_type): Likewise.
8328 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8329 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8330 (arc_final_prescan_insn): Remove inserting the nops due to
8331 hardware hazards. It is done in reorg step.
8332 (insn_length_variant_t): Remove.
8333 (insn_length_parameters_t): Likewise.
8334 (arc_insn_length_parameters): Likewise.
8335 (arc_get_insn_variants): Likewise.
8336 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8338 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8341 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8344 PR tree-optimization/84687
8345 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8347 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8349 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8351 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8352 Rename to ppc_speculation_barrier.
8353 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8354 __builtin_ppc_speculation_barrier.
8356 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8359 * combine.c (combine_simplify_rtx): Don't try to simplify if
8360 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8363 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8365 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8366 to 32 bytes when compiling for POWER9.
8368 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8371 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8372 regparm >= 3 with no arg reg available also for calls with
8373 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8376 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8378 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8380 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8383 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8385 2018-03-05 Richard Biener <rguenther@suse.de>
8387 PR tree-optimization/84486
8388 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8389 When inserting a __builtin_assume_aligned call set the LHS
8390 SSA name alignment info accordingly.
8392 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8394 PR tree-optimization/84114
8395 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8396 Avoid reassociation of FLOAT_MODE addition.
8398 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8400 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8401 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8402 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8403 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8404 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8405 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8406 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8408 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8409 __builtin_ia32_wbinvd): New builtins.
8410 (SPECIAL_ARGS2): New.
8411 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8412 (SPECIAL_ARGS2): New.
8413 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8414 (ix86_valid_target_attribute_inner_p): Ditto.
8415 (ix86_init_mmx_sse_builtins): Add special_args2.
8416 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8417 TARGET_WBNOINVD_P): New.
8418 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8419 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8420 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8421 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8422 * config/i386/pconfigintrin.h: New file.
8423 * config/i386/wbnoinvdintrin.h: Ditto.
8424 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8426 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8428 2018-03-05 Richard Biener <rguenther@suse.de>
8430 PR tree-optimization/84670
8431 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8433 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8434 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8435 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8436 doesn't grow if all successors (recursively) were visited at least
8439 2018-03-05 Richard Biener <rguenther@suse.de>
8441 PR tree-optimization/84650
8442 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8443 if executed in the loop pipeline.
8445 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8447 * doc/configfiles.texi (Configuration Files): Move info about
8448 conditionalizing $target-protos.h to...
8449 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8450 differs from $target-protos.h.
8452 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8453 Chung-Ju Wu <jasonwucj@gmail.com>
8455 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8456 * config/nds32/nds32-multiple.md (setmemsi): Define.
8457 * config/nds32/nds32-memory-manipulation.c
8458 (nds32_gen_dup_4_byte_to_word_value): New.
8459 (emit_setmem_word_loop): New.
8460 (emit_setmem_byte_loop): New.
8461 (nds32_expand_setmem_loop): New.
8462 (nds32_expand_setmem_loop_v3m): New.
8463 (nds32_expand_setmem_unroll): New.
8464 (nds32_expand_setmem): New.
8466 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8467 Chung-Ju Wu <jasonwucj@gmail.com>
8469 * config/nds32/nds32-memory-manipulation.c
8470 (nds32_emit_load_store): New.
8471 (nds32_emit_post_inc_load_store): New.
8472 (nds32_emit_mem_move): New.
8473 (nds32_emit_mem_move_block): New.
8474 (nds32_expand_movmemsi_loop_unknown_size): New.
8475 (nds32_expand_movmemsi_loop_known_size): New.
8476 (nds32_expand_movmemsi_loop): New.
8477 (nds32_expand_movmemsi_unroll): New.
8478 (nds32_expand_movmemqi): Rename ...
8479 (nds32_expand_movmemsi): ... to this.
8480 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8481 (movmemsi): ... to this.
8482 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8483 (nds32_expand_movmemsi): ... to this.
8485 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8486 Monk Chiang <sh.chiang04@gmail.com>
8487 Chung-Ju Wu <jasonwucj@gmail.com>
8489 * config/nds32/nds32-protos.h
8490 (nds32_expand_load_multiple): New arguments.
8491 (nds32_expand_store_multiple): Ditto.
8492 (nds32_valid_multiple_load_store): Rename ...
8493 (nds32_valid_multiple_load_store_p): ... to this.
8494 * config/nds32/nds32-memory-manipulation.c
8495 (nds32_expand_load_multiple): Refine implementation.
8496 (nds32_expand_store_multiple): Ditto.
8497 * config/nds32/nds32-multiple.md
8498 (load_multiple): Update nds32_expand_load_multiple interface.
8499 (store_multiple): Update nds32_expand_store_multiple interface.
8500 * config/nds32/nds32-predicates.c
8501 (nds32_valid_multiple_load_store): Rename ...
8502 (nds32_valid_multiple_load_store_p): ... to this and refine
8504 * config/nds32/predicates.md
8505 (nds32_load_multiple_and_update_address_operation): New predicate.
8506 (nds32_store_multiple_and_update_address_operation): New predicate.
8508 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8509 Chung-Ju Wu <jasonwucj@gmail.com>
8511 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8512 (combo): New attribute.
8513 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8515 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8517 * config/nds32/nds32.opt: Change -mcmodel= default value.
8519 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8520 Monk Chiang <sh.chiang04@gmail.com>
8521 Chung-Ju Wu <jasonwucj@gmail.com>
8523 * config/nds32/constants.md (unspec_element): New enum.
8524 * config/nds32/constraints.md (Umw): New constraint.
8525 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8526 * config/nds32/nds32-intrinsic.md: Likewise.
8527 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8528 (nds32_valid_smw_lwm_base_p): New.
8529 (nds32_output_smw_single_word): New.
8530 (nds32_output_lmw_single_word): New.
8531 (nds32_expand_unaligned_load): New.
8532 (nds32_expand_unaligned_store): New.
8533 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8534 (nds32_output_smw_single_word): Declare.
8535 (nds32_output_lmw_single_word): Declare.
8536 (nds32_expand_unaligned_load): Declare.
8537 (nds32_expand_unaligned_store): Declare.
8538 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8539 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8540 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8541 NDS32_BUILTIN_UASTORE_DW.
8542 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8545 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8546 Kito Cheng <kito.cheng@gmail.com>
8547 Chung-Ju Wu <jasonwucj@gmail.com>
8549 * config/nds32/nds32-intrinsic.c
8550 (nds32_expand_builtin_null_ftype_reg): Delete.
8551 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8552 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8553 (nds32_read_argument): New.
8554 (nds32_legitimize_target): Ditto.
8555 (nds32_legitimize_argument): Ditto.
8556 (nds32_check_constant_argument): Ditto.
8557 (nds32_expand_unop_builtin): Ditto.
8558 (nds32_expand_unopimm_builtin): Ditto.
8559 (nds32_expand_binop_builtin): Ditto.
8560 (nds32_builtin_decl_impl): Ditto.
8561 (builtin_description): Ditto.
8562 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8563 (nds32_init_builtins_impl): Ditto.
8564 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8565 (nds32_builtin_decl): New.
8566 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8567 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8569 2018-03-02 Jeff Law <law@redhat.com>
8571 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8572 (redundant_insn, fill_simple_delay_slots): Likewise.
8573 (fill_slots_from_thread): Likewise.
8574 * resource.c (mark_referenced_resources): Likewise.
8575 (mark_set_resources, find_dead_or_set_registers): Likewise.
8577 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8579 * substring-locations.h (format_warning_va): Formatting fix for
8581 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8583 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8584 * substring-locations.c: Include intl.h.
8585 (format_warning_va): Turned into small wrapper around
8586 format_warning_n_va, renamed to ...
8587 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8588 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8590 (format_warning_at_substring_n): New function.
8591 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8592 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8593 format_warning_at_substring with just a shorter name instead of
8594 const function pointer.
8595 (fmtwarn_n): New function.
8596 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8597 appropriate, get rid of all the fmtstr temporaries, move conditionals
8598 with G_() wrapped string literals directly into fmtwarn arguments,
8599 cast dir.len to (int), formatting fixes.
8601 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8603 * doc/invoke.texi: Remove "Cilk Plus" references.
8605 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8606 Richard Biener <rguenther@suse.de>
8609 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8610 for error or warning attributes if CALL_FROM_THUNK_P is set.
8613 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8616 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8617 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8620 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8621 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8623 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8624 instead of -1U in last predictors element's probability member.
8626 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8629 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8630 arguments if they are comparable.
8632 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8634 PR tree-optimization/84634
8635 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8636 masks and masked_loop_p with a single loop_masks, making sure it's
8637 null for bb vectorization.
8639 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8641 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8642 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8643 loop->force_vectorize to check whether there is no alias.
8645 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8648 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8650 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8652 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8653 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8654 instead of a loop around prev_real_insn.
8655 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8659 * config/i386/i386.c (ix86_print_operand): Use conditional
8660 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8663 2018-03-02 Richard Biener <rguenther@suse.de>
8665 PR tree-optimization/84427
8666 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8667 (bitmap_set_subtract_values): Rewrite to handle multiple
8670 (prune_clobbered_mems): Likewise.
8671 (phi_translate): Take edge instead of pred/phiblock.
8672 (phi_translate_1): Likewise.
8673 (phi_translate_set): Likewise. Insert all translated
8674 exprs for a value into the set, keeping possibly multiple
8675 expressions per value.
8676 (compute_antic_aux): Adjust for phi_translate changes.
8677 When intersecting union the expressions and prune those
8678 not in the final value set, keeping possibly multiple
8679 expressions per value. Do not use value-insertion
8680 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8681 all expressions. Add verification that the value-sets
8682 only shrink during iteration.
8683 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8684 (do_pre_regular_insertion): Likewise.
8685 (do_pre_partial_partial_insertion): Likewise.
8687 2018-03-02 Richard Biener <rguenther@suse.de>
8690 * config/darwin.c (saved_debug_info_level): New static global.
8691 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8692 (darwin_asm_lto_end): Restore debug info generation settings.
8694 2018-03-01 Martin Liska <mliska@suse.cz>
8697 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8700 2018-03-01 Richard Biener <rguenther@suse.de>
8703 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8704 type annotation with LTO when debug was disabled at compile-time.
8706 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8708 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8710 (mips_final_postscan_insn): Likewise.
8712 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8714 PR rtl-optimization/84528
8715 * alias.c (init_alias_target): Add commentary.
8716 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8717 a unique base value if the frame pointer is not eliminated
8718 to the stack pointer.
8720 2018-03-01 Tom de Vries <tom@codesourcery.com>
8722 PR rtl-optimization/83327
8723 * lra-int.h (hard_regs_spilled_into): Declare.
8724 * lra.c (hard_regs_spilled_into): Define.
8725 (init_reg_info): Init hard_regs_spilled_into.
8726 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8727 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8728 (process_bb_lives): Handle hard_regs_spilled_into.
8729 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8730 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8732 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8734 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8735 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8736 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8737 * config/rs6000/aix72.h: New file.
8739 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8741 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8742 instead of warning_at with conditional singular and plural messages
8746 * stor-layout.c (update_alignment_for_field): For
8747 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8748 && !DECL_PACKED (field), do the alignment update, just use
8749 only desired_align instead of MAX (type_align, desired_align)
8751 (place_field): Don't do known_align < desired_align handling
8752 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8753 is non-NULL, instead do it after rli->prev_field handling and
8754 only if not within a bitfield word. For DECL_PACKED (field)
8755 use type_align of BITS_PER_UNIT.
8757 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8759 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8760 superfluous parentheses and trailing spaces.
8762 2018-02-28 Richard Biener <rguenther@suse.de>
8764 PR tree-optimization/84584
8765 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8766 SCoPs with fake exit edge.
8768 2018-02-28 Martin Liska <mliska@suse.cz>
8771 * timevar.c (timer::print): Fix format to properly print 100%
8774 2018-02-28 Richard Biener <rguenther@suse.de>
8777 * genmatch.c (capture_info::walk_match): Do not mark
8778 captured expressions without operands as expr_p given
8779 they act more like predicates and should be subject to
8780 "lost tail" side-effect preserving.
8782 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8784 PR rtl-optimization/81611
8785 * auto-inc-dec.c (attempt_change): Move dead note from
8786 mem_insn if it's the next use of regno
8787 (find_address): Take address use of reg holding
8788 non-incremented value. Add parm to limit search to the named
8790 (merge_in_block): Attempt to use a mem insn that is the next
8791 use of the original regno.
8793 2018-02-27 Martin Sebor <msebor@redhat.com>
8796 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8797 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8799 2018-02-27 Martin Sebor <msebor@redhat.com>
8801 PR translation/84207
8802 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8803 n argument to unsigned HOST_WIDE_INT.
8804 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8805 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8806 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8807 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8809 2018-02-27 Richard Biener <rguenther@suse.de>
8811 PR tree-optimization/84512
8812 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8813 Do not use the estimate returned from record_stmt_cost for
8814 the scalar iteration cost but sum properly using add_stmt_cost.
8816 2018-02-27 Richard Biener <rguenther@suse.de>
8818 PR tree-optimization/84466
8819 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8820 Adjust last change to less strictly validate use operands.
8822 2018-02-27 Martin Liska <mliska@suse.cz>
8824 PR gcov-profile/84548
8825 * gcov.c (process_file): Allow partial overlap and consider it
8826 also as group functions.
8827 (output_lines): Properly calculate range of lines for a group.
8829 2018-02-27 Martin Liska <mliska@suse.cz>
8831 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8832 'ggc' suffixes. Change first column width.
8833 (timer::print): Fix formatting of the column.
8835 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8837 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8838 preserve inline entry blocks for the sake of debug inline
8839 entry point markers alone.
8840 (remove_unused_locals): Suggest in comments a better place to
8841 force the preservation of inline entry blocks that are
8842 otherwise unused, but do not preserve them.
8844 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8846 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8848 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8851 * config/i386/constraints.md (Bs): Replace
8852 ix86_indirect_branch_register with
8853 TARGET_INDIRECT_BRANCH_REGISTER.
8855 * config/i386/i386.md (indirect_jump): Likewise.
8856 (tablejump): Likewise.
8857 (*sibcall_memory): Likewise.
8858 (*sibcall_value_memory): Likewise.
8859 Peepholes of indirect call and jump via memory: Likewise.
8860 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8861 (*sibcall_value_GOT_32): Likewise.
8862 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8863 (GOT_memory_operand): Likewise.
8864 (call_insn_operand): Likewise.
8865 (sibcall_insn_operand): Likewise.
8866 (GOT32_symbol_operand): Likewise.
8867 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8869 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8871 PR rtl-optimization/83496
8872 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8873 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8874 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8875 redundant insn, if any.
8876 (relax_delay_slots): Likewise.
8877 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8879 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8881 PR tree-optimization/83965
8882 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8883 that grouped statements are part of a reduction chain. Return
8884 true if the statement is not marked as a reduction itself but
8886 (vect_recog_dot_prod_pattern): Don't check whether the statement
8887 is part of a group here.
8888 (vect_recog_sad_pattern): Likewise.
8889 (vect_recog_widen_sum_pattern): Likewise.
8891 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8894 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8896 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8899 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8901 (ix86_output_indirect_function_return): New prototype.
8902 (ix86_split_simple_return_pop_internal): Likewise.
8903 * config/i386/i386.c (indirect_return_via_cx): New.
8904 (indirect_return_via_cx_bnd): Likewise.
8905 (indirect_thunk_name): Handle return va CX_REG.
8906 (output_indirect_thunk_function): Create alias for
8907 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8908 (ix86_output_indirect_jmp): Remove the bool argument.
8909 (ix86_output_indirect_function_return): New function.
8910 (ix86_split_simple_return_pop_internal): Likewise.
8911 * config/i386/i386.md (*indirect_jump): Don't pass false
8912 to ix86_output_indirect_jmp.
8913 (*tablejump_1): Likewise.
8914 (simple_return_pop_internal): Change it to define_insn_and_split.
8915 Call ix86_split_simple_return_pop_internal to split it for
8917 (simple_return_indirect_internal): Call
8918 ix86_output_indirect_function_return instead of
8919 ix86_output_indirect_jmp.
8921 2018-02-26 Jakub Jelinek <jakub@redhat.com>
8924 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
8925 memset and value initialization afterwards.
8927 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
8929 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
8931 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8934 * common/config/aarch64/aarch64-common.c
8935 (aarch_option_optimization_table[]): Switch
8936 off fomit-frame-pointer
8938 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8939 Chung-Ju Wu <jasonwucj@gmail.com>
8941 * config/nds32/nds32-multiple.md (load_multiple): Disallow
8943 (store_multiple): Ditto.
8945 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
8947 * config.gcc: Add --with-cpu support for nds32 target.
8948 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
8949 * config/nds32/nds32.opt: Add -mcpu= option.
8951 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
8953 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
8954 isel=yes): Warn for these deprecated options.
8956 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
8958 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
8959 ISA_2_5_MASKS_EMBEDDED.
8961 2018-02-23 Jakub Jelinek <jakub@redhat.com>
8963 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
8964 p->max as pointers rather than using iterative_hash_expr.
8966 2018-02-23 Carl Love <cel@us.ibm.com>
8968 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
8969 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
8971 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
8972 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
8973 P8V_BUILTIN_VEC_VUNSIGNED2.
8975 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
8978 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
8979 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
8981 * lra-constraints.c (curr_insn_transform): Set up
8982 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
8984 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
8985 * lra-eliminations.c (spill_pseudos): Ditto.
8986 (process_insn_for_elimination): Ditto.
8987 * lra-lives.c (reg_early_clobber_p): Use the new macros.
8988 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
8989 LRA_NON_CLOBBERED_ALT.
8991 2018-02-22 Martin Sebor <msebor@redhat.com>
8993 PR tree-optimization/84480
8994 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
8995 to maybe_diag_stxncpy_trunc. Call it.
8996 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
8997 from gimple_fold_builtin_strcpy. Print inlining stack.
8998 (handle_builtin_stxncpy): Print inlining stack.
8999 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9001 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9004 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9005 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9006 and -fcheck-pointer-bounds are used together.
9007 (indirect_thunk_prefix): New enum.
9008 (indirect_thunk_need_prefix): New function.
9009 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9010 "_nt" instead of "_bnd" for NOTRACK prefix.
9011 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9012 (output_indirect_thunk_function): Likewise.
9014 (ix86_code_end): Update output_indirect_thunk_function calls.
9015 (ix86_output_indirect_branch_via_reg): Replace
9016 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9017 (ix86_output_indirect_branch_via_push): Likewise.
9018 (ix86_output_function_return): Likewise.
9019 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9020 incompatible with -fcf-protection=branch and
9021 -fcheck-pointer-bounds.
9023 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9026 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9027 Change gcc_assert call to output_operand_lossage.
9029 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9031 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9033 2018-02-22 DJ Delorie <dj@redhat.com>
9034 Sebastian Perta <sebastian.perta@renesas.com>
9035 Oleg Endo <olegendo@gcc.gnu.org>
9037 * config/rx/rx.c (rx_rtx_costs): New function.
9038 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9040 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9042 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9044 2018-02-22 Martin Liska <mliska@suse.cz>
9047 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9048 Add "native" as a possible value.
9050 2018-02-22 Martin Liska <mliska@suse.cz>
9053 * config/i386/i386.c (ix86_option_override_internal):
9054 Add "native" as a possible value for -march and -mtune.
9056 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9059 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9060 to all type variants.
9062 PR tree-optimization/84503
9063 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9064 width as info->bitpos + info->bitsize - start.
9065 (merged_store_group::merge_overlapping): Simplify width computation.
9066 (check_no_overlap): New function.
9067 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9068 start + width and last_order of the group, fail if check_no_overlap
9070 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9071 to group if check_no_overlap fails.
9073 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9075 * config/rs6000/altivec.md: Delete contraint arguments to
9076 define_expand, define_split, and define_peephole2, and in
9077 define_insn_and_split if always unused.
9078 * config/rs6000/darwin.md: Ditto.
9079 * config/rs6000/dfp.md: Ditto.
9080 * config/rs6000/rs6000.md: Ditto.
9081 * config/rs6000/sync.md: Ditto.
9082 * config/rs6000/vector.md: Ditto.
9083 * config/rs6000/vsx.md: Ditto.
9085 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9087 * config/rs6000/altivec.md: Write output control strings as braced
9088 blocks instead of double-quoted strings.
9089 * config/rs6000/darwin.md: Ditto.
9090 * config/rs6000/rs6000.md: Ditto.
9091 * config/rs6000/vector.md: Ditto.
9092 * config/rs6000/vsx.md: Ditto.
9094 2018-02-21 Jason Merrill <jason@redhat.com>
9096 PR c++/84314 - ICE with templates and fastcall attribute.
9097 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9099 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9101 * ipa-cp.c (determine_versionability): Fix comment typos.
9103 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9106 * ipa-cp.c (determine_versionability): Do not version functions caling
9109 2018-02-21 Martin Liska <mliska@suse.cz>
9112 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9113 Add "native" as a possible value.
9114 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9115 the macro when native cpu detection is available.
9117 2018-02-21 Martin Liska <mliska@suse.cz>
9120 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9121 Add "native" as a possible value.
9122 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9123 when native cpu detection is available.
9125 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9126 Martin Sebor <msebor@redhat.com>
9128 PR tree-optimization/84478
9129 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9131 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9132 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9133 support which is conservatively correct, for 2 only stay conservative
9134 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9135 argument to the 2 argument get_range_strlen, adjust 6 arg
9136 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9138 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9139 (gimple_fold_builtin_strlen): Pass true as last argument to
9142 2018-02-20 Martin Sebor <msebor@redhat.com>
9145 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9146 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9147 (builtin_memref::builtin_memref): Factor out parts into
9148 set_base_and_offset and call it.
9150 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9153 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9154 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9155 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9156 search at the associated MODE_INT.
9158 2018-02-20 Jeff Law <law@redhat.com>
9161 PR tree-optimization/81592
9163 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9164 for range data rather than using global data.
9165 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9166 range data rather than using global data.
9167 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9168 pass it to children as needed.
9169 (struct directive::fmtresult): Similarly.
9170 (struct directive::set_width): Similarly.
9171 (struct directive::set_precision): Similarly.
9172 (format_integer, format_directive, parse_directive): Similarly.
9173 (format_none): Accept unnamed vr_values parameter.
9174 (format_percent, format_floating, format_character): Similarly.
9175 (format_string, format_plain): Similarly.
9176 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9177 the EVRP range analyzer for range data rather than using global data.
9178 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9179 gimple-ssa-evrp-analyze.h
9180 (class sprintf_dom_walker): Add after_dom_children member function.
9181 Add evrp_range_analyzer member.
9182 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9183 range analyzer as needed.
9184 (sprintf_dom_walker::after_dom_children): New member function.
9185 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9187 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9188 (evrp_range_analyzer::pop_to_marker): Likewise.
9190 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9192 PR tree-optimization/84419
9193 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9194 with the required type if its current type is compatible but
9197 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9200 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9201 after vectorization.
9203 2018-02-20 Martin Liska <mliska@suse.cz>
9206 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9207 possible values if we don't have a hint.
9209 2018-02-20 Martin Liska <mliska@suse.cz>
9213 * final.c (shorten_branches): Build align_tab array with one
9215 * opts.c (finish_options): Add alignment option limit check.
9216 (MAX_CODE_ALIGN): Likewise.
9217 (MAX_CODE_ALIGN_VALUE): Likewise.
9218 * doc/invoke.texi: Document maximum allowed option value for
9219 all -falign-* options.
9221 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9224 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9225 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9226 * var-tracking.c (emit_note_insn_var_location): Remove all references
9227 to NOTE_INSN_CALL_ARG_LOCATION.
9228 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9229 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9230 Use copy_rtx_if_shared.
9231 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9232 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9233 (dwarf2out_var_location): Remove handling of
9234 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9236 * final.c (final_scan_insn): Remove all references to
9237 NOTE_INSN_CALL_ARG_LOCATION.
9238 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9239 before dumping final insns.
9240 * except.c (emit_note_eh_region_end): Remove all references to
9241 NOTE_INSN_CALL_ARG_LOCATION.
9242 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9243 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9244 * config/arc/arc.c (hwloop_optimize): Likewise.
9245 * config/arm/arm.c (create_fix_barrier): Likewise.
9246 * config/s390/s390.c (s390_chunkify_start): Likewise.
9247 * config/sh/sh.c (find_barrier): Likewise.
9248 * config/i386/i386.c (rest_of_insert_endbranch,
9249 ix86_seh_fixup_eh_fallthru): Likewise.
9250 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9251 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9252 * config/frv/frv.c (frv_function_prologue): Likewise.
9253 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9255 (note_outside_basic_block_p): Remove all references to
9256 NOTE_INSN_CALL_ARG_LOCATION.
9257 * gengtype.c (adjust_field_rtx_def): Likewise.
9258 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9260 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9261 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9264 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9267 PR tree-optimization/84452
9268 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9269 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9272 2018-02-19 Martin Liska <mliska@suse.cz>
9275 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9277 2018-02-19 Martin Liska <mliska@suse.cz>
9278 Richard Sandiford <richard.sandiford@linaro.org>
9280 PR tree-optimization/82491
9281 * gimple-fold.c (get_base_constructor): Make earlier bail out
9284 2018-02-19 Carl Love <cel@us.ibm.com>
9286 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9287 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9289 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9290 P8V_BUILTIN_VEC_NEG.
9292 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9294 * config/rl78/rl78.md (movdf): New define expand.
9296 2018-02-19 Martin Liska <mliska@suse.cz>
9299 * doc/invoke.texi: Fix typo.
9300 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9302 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9304 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9305 handle rs6000_single_float and rs6000_double_float specially for
9308 2018-02-16 Jeff Law <law@redhat.com>
9310 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9313 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9316 * ipa-inline.c (inline_small_functions): Fix a typo.
9318 2018-02-16 Nathan Sidwell <nathan@acm.org>
9320 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9322 2018-02-16 Carl Love <cel@us.ibm.com>
9324 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9325 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9326 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9327 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9328 expansion to P8V_BUILTIN_VEC_FLOAT2.
9330 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9332 PR rtl-optimization/70023
9333 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9334 src_regno into account.
9336 2018-02-16 Carl Love <cel@us.ibm.com>
9338 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9339 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9340 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9341 * config/rs6000/rs6000.c: Remove case statements for
9342 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9343 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9344 and P9V_BUILTIN_VEC_VINSERT4B.
9345 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9346 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9347 * config/rs6000/vsx.md:
9348 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9351 2018-02-16 Carl Love <cel@us.ibm.com>
9353 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9355 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9357 * config/rs6000/rs6000-c.c: Add the definitions for
9358 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9359 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9360 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9361 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9362 definition for insert4b and define insn *insert3b_internal.
9363 * doc/extend.texi: Add documentation for vec_extract4b.
9365 2018-02-16 Nathan Sidwell <nathan@acm.org>
9367 * doc/extend.texi (Backwards Compatibility): Mention friend
9368 injection. Note for-scope is deprecated.
9369 * doc/invoke.texi (-ffriend-injection): Deprecate.
9371 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9373 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9374 that moved to I2, also allow destinations that are a paradoxical
9375 subreg (instead of a normal reg).
9377 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9380 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9383 2018-02-16 Richard Biener <rguenther@suse.de>
9385 PR tree-optimization/84037
9386 PR tree-optimization/84016
9388 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9389 Adjust vec_construct for the fact we need additional higher latency
9390 128bit inserts for AVX256 and AVX512 vector builds.
9391 (ix86_add_stmt_cost): Scale vector construction cost for
9394 2018-02-16 Richard Biener <rguenther@suse.de>
9396 PR tree-optimization/84417
9397 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9398 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9399 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9401 2018-02-16 Martin Liska <mliska@suse.cz>
9404 * internal-fn.def (ASAN_CHECK): Set proper flags.
9405 (ASAN_MARK): Likewise.
9407 2018-02-16 Julia Koval <julia.koval@intel.com>
9409 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9410 from PTA_CANNONLAKE.
9412 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9415 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9416 Use ++iter rather than iter++ for std::list iterators.
9417 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9418 defer deleting them until all nodes in the forest are processed. Do
9419 free even leaf nodes. Change to_process into auto_vec.
9422 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9423 * vec.h (vec_default_construct): Use memset instead of placement new
9424 if BROKEN_VALUE_INITIALIZATION is defined.
9425 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9426 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9429 PR rtl-optimization/83723
9430 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9431 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9432 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9434 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9436 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9438 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9440 PR rtl-optimization/81443
9441 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9442 from inner REGs to paradoxical SUBREGs.
9444 2018-02-16 Richard Biener <rguenther@suse.de>
9446 PR tree-optimization/84399
9447 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9448 For operands we can analyze at their definition make sure we can
9449 analyze them at each use as well.
9451 2018-02-16 Richard Biener <rguenther@suse.de>
9453 PR tree-optimization/84190
9454 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9455 volatile accesses if the decl isn't volatile.
9457 2018-02-15 Jason Merrill <jason@redhat.com>
9459 PR c++/84314 - ICE with templates and fastcall attribute.
9460 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9461 TYPE_CANONICAL on an existing type.
9463 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9465 PR tree-optimization/84383
9466 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9467 dstoff nor call operand_equal_p if dstbase is NULL.
9469 PR tree-optimization/84334
9470 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9471 also a CONSTANT_CLASS_P, punt.
9473 2018-02-14 Jim Wilson <jimw@sifive.com>
9475 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9476 first SMALL_OPERAND check. New local min_second_step. Move assert
9477 to where locals are set. Add TARGET_RVC support.
9478 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9480 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9482 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9484 2018-02-14 Martin Sebor <msebor@redhat.com>
9486 PR tree-optimization/83698
9487 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9488 arrays constrain the offset range to their bounds.
9489 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9490 (builtin_access::overlap): Avoid setting the size of overlap if it's
9492 (maybe_diag_overlap): Also consider arrays when deciding what values
9493 of offsets to include in diagnostics.
9495 2018-02-14 Martin Sebor <msebor@redhat.com>
9498 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9499 that correspond to the kind of a declaration.
9501 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9504 * config/pa/pa.md: Load address of PIC label using the linkage table
9505 if the label is nonlocal.
9507 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9509 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9510 warning message if user requests -maltivec=be.
9511 * doc/invoke.texi: Document deprecation of -maltivec=be.
9513 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9516 * config/rs6000/rs6000-c.c: Update definitions for
9517 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9518 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9520 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9523 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9524 add _get_ssp intrinsics. Remove argument from
9525 __builtin_ia32_rdssp[d|q].
9526 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9527 * config/i386/i386-builtin.def: Remove argument from
9528 __builtin_ia32_rdssp[d|q].
9529 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9530 ix86_expand_special_args_builtin for _rdssp[d|q].
9531 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9532 Clear register before usage.
9533 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9534 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9536 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9538 PR tree-optimization/84357
9539 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9540 operand 1 of an ARRAY_REF too.
9542 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9545 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9546 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9548 (set_of_reg): New struct.
9549 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9550 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9551 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9553 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9554 Split into bitclr, bitset, bitinvert patterns if appropriate.
9555 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9556 use rx_fuse_in_memory_bitop.
9557 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9558 to named insn, correct maximum insn length.
9560 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9563 * machmode.def: Define a complex mode for PARTIAL_INT.
9564 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9566 * doc/rtl.texi: Document CSPImode.
9567 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9569 (msp430_hard_regno_nregs_with_padding): Likewise.
9571 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9574 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9576 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9578 PR rtl-optimization/84169
9579 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9580 we generated a parallel as new i3 and we split that to new i2 and i3
9581 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9582 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9583 those to i2, not i1. Partially rewrite this scan code.
9585 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9588 * stor-layout.c (place_field): For variable length fields, adjust
9589 offset_align afterwards not just based on the field's alignment,
9590 but also on the size.
9593 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9594 of exps and logs in the use_exp2 case.
9596 2018-02-13 Jeff Law <law@redhat.com>
9598 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9601 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9604 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9608 * common.opt (gas-loc-support, gas-locview-support): New.
9609 (ginline-points, ginternal-reset-location-views): New.
9610 * doc/invoke.texi: Document them. Use @itemx where intended.
9611 (gvariable-location-views): Adjust.
9612 * target.def (reset_location_view): New.
9613 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9614 (TARGET_RESET_LOCATION_VIEW): New.
9615 * doc/tm.texi: Rebuilt.
9616 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9617 (dwarf2out_default_as_locview_support): New.
9618 (output_asm_line_debug_info): Use option variables.
9619 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9620 (output_loc_list): Likewise.
9621 (add_high_low_attributes): Check option variables.
9622 Don't output entry view attribute in strict mode.
9623 (gen_inlined_subroutine_die): Check option variables.
9624 (dwarf2out_inline_entry): Likewise.
9625 (init_sections_and_labels): Likewise.
9626 (dwarf2out_early_finish): Likewise.
9627 (maybe_reset_location_view): New, from...
9628 (dwarf2out_var_location): ... here. Call it.
9629 * debug.h (dwarf2out_default_as_loc_support): Declare.
9630 (dwarf2out_default_as_locview_support): Declare.
9631 * hooks.c (hook_int_rtx_insn_0): New.
9632 * hooks.h (hook_int_rtx_insn_0): Declare.
9633 * toplev.c (process_options): Take -gas-loc-support and
9634 -gas-locview-support from dwarf2out. Enable
9635 -gvariable-location-views by default only with locview
9636 assembler support. Enable -ginternal-reset-location-views by
9637 default only if the target defines the corresponding hook.
9638 Enable -ginline-points by default if location views are
9639 enabled; force it disabled if statement frontiers are
9641 * tree-inline.c (expand_call_inline): Check option variables.
9642 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9644 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9646 PR tree-optimization/84321
9647 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9648 handling. Also check whether the anti-range contains any values
9649 that satisfy the mask; switch to a VR_RANGE if not.
9651 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9654 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9656 2018-02-13 Martin Jambor <mjambor@suse.cz>
9659 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9660 of call statements, also set location of a load to a temporary.
9662 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9664 * config/rl78/rl78.c (add_vector_labels): New function.
9665 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9666 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9667 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9668 which checks that no arguments are passed.
9669 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9670 * doc/extend.texi: Documentation for the new attribute.
9672 2018-02-13 Andreas Schwab <schwab@suse.de>
9674 * config/riscv/linux.h (CPP_SPEC): Define.
9676 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9679 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9680 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9681 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9682 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9683 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9684 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9685 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9686 temporarily for AES and PCLMUL builtins.
9688 PR tree-optimization/84339
9689 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9690 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9694 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9695 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9696 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9698 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9700 * omp-simd-clone.h: New file.
9701 * omp-simd-clone.c: Include omp-simd-clone.h.
9702 (expand_simd_clones): No longer static.
9703 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9704 cgraph.h and omp-simd-clone.h.
9705 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9706 (vect_recog_widen_shift_pattern): Formatting fix.
9707 (vect_pattern_recog_1): Don't check optab for calls.
9710 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9711 operands[2] into a REG before using gen_lowpart on it.
9713 2018-02-12 Jeff Law <law@redhat.com>
9716 * config/sh/sh.c (find_barrier): Consider a sibling call
9719 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9720 successfully back substituting a reg.
9722 2018-02-12 Richard Biener <rguenther@suse.de>
9724 PR tree-optimization/84037
9725 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9726 parameter, move visited init to caller.
9727 (vect_slp_analyze_operations): Separate cost from validity
9728 check, initialize visited once for all instances.
9729 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9731 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9733 (vect_model_promotion_demotion_cost): Likewise.
9734 (vectorizable_bswap): Guard cost modeling with !slp_node
9735 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9737 (vectorizable_call): Likewise.
9738 (vectorizable_conversion): Likewise.
9739 (vectorizable_assignment): Likewise.
9740 (vectorizable_shift): Likewise.
9741 (vectorizable_operation): Likewise.
9742 (vectorizable_store): Likewise.
9743 (vectorizable_load): Likewise.
9744 (vectorizable_condition): Likewise.
9745 (vectorizable_comparison): Likewise.
9747 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9750 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9751 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9752 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9754 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9757 * params.def (inline-min-speedup): Increase from 8 to 15.
9758 (max-inline-insns-auto): Decrease from 40 to 30.
9759 * ipa-split.c (consider_split): Add some buffer for function to
9760 be considered inlining candidate.
9761 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9764 2018-02-12 Richard Biener <rguenther@suse.de>
9766 PR tree-optimization/84037
9767 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9768 matched stmts if we cannot swap the non-matched ones.
9770 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9772 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9773 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9774 _mm_maskz_scalef_round_ss): New intrinsics.
9775 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9776 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9777 __builtin_ia32_scalefss_round): Remove.
9778 (__builtin_ia32_scalefsd_mask_round,
9779 __builtin_ia32_scalefss_mask_round): New intrinsics.
9780 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9781 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9782 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9783 "<round_constraint>")): Changed to ...
9784 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9785 "<round_scalar_constraint>")): ... this.
9786 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9787 %0, %1, %2<round_op3>}"): Changed to ...
9788 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9789 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9790 %2<round_scalar_mask_op3>}"): ... this.
9791 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9793 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9795 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9796 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9797 (_mm_maskz_sqrt_round_ss): New intrinsics.
9798 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9799 (__builtin_ia32_sqrtsd_mask_round)
9800 (__builtin_ia32_sqrtss_mask_round): New builtins.
9801 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9802 (__builtin_ia32_sqrtss_round): Remove.
9803 (__builtin_ia32_sqrtsd_mask_round)
9804 (__builtin_ia32_sqrtss_mask_round): New builtins.
9805 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9806 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9807 ((match_operand:VF_128 1 "vector_operand"
9808 "xBm,<round_constraint>")): Changed to ...
9809 ((match_operand:VF_128 1 "vector_operand"
9810 "xBm,<round_scalar_constraint>")): ... this.
9811 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9812 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9813 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9814 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9815 %<iptr>1<round_scalar_mask_op3>}): ... this.
9816 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9817 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9819 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9822 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9823 Cast vec_cmpeq result to correct type.
9824 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9825 Cast vec_cmpgt result to correct type.
9827 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9829 * final.c (final_scan_insn_1): Renamed from...
9830 (final_scan_insn): ... this. New wrapper, to recover
9831 seen from the outermost call in recursive ones.
9832 * config/sparc/sparc.c (output_return): Drop seen from call.
9833 (output_sibcall): Likewise.
9834 * config/visium/visium.c (output_branch): Likewise.
9836 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9838 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9841 2018-02-10 Alan Modra <amodra@gmail.com>
9844 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9845 Specify LR as an input.
9847 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9850 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9851 remove_member_access_dummy_vars): New functions.
9852 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9853 lower_omp_1, execute_lower_omp): Use them.
9855 PR rtl-optimization/84308
9856 * shrink-wrap.c (spread_components): Release todo vector.
9858 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9860 PR rtl-optimization/57193
9861 * ira-color.c (struct allocno_color_data): Add member
9862 conflict_allocno_hard_prefs.
9863 (update_conflict_allocno_hard_prefs): New.
9864 (bucket_allocno_compare_func): Add a preference based on
9865 conflict_allocno_hard_prefs.
9866 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9867 (color_allocnos): Remove a dead code. Initiate
9868 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9870 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9873 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9874 constraint from =wa to wa. Avoid a subreg on the output operand,
9875 instead use a pseudo and subreg it in a move.
9876 (p9_xxbrd_<mode>): Changed to ...
9877 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9878 (p9_xxbrd_v2df): New expander.
9879 (p9_xxbrw_<mode>): Changed to ...
9880 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9881 (p9_xxbrw_v4sf): New expander.
9883 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9885 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9887 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9890 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9891 multiply in 32-bit mode.
9892 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9893 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9896 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9898 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9899 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9900 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9901 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9903 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9906 * dwarf2out.c (is_trivial_indirect_ref): New function.
9907 (dwarf2out_late_global_decl): Do not generate a location
9908 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9909 and that are not defined in the current unit.
9911 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9913 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
9914 instead of a libcall for UNORDERED.
9916 2018-02-09 Tamar Christina <tamar.christina@arm.com>
9919 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
9920 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
9922 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9925 * config/s390/s390.c (s390_set_current_function): Invoke
9926 s390_indirect_branch_settings also if fndecl didn't change.
9928 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9930 * config/rs6000/rs6000.md (blockage): Set length to zero.
9932 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
9934 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
9936 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9939 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
9940 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
9944 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
9945 PARALLEL incoming that failed vt_get_decl_and_offset check.
9948 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
9949 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
9951 (get_variable_section): For decls in named .bss* sections pass true as
9952 second argument to bss_initializer_p.
9954 2018-02-09 Marek Polacek <polacek@redhat.com>
9955 Jakub Jelinek <jakub@redhat.com>
9958 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
9959 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
9960 Sync some changes from cxx_fold_indirect_ref.
9962 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
9964 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
9966 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
9967 (BLOCK_INLINE_ENTRY_LABEL): New.
9968 (dwarf2out_var_location): Disregard inline entry markers.
9969 (inline_entry_data): New struct.
9970 (inline_entry_data_hasher): New hashtable type.
9971 (inline_entry_data_hasher::hash): New.
9972 (inline_entry_data_hasher::equal): New.
9973 (inline_entry_data_table): New variable.
9974 (add_high_low_attributes): Add DW_AT_entry_pc and
9975 DW_AT_GNU_entry_view attributes if a pending entry is found
9976 in inline_entry_data_table. Add old entry_pc attribute only
9977 if debug nonbinding markers are disabled.
9978 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
9979 markers are enabled.
9980 (block_within_block_p, dwarf2out_inline_entry): New.
9981 (dwarf2out_finish): Check that no entries remained in
9982 inline_entry_data_table.
9983 * final.c (reemit_insn_block_notes): Handle inline entry notes.
9984 (final_scan_insn, notice_source_line): Likewise.
9985 (rest_of_clean_state): Skip inline entry markers.
9986 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
9988 * gimple.c (gimple_build_debug_inline_entry): New.
9989 * gimple.h (enum gimple_debug_subcode): Add
9990 GIMPLE_DEBUG_INLINE_ENTRY.
9991 (gimple_build_debug_inline_entry): Declare.
9992 (gimple_debug_inline_entry_p): New.
9993 (gimple_debug_nonbind_marker_p): Adjust.
9994 * insn-notes.def (INLINE_ENTRY): New.
9995 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
9996 inline entry marker notes.
9997 (print_insn): Likewise.
9998 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
9999 (INSN_DEBUG_MARKER_KIND): Likewise.
10000 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10001 * tree-inline.c (expand_call_inline): Build and insert
10002 debug_inline_entry stmt.
10003 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10004 inline entry blocks early, if nonbind markers are enabled.
10005 (dump_scope_block): Dump fragment info.
10006 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10007 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10008 (gimple_build_debug_inline_entry): New.
10009 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10010 Enable/disable inline entry points too.
10011 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10012 (DEBUG_INSN): Describe inline entry markers.
10014 * common.opt (gvariable-location-views): New.
10015 (gvariable-location-views=incompat5): New.
10016 * config.in: Rebuilt.
10017 * configure: Rebuilt.
10018 * configure.ac: Test assembler for view support.
10019 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10020 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10021 * dwarf2out.c (var_loc_view): New typedef.
10022 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10023 (dwarf2out_locviews_in_attribute): New.
10024 (dwarf2out_locviews_in_loclist): New.
10025 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10026 (enum dw_line_info_opcode): Add LI_adv_address.
10027 (struct dw_line_info_table): Add view.
10028 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10029 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10030 (zero_view_p): New variable.
10031 (ZERO_VIEW_P): New macro.
10032 (output_asm_line_debug_info): New.
10033 (struct var_loc_node): Add view.
10034 (add_AT_view_list, AT_loc_list): New.
10035 (add_var_loc_to_decl): Add view param. Test it against last.
10036 (new_loc_list): Add view params. Record them.
10037 (AT_loc_list_ptr): Handle loc and view lists.
10038 (view_list_to_loc_list_val_node): New.
10039 (print_dw_val): Handle dw_val_class_view_list.
10040 (size_of_die): Likewise.
10041 (value_format): Likewise.
10042 (loc_list_has_views): New.
10043 (gen_llsym): Set vl_symbol too.
10044 (maybe_gen_llsym, skip_loc_list_entry): New.
10045 (dwarf2out_maybe_output_loclist_view_pair): New.
10046 (output_loc_list): Output view list or entries too.
10047 (output_view_list_offset): New.
10048 (output_die): Handle dw_val_class_view_list.
10049 (output_dwarf_version): New.
10050 (output_compilation_unit_header): Use it.
10051 (output_skeleton_debug_sections): Likewise.
10052 (output_rnglists, output_line_info): Likewise.
10053 (output_pubnames, output_aranges): Update version comments.
10054 (output_one_line_info_table): Output view numbers in asm comments.
10055 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10056 Call maybe_gen_llsym.
10057 (loc_list_from_tree_1): Adjust.
10058 (add_AT_location_description): Create view list attribute if
10059 needed, check it's absent otherwise.
10060 (convert_cfa_to_fb_loc_list): Adjust.
10061 (maybe_emit_file): Call output_asm_line_debug_info for test.
10062 (dwarf2out_var_location): Reset views as needed. Precompute
10063 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10064 attribute. Set view.
10065 (new_line_info_table): Reset next view.
10066 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10067 (dwarf2out_source_line): Likewise. Output view resets and labels to
10068 the assembler, or select appropriate line info opcodes.
10069 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10070 (optimize_string_length): Catch it. Adjust.
10071 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10072 dw_val_class_view_list, and remove it if no longer needed.
10073 (hash_loc_list): Hash view numbers.
10074 (loc_list_hasher::equal): Compare them.
10075 (optimize_location_lists): Check whether a view list symbol is
10076 needed, and whether the locview attribute is present, and
10077 whether they match. Remove the locview attribute if no longer
10079 (index_location_lists): Call skip_loc_list_entry for test.
10080 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10081 Use output_dwarf_version.
10082 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10083 (struct dw_val_node): Add val_view_list.
10084 * final.c (SEEN_NEXT_VIEW): New.
10085 (set_next_view_needed): New.
10086 (clear_next_view_needed): New.
10087 (maybe_output_next_view): New.
10088 (final_start_function): Rename to...
10089 (final_start_function_1): ... this. Take pointer to FIRST,
10090 add SEEN parameter. Emit param bindings in the initial view.
10091 (final_start_function): Reintroduce SEEN-less interface.
10092 (final): Rename to...
10093 (final_1): ... this. Take SEEN parameter. Output final pending
10094 next view at the end.
10095 (final): Reintroduce seen-less interface.
10096 (final_scan_insn): Output pending next view before switching
10097 sections or ending a block. Mark the next view as needed when
10098 outputting variable locations. Notify debug backend of section
10099 changes, and of location view changes.
10100 (rest_of_handle_final): Adjust.
10101 * toplev.c (process_options): Autodetect value for debug variable
10102 location views option. Warn on incompat5 without -gdwarf-5.
10103 * doc/invoke.texi (gvariable-location-views): New.
10104 (gvariable-location-views=incompat5): New.
10105 (gno-variable-location-views): New.
10107 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10109 PR tree-optimization/84136
10110 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10111 that the result of find_edge is non-NULL.
10113 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10116 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10117 storing integer register in SImode. Fix cost of 256 and 512
10118 byte aligned SSE register store.
10120 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10122 * config/i386/i386.c (ix86_multiplication_cost): Fix
10123 multiplication cost for TARGET_AVX512DQ.
10125 2018-02-08 Marek Polacek <polacek@redhat.com>
10127 PR tree-optimization/84238
10128 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10131 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10133 PR tree-optimization/84265
10134 * tree-vect-stmts.c (vectorizable_store): Don't treat
10135 VMAT_CONTIGUOUS accesses as grouped.
10136 (vectorizable_load): Likewise.
10138 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10140 PR tree-optimization/81635
10141 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10142 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10143 (test_round_for_mask): New functions.
10144 (wide_int_cc_tests): Call test_round_for_mask.
10145 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10146 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10147 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10148 range returned by get_range_info.
10150 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10153 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10154 * symtab.c: Include builtins.h
10155 (symtab_node::output_to_lto_symbol_table_p): Move here
10156 from lto-streamer-out.c:output_symbol_p.
10157 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10158 (output_symbol_p): Move all logic to symtab.c
10159 (produce_symtab): Update.
10161 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10163 * config/s390/s390-opts.h (enum indirect_branch): Define.
10164 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10165 (s390_indirect_branch_via_thunk)
10166 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10167 (enum s390_indirect_branch_type): Define.
10168 * config/s390/s390.c (struct s390_frame_layout, struct
10169 machine_function): Remove.
10170 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10171 (indirect_branch_table_label_no, indirect_branch_table_name):
10173 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10174 (enum s390_indirect_branch_option): Define.
10175 (s390_return_addr_from_memory): New function.
10176 (s390_handle_string_attribute): New function.
10177 (s390_attribute_table): Add new attribute handler.
10178 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10179 (s390_indirect_branch_via_thunk): New function.
10180 (s390_indirect_branch_via_inline_thunk): New function.
10181 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10182 sibling call optimization for non z10 compiles.
10183 (s390_emit_call): Force indirect branch target to be a single
10184 register. Add r1 clobber for non-z10 compiles.
10185 (s390_emit_epilogue): Emit return jump via return_use expander.
10186 (s390_reorg): Handle JUMP_INSNs as execute targets.
10187 (s390_option_override_internal): Perform validity checks for the
10188 new command line options.
10189 (s390_indirect_branch_attrvalue): New function.
10190 (s390_indirect_branch_settings): New function.
10191 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10192 (s390_output_indirect_thunk_function): New function.
10193 (s390_code_end): Implement target hook.
10194 (s390_case_values_threshold): Implement target hook.
10195 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10197 * config/s390/s390.h (struct s390_frame_layout)
10198 (struct machine_function): Move here from s390.c.
10199 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10200 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10201 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10202 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10203 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10204 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10205 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10206 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10207 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10208 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10209 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10210 (mnemonic attribute): Add values which aren't recognized
10212 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10213 pattern for branch conversion. Fix mnemonic attribute.
10214 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10215 indirect branch via thunk if requested.
10216 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10217 ("*indirect_jump"): Disable for branch conversion using out of
10219 ("indirect_jump_via_thunk<mode>_z10")
10220 ("indirect_jump_via_thunk<mode>")
10221 ("indirect_jump_via_inlinethunk<mode>_z10")
10222 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10223 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10224 ("casesi_jump_via_inlinethunk<mode>_z10")
10225 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10226 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10227 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10228 ("*indirect2_jump"): Disable for branch conversion.
10229 ("casesi_jump"): Turn into expander and expand patterns for branch
10231 ("return_use"): New expander.
10232 ("*return"): Emit return via thunk and rename it to ...
10233 ("*return<mode>"): ... this one.
10234 * config/s390/s390.opt: Add new options and and enum for the
10237 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10239 * lra-constraints.c (match_reload): Unconditionally use
10240 gen_lowpart_SUBREG, rather than selecting between that
10241 and equivalent gen_rtx_SUBREG code.
10243 2018-02-08 Richard Biener <rguenther@suse.de>
10245 PR tree-optimization/84233
10246 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10247 changed flag instead of boguously re-using phi_inserted.
10249 2018-02-08 Martin Jambor <mjambor@suse.cz>
10251 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10252 static local variables.
10254 2018-02-08 Richard Biener <rguenther@suse.de>
10256 PR tree-optimization/84278
10257 * tree-vect-stmts.c (vectorizable_store): When looking for
10258 smaller vector types to perform grouped strided loads/stores
10259 make sure the mode is supported by the target.
10260 (vectorizable_load): Likewise.
10262 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10264 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10265 Increase LDP/STP opportunities by adding adjacent callee-saves.
10267 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10269 PR rtl-optimization/84068
10270 PR rtl-optimization/83459
10271 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10273 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10275 PR tree-optimization/84224
10276 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10277 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10278 non-zero arguments.
10280 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10283 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10284 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10285 count, remove one USE.
10287 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10289 * doc/install.texi (Configuration): Document the
10290 --with-long-double-format={ibm,ieee} PowerPC configuration
10294 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10295 Convert from define_expand to be define_insn_and_split. Rework
10296 float/double/_Float128 conversions to QI/HI/SImode to work with
10297 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10298 conversions to QI/HImode types did a store and then a load to
10299 truncate the value. For conversions to VSX registers, don't split
10300 the insn, instead emit the code directly. Use the code iterator
10301 any_fix to combine signed and unsigned conversions.
10302 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10303 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10304 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10305 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10306 (fix_<mode>di2_hw): Likewise.
10307 (fixuns_<mode>di2_hw): Likewise.
10308 (fix_<mode>si2_hw): Likewise.
10309 (fixuns_<mode>si2_hw): Likewise.
10310 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10311 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10312 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10313 fix<uns>_trunc<SFDF:mode>si2_p8.
10314 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10315 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10316 (fix<uns>_<mode>_mem): Likewise.
10317 (fctiw<u>z_<mode>_mem): Likewise.
10318 (fix<uns>_<mode>_mem): Likewise.
10319 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10320 the register allocator from doing a direct move to the GPRs to do
10321 a store, and instead use the ISA 3.0 store byte/half-word from
10322 vector register instruction. For IEEE 128-bit floating point,
10323 also optimize stores of 32-bit ints.
10324 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10326 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10328 * genextract.c (push_pathstr_operand): New function to support
10330 (walk_rtx): Call push_pathstr_operand.
10331 (print_path): Support [a-zA-Z].
10333 2018-02-07 Richard Biener <rguenther@suse.de>
10335 PR tree-optimization/84037
10336 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10337 (cse_and_gimplify_to_preheader): Declare.
10338 (vect_get_place_in_interleaving_chain): Likewise.
10339 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10341 (_loop_vec_info::~_loop_vec_info): Delete it.
10342 (cse_and_gimplify_to_preheader): New function.
10343 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10344 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10345 (vectorizable_load): Likewise. For grouped stores always base
10346 the IV on the first element.
10347 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10348 condition before gimplifying.
10350 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10352 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10353 *DIV_EXPR and *MOD_EXPR.
10355 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10358 * config/i386/i386.c (ix86_option_override_internal): Mask out
10359 the CF_SET bit when checking -fcf-protection.
10361 2018-02-07 Tom de Vries <tom@codesourcery.com>
10364 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10367 2018-02-07 Richard Biener <rguenther@suse.de>
10369 PR tree-optimization/84204
10370 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10373 PR tree-optimization/84205
10374 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10375 special-case isl_ast_op_zdiv_r.
10377 PR tree-optimization/84223
10378 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10379 Only add conditions from within the region.
10380 (gather_bbs::after_dom_children): Adjust.
10382 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10385 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10386 * config/avr/avr.md: Only post-reload split REG-REG moves if
10387 either register is GENERAL_REG_P.
10389 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10391 PR tree-optimization/84235
10392 * tree-ssa-scopedtables.c
10393 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10394 if the subtraction is performed in floating point type where NaNs are
10395 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10396 build 1. Formatting fix.
10398 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10401 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10402 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10403 and skip it regardless of bb boundaries. Use CALL_P macro,
10404 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10405 unnecessarily, formatting fix.
10407 2018-02-06 Michael Collison <michael.collison@arm.com>
10409 * config/arm/thumb2.md:
10410 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10411 (*thumb_mov_notscc): Ditto.
10413 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10416 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10417 unsigned_fix, not "s".
10419 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10421 * configure.ac (gcc_fn_eh_frame_ro): New function.
10422 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10423 correct .eh_frame permissions.
10424 * configure: Regenerate.
10426 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10428 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10429 irrelevant options.
10431 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10433 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10434 Display warning message for -mno-speculate-indirect-jumps.
10436 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10438 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10440 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10442 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10444 PR tree-optimization/84225
10445 * tree-eh.c (find_trapping_overflow): Only call
10446 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10448 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10451 * config/i386/i386.c: Reimplement the check of possible options
10452 -mibt/-mshstk conbination. Change error messages.
10453 * doc/invoke.texi: Fix a typo: remove extra '='.
10455 2018-02-06 Marek Polacek <polacek@redhat.com>
10457 PR tree-optimization/84228
10458 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10460 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10463 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10464 emitted arch directives.
10465 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10466 __ARM_FEATURE_COPROC before changing architectures.
10468 2018-02-06 Richard Biener <rguenther@suse.de>
10470 * config/i386/i386.c (print_reg): Fix typo.
10471 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10473 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10475 * configure: Regenerate.
10477 2018-02-05 Martin Sebor <msebor@redhat.com>
10479 PR tree-optimization/83369
10480 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10483 2018-02-05 Martin Liska <mliska@suse.cz>
10485 * doc/invoke.texi: Cherry-pick upstream r323995.
10487 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10489 * ira.c (ira_init_register_move_cost): Adjust comment.
10491 2018-02-05 Martin Liska <mliska@suse.cz>
10493 PR gcov-profile/84137
10494 * doc/gcov.texi: Fix typo in documentation.
10496 2018-02-05 Martin Liska <mliska@suse.cz>
10498 PR gcov-profile/83879
10499 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10500 using dlopen functionality.
10502 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10504 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10505 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10506 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10507 _mm_maskz_range_round_ss): New intrinsics.
10508 (__builtin_ia32_rangesd128_round)
10509 (__builtin_ia32_rangess128_round): Remove.
10510 (__builtin_ia32_rangesd128_mask_round,
10511 __builtin_ia32_rangess128_mask_round): New builtins.
10512 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10513 __builtin_ia32_rangess128_round): Remove.
10514 (__builtin_ia32_rangesd128_mask_round,
10515 __builtin_ia32_rangess128_mask_round): New builtins.
10516 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10517 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10518 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10519 "<round_saeonly_constraint>")): Changed to ...
10520 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10521 "<round_saeonly_scalar_constraint>")): ... this.
10522 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10523 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10524 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10525 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10526 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10528 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10530 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10532 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10533 Remove all values except native, 8540 and 8548.
10535 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10537 * config/i386/i386.c (ix86_output_function_return): Pass
10538 INVALID_REGNUM, instead of -1, as invalid register number to
10539 indirect_thunk_name and output_indirect_thunk.
10541 2018-02-02 Julia Koval <julia.koval@intel.com>
10543 * config.gcc: Add -march=icelake.
10544 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10545 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10546 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10547 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10548 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10549 (processor_target_table): Add icelake.
10550 (ix86_option_override_internal): Handle new PTAs.
10551 (get_builtin_code_for_version): Handle icelake.
10552 (M_INTEL_COREI7_ICELAKE): New.
10553 (fold_builtin_cpu): Handle icelake.
10554 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10555 * doc/invoke.texi: Add -march=icelake.
10557 2018-02-02 Julia Koval <julia.koval@intel.com>
10559 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10560 to wide_int_bitmask.
10561 * wide-int-bitmask.h: New.
10563 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10566 * config/i386/i386.md: Replace Pmode with word_mode in
10567 builtin_setjmp_setup and builtin_longjmp to support x32.
10569 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10573 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10575 (rs6000_supported_cpu_names): New static variable.
10576 (linux_cpu_translation_table): Likewise.
10577 (elf_platform) <cpu>: Define new static variable and use it.
10578 Translate kernel AT_PLATFORM name to canonical name if needed.
10579 Error if platform name is unknown.
10581 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10584 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10586 2018-02-01 Jeff Law <law@redhat.com>
10589 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10590 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10591 the scratch if RELEASE_VIA_POP is false.
10592 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10593 If we have to save a temporary register, decrement SIZE appropriately.
10594 Pass new arguments to release_scratch_register_on_entry.
10595 (ix86_adjust_stack_and_probe): Likewise.
10596 (ix86_emit_probe_stack_range): Pass new arguments to
10597 release_scratch_register_on_entry.
10599 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10601 PR rtl-optimization/84157
10602 * combine.c (change_zero_ext): Use REG_P predicate in
10603 front of HARD_REGISTER_P predicate.
10605 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10607 * config/avr/avr.c (avr_option_override): Move disabling of
10608 -fdelete-null-pointer-checks to...
10609 * common/config/avr/avr-common.c (avr_option_optimization_table):
10612 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10614 PR tree-optimization/81635
10615 * tree-data-ref.c (split_constant_offset_1): For types that
10616 wrap on overflow, try to use range info to prove that wrapping
10619 2018-02-01 Renlin Li <renlin.li@arm.com>
10622 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10623 TAILCALL_ADDR_REGS.
10624 (aarch64_register_move_cost): Likewise.
10625 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10626 TAILCALL_ADDR_REGS.
10627 (REG_CLASS_NAMES): Likewise.
10628 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10629 TAILCALL_ADDR_REGS. Remove IP registers.
10630 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10632 2018-02-01 Richard Biener <rguenther@suse.de>
10634 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10635 for specifying RPO order and allow NULL for that.
10636 * domwalk.c (dom_walker::dom_walker): Likewise.
10637 (dom_walker::walk): Handle NULL RPO order.
10638 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10640 (rewrite_update_dom_walker): Likewise.
10641 (mark_def_dom_walker): Likewise.
10643 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10645 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10646 (aarch64_maybe_expand_sve_subreg_move): Declare.
10647 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10648 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10650 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10651 that are semantically a reverse operation.
10652 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10653 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10654 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10656 (aarch64_can_change_mode_class): For big-endian, forbid changes
10657 between two SVE modes if they have different element sizes.
10659 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10661 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10662 the TImode handling for big-endian targets.
10664 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10666 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10667 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10669 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10670 Remove BSWAP handing for big-endian targets and use the form of
10671 LD1RQ appropariate for the mode.
10673 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10675 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10676 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10677 duplicated element.
10679 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10682 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10683 check for operands that need to go through aarch64_sve_reload_be.
10685 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10687 PR tree-optimization/81661
10688 PR tree-optimization/84117
10689 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10690 * tree-eh.c: Include gimplify.h.
10691 (find_trapping_overflow, replace_trapping_overflow,
10692 rewrite_to_non_trapping_overflow): New functions.
10693 * tree-vect-loop.c: Include tree-eh.h.
10694 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10695 * tree-data-ref.c: Include tree-eh.h.
10696 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10698 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10700 PR rtl-optimization/84123
10701 * combine.c (change_zero_ext): Check if hard register satisfies
10702 can_change_dest_mode before calling gen_lowpart_SUBREG.
10704 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10707 * ira.c (ira_init_register_move_cost): Remove assert.
10709 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10711 PR rtl-optimization/84071
10712 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10713 * doc/tm.texi: Regenerate.
10715 2018-01-31 Richard Biener <rguenther@suse.de>
10717 PR tree-optimization/84132
10718 * tree-data-ref.c (analyze_miv_subscript): Properly
10719 check whether evolution_function_is_affine_multivariate_p
10720 before calling gcd_of_steps_may_divide_p.
10722 2018-01-31 Julia Koval <julia.koval@intel.com>
10725 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10726 * config/i386/i386.md (rdpid_rex64) New.
10727 (rdpid): Make 32bit only.
10729 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10732 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10733 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10735 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10738 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10740 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10742 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10744 PR rtl-optimization/84071
10745 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10746 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10748 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10750 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10751 (arc_attribute_table): Add 'aux' attribute.
10752 (arc_in_small_data_p): Consider aux like variables.
10753 (arc_is_aux_reg_p): New function.
10754 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10755 (arc_get_aux_arg): New function.
10756 (prepare_move_operands): Handle aux-register access.
10757 (arc_handle_aux_attribute): New function.
10758 * doc/extend.texi (ARC Variable attributes): Add subsection.
10760 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10762 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10763 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10764 (arc_attribute_table): Add 'uncached' attribute.
10765 (arc_print_operand): Print '.di' flag for uncached memory
10767 (arc_in_small_data_p): Do not consider for small data the uncached
10769 (arc_is_uncached_mem_p): New function.
10770 * config/arc/predicates.md (compact_store_memory_operand): Check
10771 for uncached memory accesses.
10772 (nonvol_nonimm_operand): Likewise.
10773 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10775 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10778 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10779 falign-loops=): Add Optimization flag.
10781 2018-01-30 Jeff Law <law@redhat.com>
10784 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10785 INT_REGISTERS_SAVED. Check it prior to calling
10786 get_scratch_register_on_entry.
10787 (ix86_adjust_stack_and_probe): Similarly.
10788 (ix86_emit_probe_stack_range): Similarly.
10789 (ix86_expand_prologue): Corresponding changes.
10791 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10794 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10795 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10797 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10800 * lra-constraints.c (curr_insn_transform): Process AND in the
10803 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10805 PR rtl-optimization/83986
10806 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10807 dependence against last_pending_memory_flush in addition to
10808 pending_jump_insns.
10810 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10812 PR tree-optimization/81611
10813 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10816 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10819 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10822 2018-01-30 Richard Biener <rguenther@suse.de>
10823 Jakub Jelinek <jakub@redhat.com>
10825 PR tree-optimization/84111
10826 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10827 inner loops added during recursion, as they don't have up-to-date
10830 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10833 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10834 (can_inline_edge_by_limits_p): ... here.
10835 (can_early_inline_edge_p, check_callers,
10836 update_caller_keys, update_callee_keys, recursive_inlining,
10837 add_new_edges_to_heap, speculation_useful_p,
10838 inline_small_functions,
10839 inline_small_functions, flatten_function,
10840 inline_to_all_callers_1): Update.
10842 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10844 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10847 2018-01-30 Richard Biener <rguenther@suse.de>
10849 PR tree-optimization/83008
10850 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10851 invariant and constant vector uses in stmts when they need
10852 more than one stmt.
10854 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10857 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10858 * configure: Regenerate.
10860 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10862 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10864 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10865 Use gen_rtx_REG rather than gen_lowpart.
10867 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10869 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10870 rather than 0 when creating partial subregs.
10872 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10874 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10877 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10880 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10881 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10882 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10883 flags. This restores the settings used before the 2017-07-24.
10884 Turning off pre increment/decrement/modify allows IVOPTS to
10885 optimize DF/SF loops where the index is an int.
10887 2018-01-29 Richard Biener <rguenther@suse.de>
10888 Kelvin Nilsen <kelvin@gcc.gnu.org>
10891 * tree-vect-stmts.c (vectorizable_call): Don't call
10892 targetm.vectorize_builtin_md_vectorized_function if callee is
10895 2018-01-22 Carl Love <cel@us.ibm.com>
10897 * doc/extend.tex: Fix typo in second arg in
10898 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10900 2018-01-29 Richard Biener <rguenther@suse.de>
10902 PR tree-optimization/84086
10903 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10904 (flush_ssaname_freelist): When SSA names were released reset
10905 the SCEV hash table.
10907 2018-01-29 Richard Biener <rguenther@suse.de>
10909 PR tree-optimization/84057
10910 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10911 removed paths when removing edges.
10913 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
10915 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
10916 -mfunction-return=@var{choice}.
10918 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
10920 PR diagnostic/84034
10921 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
10922 Handle CR like TAB.
10923 (layout::print_source_line): Likewise.
10924 (test_get_line_width_without_trailing_whitespace): Add test cases.
10926 2018-01-27 Jakub Jelinek <jakub@redhat.com>
10928 PR middle-end/84040
10929 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
10932 2018-01-26 Jim Wilson <jimw@sifive.com>
10934 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
10936 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
10939 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10941 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
10942 and CMP + SUB-immediate -> SUBS.
10944 2018-01-26 Martin Sebor <msebor@redhat.com>
10946 PR tree-optimization/83896
10947 * tree-ssa-strlen.c (get_string_len): Rename...
10948 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
10949 Avoid assuming length is constant.
10950 (handle_char_store): Use HOST_WIDE_INT for string length.
10952 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
10955 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
10956 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
10958 2018-01-26 Richard Biener <rguenther@suse.de>
10960 PR rtl-optimization/84003
10961 * dse.c (record_store): Only record redundant stores when
10962 the earlier store aliases at least all accesses the later one does.
10964 2018-01-26 Jakub Jelinek <jakub@redhat.com>
10966 PR rtl-optimization/83985
10967 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
10968 REG_CFA_RESTORE insns.
10969 (delete_unmarked_insns): Don't ignore separate shrink wrapping
10970 REG_CFA_RESTORE insns here.
10973 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
10974 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
10976 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10978 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
10979 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
10980 (arc_init): Likewise.
10981 (arc_override_options): Likewise.
10982 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
10984 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
10986 * config/arc/arc.h (TARGET_DBNZ): Define.
10987 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
10988 properly set the tune attribute.
10989 (dbnz): Use TARGET_DBNZ guard.
10990 * config/arc/arc.opt (mtune): Add core3 option.
10992 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
10994 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
10995 recognize new pic like addresses.
10996 (arc_delegitimize_address): Clean up.
10998 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11000 * config/arc/arc-arches.def: Option mrf16 valid for all
11002 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11003 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11004 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11005 * config/arc/arc-tables.opt: Regenerate.
11006 * config/arc/arc.c (arc_conditional_register_usage): Handle
11007 reduced register file case.
11008 (arc_file_start): Set must have build attributes.
11009 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11010 mrf16 option value.
11011 * config/arc/arc.opt (mrf16): Add new option.
11012 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11013 * config/arc/genmultilib.awk: Handle new mrf16 option.
11014 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11015 * config/arc/t-multilib: Regenerate.
11016 * doc/invoke.texi (ARC Options): Document mrf16 option.
11018 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11020 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11021 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11022 (arc_attribute_table): Add 'secure_call' attribute.
11023 (arc_print_operand): Print secure call operand.
11024 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11026 (arc_is_secure_call_p): New function. * config/arc/arc.md
11027 (call_i): Add support for sjli instruction.
11028 (call_value_i): Likewise.
11029 * config/arc/constraints.md (Csc): New constraint.
11031 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11032 John Eric Martin <John.Martin@emmicro-us.com>
11034 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11035 * config/arc/arc.c (_arc_jli_section): New struct.
11036 (arc_jli_section): New type.
11037 (rc_jli_sections): New static variable.
11038 (arc_handle_jli_attribute): New function.
11039 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11040 (arc_file_end): New function.
11041 (TARGET_ASM_FILE_END): Define.
11042 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11043 (arc_add_jli_section): New function.
11044 (jli_call_scan): Likewise.
11045 (arc_reorg): Call jli_call_scan.
11046 (arc_output_addsi): Remove 'S' from printing asm operand.
11047 (arc_is_jli_call_p): New function.
11048 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11050 (movhi_insn): Likewise.
11051 (movsi_insn): Likewise.
11052 (movsi_set_cc_insn): Likewise.
11053 (loadqi_update): Likewise.
11054 (load_zeroextendqisi_update): Likewise.
11055 (load_signextendqisi_update): Likewise.
11056 (loadhi_update): Likewise.
11057 (load_zeroextendhisi_update): Likewise.
11058 (load_signextendhisi_update): Likewise.
11059 (loadsi_update): Likewise.
11060 (loadsf_update): Likewise.
11061 (movsicc_insn): Likewise.
11062 (bset_insn): Likewise.
11063 (bxor_insn): Likewise.
11064 (bclr_insn): Likewise.
11065 (bmsk_insn): Likewise.
11066 (bicsi3_insn): Likewise.
11067 (cmpsi_cc_c_insn): Likewise.
11068 (movsi_ne): Likewise.
11069 (movsi_cond_exec): Likewise.
11070 (clrsbsi2): Likewise.
11071 (norm_f): Likewise.
11081 (call_i): Remove 'S' asm letter, add jli instruction.
11082 (call_value_i): Likewise.
11083 * config/arc/arc.op (mjli-always): New option.
11084 * config/arc/constraints.md (Cji): New constraint.
11085 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11087 (subsf3_fpx): Likewise.
11088 (mulsf3_fpx): Likewise.
11089 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11091 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11092 function attrbutes.
11093 * doc/invoke.texi (ARC): Document mjli-always option.
11095 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11097 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11098 avoid addition with 0 and use incw and decw where possible.
11100 2018-01-26 Richard Biener <rguenther@suse.de>
11102 PR tree-optimization/81082
11103 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11104 association if it requires casting to unsigned.
11105 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11106 from fold_plusminus_mult_expr to catch important cases late when
11107 range info is available.
11109 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11111 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11112 * configure.ac (hidden_linkonce): New test.
11113 * configure: Regenerate.
11114 * config.in: Regenerate.
11116 2018-01-26 Julia Koval <julia.koval@intel.com>
11118 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11119 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11120 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11121 _mm_mask_bitshuffle_epi64_mask): Fix type.
11122 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11123 USI_FTYPE_V4DI_V4DI_USI): Remove.
11124 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11125 __builtin_ia32_vpshufbitqmb256_mask,
11126 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11127 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11128 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11130 2018-01-26 Alan Modra <amodra@gmail.com>
11133 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11134 UNSPEC_VBPERMQ. Sort other unspecs.
11136 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11138 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11140 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11142 PR middle-end/83055
11143 * predict.c (drop_profile): Do not push/pop cfun; update also
11145 (handle_missing_profiles): Fix logic looking for zero profiles.
11147 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11149 PR middle-end/83977
11150 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11151 on functions with #pragma omp declare simd or functions with simd
11153 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11154 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11155 Remove trailing \n from warning_at calls.
11157 2018-01-25 Tom de Vries <tom@codesourcery.com>
11160 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11161 for neutered workers.
11163 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11166 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11167 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11169 2018-01-24 Jeff Law <law@redhat.com>
11172 * i386.c (get_probe_interval): Move to earlier point.
11173 (ix86_compute_frame_layout): If -fstack-clash-protection and
11174 the frame is larger than the probe interval, then use pushes
11175 to save registers rather than reg->mem moves.
11176 (ix86_expand_prologue): Remove conditional for int_registers_saved
11179 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11182 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11183 min/max for never referenced object.
11185 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11187 PR middle-end/83977
11188 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11190 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11191 attributes from DECL_ATTRIBUTES (decl) without affecting
11192 DECL_ATTRIBUTES (current_function_decl).
11193 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11194 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11196 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11198 PR tree-optimization/83979
11199 * fold-const.c (fold_comparison): Use constant_boolean_node
11200 instead of boolean_{true,false}_node.
11202 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11204 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11207 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11209 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11210 Simplify the clause that sets the length attribute.
11211 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11212 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11213 clause that sets the length attribute.
11214 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11216 2018-01-24 Tom de Vries <tom@codesourcery.com>
11219 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11220 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11221 Add strict parameter.
11222 (prevent_branch_around_nothing): Insert dummy insn between branch to
11223 label and label with no ptx insn inbetween.
11224 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11226 2018-01-24 Tom de Vries <tom@codesourcery.com>
11229 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11230 for neutered threads in warp.
11231 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11233 2018-01-24 Richard Biener <rguenther@suse.de>
11235 PR tree-optimization/83176
11236 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11239 2018-01-24 Richard Biener <rguenther@suse.de>
11241 PR tree-optimization/82819
11242 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11243 code generating pluses that are no-ops in the target precision.
11245 2018-01-24 Richard Biener <rguenther@suse.de>
11247 PR middle-end/84000
11248 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11250 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11252 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11253 to merge probabilities.
11254 * predict.c (probably_never_executed): Also mark as cold functions
11255 with global 0 profile and guessed local profile.
11256 * profile-count.c (profile_probability::combine_with_count): New
11258 * profile-count.h (profile_probability::operator*,
11259 profile_probability::operator*=, profile_probability::operator/,
11260 profile_probability::operator/=): Reduce precision to adjusted
11261 and set value to guessed on contradictory divisions.
11262 (profile_probability::combine_with_freq): Remove.
11263 (profile_probability::combine_wiht_count): Declare.
11264 (profile_count::force_nonzero):: Set to adjusted.
11265 (profile_count::probability_in):: Set quality to adjusted.
11266 * tree-ssa-tail-merge.c (replace_block_by): Use
11267 combine_with_count.
11269 2018-01-23 Andrew Waterman <andrew@sifive.com>
11270 Jim Wilson <jimw@sifive.com>
11272 * config/riscv/riscv.c (riscv_stack_boundary): New.
11273 (riscv_option_override): Set riscv_stack_boundary. Handle
11274 riscv_preferred_stack_boundary_arg.
11275 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11276 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11277 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11278 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11279 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11280 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11282 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11285 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11286 of struct ix86_frame.
11287 (ix86_expand_epilogue): Likewise. Add a local variable for
11288 the reg_save_offset field in struct ix86_frame.
11290 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11292 PR tree-optimization/82604
11293 * tree-loop-distribution.c (enum partition_kind): New enum item
11294 PKIND_PARTIAL_MEMSET.
11295 (partition_builtin_p): Support above new enum item.
11296 (generate_code_for_partition): Ditto.
11297 (compute_access_range): Differentiate cases that equality can be
11298 proven at all loops, the innermost loops or no loops.
11299 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11300 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11301 (finalize_partitions, distribute_loop): Don't fuse partition of
11302 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11303 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11304 parloop is enabled.
11306 2018-01-23 Martin Liska <mliska@suse.cz>
11308 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11309 order to ignore the predictor.
11310 (PRED_POLYMORPHIC_CALL): Likewise.
11311 (PRED_RECURSIVE_CALL): Likewise.
11313 2018-01-23 Martin Liska <mliska@suse.cz>
11315 * tree-profile.c (tree_profiling): Print function header to
11316 aware reader which function we are working on.
11317 * value-prof.c (gimple_find_values_to_profile): Do not print
11318 not interesting value histograms.
11320 2018-01-23 Martin Liska <mliska@suse.cz>
11322 * profile-count.h (enum profile_quality): Add
11323 profile_uninitialized as the first value. Do not number values
11324 as they are zero based.
11325 (profile_count::verify): Update sanity check.
11326 (profile_probability::verify): Likewise.
11328 2018-01-23 Nathan Sidwell <nathan@acm.org>
11330 * doc/invoke.texi (ffor-scope): Deprecate.
11332 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11334 PR tree-optimization/83510
11335 * domwalk.c (set_all_edges_as_executable): New function.
11336 (dom_walker::dom_walker): Convert bool param
11337 "skip_unreachable_blocks" to enum reachability. Move setup of
11338 edge flags to set_all_edges_as_executable and only do it when
11339 reachability is REACHABLE_BLOCKS.
11340 * domwalk.h (enum dom_walker::reachability): New enum.
11341 (dom_walker::dom_walker): Convert bool param
11342 "skip_unreachable_blocks" to enum reachability.
11343 (set_all_edges_as_executable): New decl.
11344 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11345 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11347 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11348 but converting true to REACHABLE_BLOCKS.
11349 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11351 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11352 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11353 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11355 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11356 if check_all_array_refs will be called.
11358 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11360 * tree.c (selftest::test_location_wrappers): Add more test
11363 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11365 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11366 (selftest::test_bit_in_range): Likewise.
11368 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11371 * doc/sourcebuild.texi (vect_float): Say that the selector
11372 only describes the situation when -funsafe-math-optimizations is on.
11373 (vect_float_strict): Document.
11375 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11377 PR tree-optimization/83965
11378 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11379 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11380 instead of checking only for a reduction.
11381 (vect_recog_widen_sum_pattern): Likewise.
11383 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11385 * predict.c (probably_never_executed): Only use precise profile info.
11386 (compute_function_frequency): Skip after inlining hack since we now
11387 have quality checking.
11389 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11391 * profile-count.h (profile_probability::very_unlikely,
11392 profile_probability::unlikely, profile_probability::even): Set
11393 precision to guessed.
11395 2018-01-23 Richard Biener <rguenther@suse.de>
11397 PR tree-optimization/83963
11398 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11399 Properly terminate dominator walk when crossing the exit edge not
11400 when visiting its source block.
11402 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11405 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11406 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11408 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11410 PR tree-optimization/83957
11411 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11412 semicolon after for body surrounded by braces.
11414 PR tree-optimization/83081
11415 * profile-count.h (profile_probability::split): New method.
11416 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11417 Use profile_probability::split.
11418 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11419 when splitting a single conditional jump into 2.
11421 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11423 PR tree-optimization/69452
11424 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11427 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11429 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11430 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11431 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11433 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11435 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11437 * config/rl78/rl78.md (movdi): New define_expand.
11438 * config/rl78/rl78.c (rl78_split_movdi): New function.
11440 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11443 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11445 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11446 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11447 128-bit to produce an UNSPEC move to get the double word with the
11448 signbit and then a shift directly to do signbit.
11449 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11450 implementation with a new version that just does either a direct
11451 move or a regular move. Move memory interface to separate insns.
11452 Move insns so they are next to the expander.
11453 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11454 with signbit move. Split big and little endian case.
11455 (signbit<mode>2_dm_mem_le): Likewise.
11456 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11457 (signbit<mode>2_dm2): Likewise.
11459 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11461 * config/rl78/rl78.md (anddi3): New define_expand.
11463 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11465 * config/rl78/rl78.md (umindi3): New define_expand.
11467 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11469 * config/rl78/rl78.md (smindi3): New define_expand.
11471 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11473 * config/rl78/rl78.md (smaxdi3): New define_expand.
11475 2018-01-22 Carl Love <cel@us.ibm.com>
11477 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11478 LVX_V1TI): Add macro expansion.
11479 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11480 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11481 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11482 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11483 Change check to determine if the instruction is a byte reversing
11484 entry. Fix typo in comment.
11485 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11486 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11487 Add def_builtin calls for new builtins.
11488 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11489 Add define_insn expansion.
11491 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11493 * config/rl78/rl78.md (umaxdi3): New define_expand.
11495 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11497 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11498 for non-QImode registers.
11500 2018-01-22 Richard Biener <rguenther@suse.de>
11502 PR tree-optimization/83963
11503 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11504 including the loop exit block.
11505 (scop_detection::merge_sese): Likewise.
11506 (scop_detection::add_scop): Do it here instead.
11508 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11510 * doc/sourcebuild.texi (arm_softfloat): Document.
11512 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11515 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11516 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11517 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11519 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11520 David Edelsohn <dje.gcc@gmail.com>
11523 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11524 Change "crset eq" to "crset 2".
11525 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11526 (*call_indirect_aix<mode>_nospec): Likewise.
11527 (*call_value_indirect_aix<mode>_nospec): Likewise.
11528 (*call_indirect_elfv2<mode>_nospec): Likewise.
11529 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11530 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11531 change assembly output from . to $.
11532 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11533 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11534 (*tablejump<mode>_internal1_nospec): Likewise.
11536 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11539 * config/sh/sh_optimize_sett_clrt.cc:
11540 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11542 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11544 PR tree-optimization/83940
11545 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11546 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11547 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11548 use it to pass back the definition type.
11549 (vect_check_store_rhs): Likewise rhs_dt_out.
11550 (vect_build_gather_load_calls): Add a mask_dt argument and use
11551 it instead of a call to vect_is_simple_use.
11552 (vectorizable_store): Update calls to vect_check_load_store_mask
11553 and vect_check_store_rhs. Use the dt returned by the latter instead
11554 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11555 instead of calls to vect_is_simple_use. Pass the scalar rather
11556 than the vector operand to vect_is_simple_use when handling
11557 second and subsequent copies of an rhs value.
11558 (vectorizable_load): Update calls to vect_check_load_store_mask
11559 and vect_build_gather_load_calls. Use the cached mask_dt and
11560 gs_info.offset_dt instead of calls to vect_is_simple_use.
11562 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11564 PR middle-end/83945
11565 * tree-emutls.c: Include gimplify.h.
11566 (lower_emutls_2): New function.
11567 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11568 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11569 it before further processing.
11572 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11573 UINTVAL (trueop1) instead of INTVAL (op1).
11575 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11579 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11580 INCOMING_FRAME_SP_OFFSET if not defined.
11581 (scan_trace): Add ENTRY argument. If true and
11582 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11583 emit a note to adjust the CFA offset.
11584 (create_cfi_notes): Adjust scan_trace callers.
11585 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11586 INCOMING_FRAME_SP_OFFSET in the CIE.
11587 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11588 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11590 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11591 * doc/tm.texi: Regenerated.
11593 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11595 PR rtl-optimization/83147
11596 * lra-constraints.c (remove_inheritance_pseudos): Use
11597 lra_substitute_pseudo_within_insn.
11599 2018-01-19 Tom de Vries <tom@codesourcery.com>
11600 Cesar Philippidis <cesar@codesourcery.com>
11603 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11605 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11608 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11609 spaces for function labels.
11611 2018-01-19 Martin Liska <mliska@suse.cz>
11613 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11614 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11615 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11616 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11617 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11618 (PRED_CONST_RETURN): Change from 69 to 65.
11619 (PRED_NULL_RETURN): Change from 91 to 71.
11620 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11621 (PRED_LOOP_GUARD): Change from 66 to 73.
11623 2018-01-19 Martin Liska <mliska@suse.cz>
11625 * predict.c (predict_insn_def): Add new assert.
11626 (struct branch_predictor): Change type to signed integer.
11627 (test_prediction_value_range): Amend test to cover
11628 PROB_UNINITIALIZED.
11629 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11630 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11631 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11632 (PRED_LOOP_IV_COMPARE): Likewise.
11633 * predict.h (PROB_UNINITIALIZED): Define new constant.
11635 2018-01-19 Martin Liska <mliska@suse.cz>
11637 * predict.c (dump_prediction): Add new format for
11638 analyze_brprob.py script which is enabled with -details
11640 * profile-count.h (precise_p): New function.
11642 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11644 PR tree-optimization/83922
11645 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11646 there are no statements that need masking.
11647 (vect_active_double_reduction_p): New function.
11648 (vect_analyze_loop_operations): Use it when handling phis that
11649 are not in the loop header.
11651 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11653 PR tree-optimization/83914
11654 * tree-vect-loop.c (vectorizable_induction): Don't convert
11655 init_expr or apply the peeling adjustment for inductions
11656 that are nested within the vectorized loop.
11658 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11660 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11663 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11667 * function.h (gimplify_parameters): Add gimple_seq * argument.
11668 * function.c: Include gimple.h and options.h.
11669 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11670 for the added local temporaries if needed.
11671 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11672 if there are any parameter cleanups, wrap whole body into a
11673 try/finally with the cleanups.
11675 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11678 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11679 Use GET_MODE_CLASS for scalar floating point.
11681 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11685 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11686 Fix call of call_cgraph_insertion_hooks.
11688 2018-01-18 Martin Sebor <msebor@redhat.com>
11690 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11692 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11695 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11698 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11701 * common.opt: (-ffile-prefix-map): New option.
11702 * opts.c (common_handle_option): Defer it.
11703 * opts-global.c (handle_common_deferred_options): Handle it.
11704 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11705 * file-prefix-map.h: New file.
11706 (remap_debug_filename, add_debug_prefix_map): ...here.
11707 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11708 * final.c (debug_prefix_map, add_debug_prefix_map
11709 remap_debug_filename): Move to...
11710 * file-prefix-map.c: New file.
11711 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11712 generalize, get rid of alloca(), use strrchr() instead of strchr().
11713 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11714 Implement in terms of add_prefix_map().
11715 (remap_macro_filename, remap_debug_filename): Implement in term of
11717 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11718 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11719 * dbxout.c: Include file-prefix-map.h.
11720 * varasm.c: Likewise.
11721 * vmsdbgout.c: Likewise.
11722 * xcoffout.c: Likewise.
11723 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11724 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11725 * doc/invoke.texi (-ffile-prefix-map): Document.
11726 (-fdebug-prefix-map): Update description.
11728 2018-01-18 Martin Liska <mliska@suse.cz>
11730 * config/i386/i386.c (indirect_thunk_name): Document that also
11732 (output_indirect_thunk): Document why both instructions
11733 (pause and lfence) are generated.
11735 2018-01-18 Richard Biener <rguenther@suse.de>
11737 PR tree-optimization/83887
11738 * graphite-scop-detection.c
11739 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11740 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11741 (scop_detection::merge_sese): Re-implement with a flood-fill
11742 algorithm that properly finds a SESE region if it exists.
11744 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11747 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11748 pointer_diff optimizations use view_convert instead of convert.
11750 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11752 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11753 Generate different code for -mno-speculate-indirect-jumps.
11754 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11755 (*call_indirect_aix<mode>): Disable for
11756 -mno-speculate-indirect-jumps.
11757 (*call_indirect_aix<mode>_nospec): New define_insn.
11758 (*call_value_indirect_aix<mode>): Disable for
11759 -mno-speculate-indirect-jumps.
11760 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11761 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11762 -mno-speculate-indirect-jumps.
11763 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11765 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11767 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11768 long double type, set the flags for noting the default long double
11769 type, even if we don't pass or return a long double type.
11771 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11774 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11777 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11779 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11780 support for merge[hl].
11781 (fold_mergehl_helper): New helper function.
11782 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11783 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11784 (altivec_vmrglw_direct): Add xxmrglw insn.
11786 2018-01-17 Andrew Waterman <andrew@sifive.com>
11788 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11789 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11791 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11794 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11795 call the lto_location_cache before reading the
11796 DECL_SOURCE_LOCATION of the types.
11798 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11799 Richard Sandiford <richard.sandiford@linaro.org>
11801 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11802 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11803 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11804 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11805 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11807 * config/aarch64/constraints.md (aarch64_movti_operand):
11809 * config/aarch64/predicates.md (Uti): Add new constraint.
11811 2018-01-17 Carl Love <cel@us.ibm.com>
11813 * config/rs6000/vsx.md (define_expand xl_len_r,
11814 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11815 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11817 (define_expand, define_insn): Move the shift left from the
11818 define_insn to the define_expand for lxvl and stxvl instructions.
11819 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11820 and XL_LEN_R definitions to PURE.
11822 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11824 * config/i386/i386.c (indirect_thunk_name): Declare regno
11825 as unsigned int. Compare regno with INVALID_REGNUM.
11826 (output_indirect_thunk): Ditto.
11827 (output_indirect_thunk_function): Ditto.
11828 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11829 in the call to output_indirect_thunk_function.
11831 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11833 PR middle-end/83884
11834 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11835 rather than the size of inner_type to determine the stack slot size
11836 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11838 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11841 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11844 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11846 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11847 endian Linux systems to optionally enable multilibs for selecting
11848 the long double type if the user configured an explicit type.
11849 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11850 have no long double multilibs if not defined.
11851 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11852 warn if the user used -mabi={ieee,ibm}longdouble and we built
11853 multilibs for long double.
11854 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11855 appropriate multilib option.
11856 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11858 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11859 for building long double multilibs.
11860 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11862 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11864 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11867 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11869 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11872 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11875 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11878 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11880 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11881 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11883 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11885 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11886 different rtl trees depending on TARGET_64BIT.
11887 (rs6000_gen_lvx): Likewise.
11889 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11891 * config/visium/visium.md (nop): Tweak comment.
11892 (hazard_nop): Likewise.
11894 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11896 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11897 -mspeculate-indirect-jumps.
11898 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11899 for -mno-speculate-indirect-jumps.
11900 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11901 (*call_value_indirect_elfv2<mode>): Disable for
11902 -mno-speculate-indirect-jumps.
11903 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11904 (indirect_jump): Emit different RTL for
11905 -mno-speculate-indirect-jumps.
11906 (*indirect_jump<mode>): Disable for
11907 -mno-speculate-indirect-jumps.
11908 (*indirect_jump<mode>_nospec): New define_insn.
11909 (tablejump): Emit different RTL for
11910 -mno-speculate-indirect-jumps.
11911 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
11912 (tablejumpsi_nospec): New define_expand.
11913 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
11914 (tablejumpdi_nospec): New define_expand.
11915 (*tablejump<mode>_internal1): Disable for
11916 -mno-speculate-indirect-jumps.
11917 (*tablejump<mode>_internal1_nospec): New define_insn.
11918 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
11921 2018-01-16 Artyom Skrobov tyomitch@gmail.com
11923 * caller-save.c (insert_save): Drop unnecessary parameter. All
11926 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11927 Richard Biener <rguenth@suse.de>
11930 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
11931 return early, inline manually is_gimple_sizepos. Make sure if we
11932 call gimplify_expr we don't end up with a gimple constant.
11933 * tree.c (variably_modified_type_p): Don't return true for
11934 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
11935 * gimplify.h (is_gimple_sizepos): Remove.
11937 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11939 PR tree-optimization/83857
11940 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
11941 vectorizable_live_operation for pure SLP statements.
11942 (vectorizable_live_operation): Handle PHIs.
11944 2018-01-16 Richard Biener <rguenther@suse.de>
11946 PR tree-optimization/83867
11947 * tree-vect-stmts.c (vect_transform_stmt): Precompute
11948 nested_in_vect_loop_p since the scalar stmt may get invalidated.
11950 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11953 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
11954 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
11955 If off is not INTEGER_CST, issue a may not be aligned warning
11956 rather than isn't aligned. Use isn%'t rather than isn't.
11957 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
11959 <case MULT_EXPR>: Improve the case when bottom and one of the
11960 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
11961 operand, in that case check if the other operand is multiple of
11962 bottom divided by the INTEGER_CST operand.
11964 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11967 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
11968 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
11969 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
11970 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
11971 * config/pa/pa.c (pa_function_arg_advance): Likewise.
11972 (pa_function_arg, pa_arg_partial_bytes): Likewise.
11973 (pa_function_arg_size): New function.
11975 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11977 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
11978 in a separate statement.
11980 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
11982 PR tree-optimization/83847
11983 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
11984 group gathers and scatters.
11986 2018-01-16 Jakub Jelinek <jakub@redhat.com>
11988 PR rtl-optimization/86620
11989 * params.def (max-sched-ready-insns): Bump minimum value to 1.
11991 PR rtl-optimization/83213
11992 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
11993 to last if both are JUMP_INSNs.
11995 PR tree-optimization/83843
11996 * gimple-ssa-store-merging.c
11997 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
11998 store_immediate_info for bswap/nop orig_stores.
12000 2018-01-15 Andrew Waterman <andrew@sifive.com>
12002 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12004 <UDIV>: Increase cost if !TARGET_DIV.
12006 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12008 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12009 (define_attr "cr_logical_3op"): New.
12010 (cceq_ior_compare): Adjust.
12011 (cceq_ior_compare_complement): Adjust.
12012 (*cceq_rev_compare): Adjust.
12013 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12014 (is_cracked_insn): Adjust.
12015 (insn_must_be_first_in_group): Adjust.
12016 * config/rs6000/40x.md: Adjust.
12017 * config/rs6000/440.md: Adjust.
12018 * config/rs6000/476.md: Adjust.
12019 * config/rs6000/601.md: Adjust.
12020 * config/rs6000/603.md: Adjust.
12021 * config/rs6000/6xx.md: Adjust.
12022 * config/rs6000/7450.md: Adjust.
12023 * config/rs6000/7xx.md: Adjust.
12024 * config/rs6000/8540.md: Adjust.
12025 * config/rs6000/cell.md: Adjust.
12026 * config/rs6000/e300c2c3.md: Adjust.
12027 * config/rs6000/e500mc.md: Adjust.
12028 * config/rs6000/e500mc64.md: Adjust.
12029 * config/rs6000/e5500.md: Adjust.
12030 * config/rs6000/e6500.md: Adjust.
12031 * config/rs6000/mpc.md: Adjust.
12032 * config/rs6000/power4.md: Adjust.
12033 * config/rs6000/power5.md: Adjust.
12034 * config/rs6000/power6.md: Adjust.
12035 * config/rs6000/power7.md: Adjust.
12036 * config/rs6000/power8.md: Adjust.
12037 * config/rs6000/power9.md: Adjust.
12038 * config/rs6000/rs64.md: Adjust.
12039 * config/rs6000/titan.md: Adjust.
12041 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12043 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12044 ix86_indirect_branch_register logic.
12046 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12048 * config/i386/constraints.md (Bs): Update
12049 ix86_indirect_branch_register check. Don't check
12050 ix86_indirect_branch_register with GOT_memory_operand.
12052 * config/i386/predicates.md (GOT_memory_operand): Don't check
12053 ix86_indirect_branch_register here.
12054 (GOT32_symbol_operand): Likewise.
12056 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12058 * config/i386/predicates.md (constant_call_address_operand):
12059 Rewrite ix86_indirect_branch_register logic.
12060 (sibcall_insn_operand): Likewise.
12062 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12064 * config/i386/constraints.md (Bs): Replace
12065 ix86_indirect_branch_thunk_register with
12066 ix86_indirect_branch_register.
12068 * config/i386/i386.md (indirect_jump): Likewise.
12069 (tablejump): Likewise.
12070 (*sibcall_memory): Likewise.
12071 (*sibcall_value_memory): Likewise.
12072 Peepholes of indirect call and jump via memory: Likewise.
12073 * config/i386/i386.opt: Likewise.
12074 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12075 (GOT_memory_operand): Likewise.
12076 (call_insn_operand): Likewise.
12077 (sibcall_insn_operand): Likewise.
12078 (GOT32_symbol_operand): Likewise.
12080 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12082 PR middle-end/83837
12083 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12084 type rather than type addr's type points to.
12085 (expand_omp_atomic_mutex): Likewise.
12086 (expand_omp_atomic): Likewise.
12088 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12091 * config/i386/i386.c (output_indirect_thunk_function): Use
12092 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12093 for __x86_return_thunk.
12095 2018-01-15 Richard Biener <rguenther@suse.de>
12097 PR middle-end/83850
12098 * expmed.c (extract_bit_field_1): Fix typo.
12100 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12103 * config/arm/iterators.md (VF): New mode iterator.
12104 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12105 Remove integer-related logic from pattern.
12106 (neon_vabd<mode>_3): Likewise.
12108 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12110 PR middle-end/82694
12111 * common.opt (fstrict-overflow): No longer an alias.
12112 (fwrapv-pointer): New option.
12113 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12114 also for pointer types based on flag_wrapv_pointer.
12115 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12116 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12117 opts->x_flag_wrapv got set.
12118 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12119 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12120 POINTER_TYPE_OVERFLOW_UNDEFINED.
12121 * match.pd: Likewise in address comparison pattern.
12122 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12124 2018-01-15 Richard Biener <rguenther@suse.de>
12127 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12128 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12129 Reset type names to their identifier if their TYPE_DECL doesn't
12130 have linkage (and thus is used for ODR and devirt).
12131 (save_debug_info_for_decl): Remove.
12132 (save_debug_info_for_type): Likewise.
12133 (add_tree_to_fld_list): Adjust.
12134 * tree-pretty-print.c (dump_generic_node): Make dumping of
12135 type names more robust.
12137 2018-01-15 Richard Biener <rguenther@suse.de>
12139 * BASE-VER: Bump to 8.0.1.
12141 2018-01-14 Martin Sebor <msebor@redhat.com>
12144 * builtins.c (check_access): Avoid warning when the no-warning bit
12147 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12149 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12150 * ira-color (allocno_hard_regs_compare): Likewise.
12152 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12155 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12156 Use .pushsection/.popsection.
12158 2018-01-14 Martin Sebor <msebor@redhat.com>
12161 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12163 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12165 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12166 entry from extra_headers.
12167 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12168 extra_headers, make the list bitwise identical to the i?86-*-* one.
12170 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12172 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12173 -mcmodel=large with -mindirect-branch=thunk,
12174 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12175 -mfunction-return=thunk-extern.
12176 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12177 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12178 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12180 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12182 * config/i386/i386.c (print_reg): Print the name of the full
12183 integer register without '%'.
12184 (ix86_print_operand): Handle 'V'.
12185 * doc/extend.texi: Document 'V' modifier.
12187 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12189 * config/i386/constraints.md (Bs): Disallow memory operand for
12190 -mindirect-branch-register.
12192 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12193 (GOT_memory_operand): Likewise.
12194 (call_insn_operand): Likewise.
12195 (sibcall_insn_operand): Likewise.
12196 (GOT32_symbol_operand): Likewise.
12197 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12198 for -mindirect-branch-register.
12199 (tablejump): Likewise.
12200 (*sibcall_memory): Likewise.
12201 (*sibcall_value_memory): Likewise.
12202 Disallow peepholes of indirect call and jump via memory for
12203 -mindirect-branch-register.
12204 (*call_pop): Replace m with Bw.
12205 (*call_value_pop): Likewise.
12206 (*sibcall_pop_memory): Replace m with Bs.
12207 * config/i386/i386.opt (mindirect-branch-register): New option.
12208 * doc/invoke.texi: Document -mindirect-branch-register option.
12210 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12212 * config/i386/i386-protos.h (ix86_output_function_return): New.
12213 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12214 set function_return_type.
12215 (indirect_thunk_name): Add ret_p to indicate thunk for function
12217 (output_indirect_thunk_function): Pass false to
12218 indirect_thunk_name.
12219 (ix86_output_indirect_branch_via_reg): Likewise.
12220 (ix86_output_indirect_branch_via_push): Likewise.
12221 (output_indirect_thunk_function): Create alias for function
12222 return thunk if regno < 0.
12223 (ix86_output_function_return): New function.
12224 (ix86_handle_fndecl_attribute): Handle function_return.
12225 (ix86_attribute_table): Add function_return.
12226 * config/i386/i386.h (machine_function): Add
12227 function_return_type.
12228 * config/i386/i386.md (simple_return_internal): Use
12229 ix86_output_function_return.
12230 (simple_return_internal_long): Likewise.
12231 * config/i386/i386.opt (mfunction-return=): New option.
12232 (indirect_branch): Mention -mfunction-return=.
12233 * doc/extend.texi: Document function_return function attribute.
12234 * doc/invoke.texi: Document -mfunction-return= option.
12236 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12238 * config/i386/i386-opts.h (indirect_branch): New.
12239 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12240 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12241 with local indirect jump when converting indirect call and jump.
12242 (ix86_set_indirect_branch_type): New.
12243 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12244 (indirectlabelno): New.
12245 (indirect_thunk_needed): Likewise.
12246 (indirect_thunk_bnd_needed): Likewise.
12247 (indirect_thunks_used): Likewise.
12248 (indirect_thunks_bnd_used): Likewise.
12249 (INDIRECT_LABEL): Likewise.
12250 (indirect_thunk_name): Likewise.
12251 (output_indirect_thunk): Likewise.
12252 (output_indirect_thunk_function): Likewise.
12253 (ix86_output_indirect_branch_via_reg): Likewise.
12254 (ix86_output_indirect_branch_via_push): Likewise.
12255 (ix86_output_indirect_branch): Likewise.
12256 (ix86_output_indirect_jmp): Likewise.
12257 (ix86_code_end): Call output_indirect_thunk_function if needed.
12258 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12260 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12261 (ix86_attribute_table): Add indirect_branch.
12262 * config/i386/i386.h (machine_function): Add indirect_branch_type
12263 and has_local_indirect_jump.
12264 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12266 (tablejump): Likewise.
12267 (*indirect_jump): Use ix86_output_indirect_jmp.
12268 (*tablejump_1): Likewise.
12269 (simple_return_indirect_internal): Likewise.
12270 * config/i386/i386.opt (mindirect-branch=): New option.
12271 (indirect_branch): New.
12274 (thunk-inline): Likewise.
12275 (thunk-extern): Likewise.
12276 * doc/extend.texi: Document indirect_branch function attribute.
12277 * doc/invoke.texi: Document -mindirect-branch= option.
12279 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12282 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12284 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12286 * ipa-inline.c (want_inline_small_function_p): Return false if
12287 inlining has already failed with CIF_FINAL_ERROR.
12288 (update_caller_keys): Call want_inline_small_function_p before
12290 (update_callee_keys): Likewise.
12292 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12294 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12296 (rs6000_quadword_masked_address_p): Likewise.
12297 (quad_aligned_load_p): Likewise.
12298 (quad_aligned_store_p): Likewise.
12299 (const_load_sequence_p): Add comment to describe the outer-most loop.
12300 (mimic_memory_attributes_and_flags): New function.
12301 (rs6000_gen_stvx): Likewise.
12302 (replace_swapped_aligned_store): Likewise.
12303 (rs6000_gen_lvx): Likewise.
12304 (replace_swapped_aligned_load): Likewise.
12305 (replace_swapped_load_constant): Capitalize argument name in
12306 comment describing this function.
12307 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12308 and stores that access quad-word aligned addresses and replace
12309 with stvx or lvx instructions when appropriate.
12310 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12311 New function prototype.
12312 (rs6000_quadword_masked_address_p): Likewise.
12313 (rs6000_gen_lvx): Likewise.
12314 (rs6000_gen_stvx): Likewise.
12315 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12316 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12317 when memory address is aligned.
12318 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12319 this split to select lvx instruction when memory address is aligned.
12320 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12321 instruction when memory address is aligned.
12322 (*vsx_le_perm_load_v16qi): Likewise.
12323 (four unnamed splitters): Modify to select the stvx instruction
12324 when memory is aligned.
12326 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12328 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12329 which appears in the queue multiple times.
12331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12332 Alan Hayward <alan.hayward@arm.com>
12333 David Sherwood <david.sherwood@arm.com>
12335 * tree-vectorizer.h (vec_lower_bound): New structure.
12336 (_loop_vec_info): Add check_nonzero and lower_bounds.
12337 (LOOP_VINFO_CHECK_NONZERO): New macro.
12338 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12339 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12340 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12341 fields. Make seg_len the distance travelled, not including the
12343 (dr_direction_indicator): Declare.
12344 (dr_zero_step_indicator): Likewise.
12345 (dr_known_forward_stride_p): Likewise.
12346 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12348 (runtime_alias_check_p): Allow runtime alias checks with
12350 (operator ==): Compare access_size and align.
12351 (prune_runtime_alias_test_list): Rework for new distinction between
12352 the access_size and seg_len.
12353 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12355 (get_segment_min_max): New function.
12356 (create_intersect_range_checks): Use it.
12357 (dr_step_indicator): New function.
12358 (dr_direction_indicator): Likewise.
12359 (dr_zero_step_indicator): Likewise.
12360 (dr_known_forward_stride_p): Likewise.
12361 * tree-loop-distribution.c (data_ref_segment_size): Return
12362 DR_STEP * (niters - 1).
12363 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12365 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12366 (vect_preserves_scalar_order_p): New function, split out from...
12367 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12368 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12369 (vect_vfa_access_size): New function.
12370 (vect_vfa_align): Likewise.
12371 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12372 (dump_lower_bound): New function.
12373 (vect_check_lower_bound): Likewise.
12374 (vect_small_gap_p): Likewise.
12375 (vectorizable_with_step_bound_p): Likewise.
12376 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12377 depencies if the vectorization factor is 1. Convert the checks
12378 for nonzero steps into checks on the bounds of DR_STEP. Try using
12379 a bunds check for variable steps if the minimum required step is
12380 relatively small. Update calls to the dr_with_seg_len
12381 constructor and to vect_compile_time_alias.
12382 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12384 (vect_loop_versioning): Call it.
12385 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12387 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12390 Alan Hayward <alan.hayward@arm.com>
12391 David Sherwood <david.sherwood@arm.com>
12393 * doc/sourcebuild.texi (vect_scatter_store): Document.
12394 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12396 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12398 * genopinit.c (main): Add supports_vec_scatter_store and
12399 supports_vec_scatter_store_cached to target_optabs.
12400 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12401 IFN_MASK_SCATTER_STORE.
12402 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12404 * internal-fn.h (internal_store_fn_p): Declare.
12405 (internal_fn_stored_value_index): Likewise.
12406 * internal-fn.c (scatter_store_direct): New macro.
12407 (expand_scatter_store_optab_fn): New function.
12408 (direct_scatter_store_optab_supported_p): New macro.
12409 (internal_store_fn_p): New function.
12410 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12411 IFN_MASK_SCATTER_STORE.
12412 (internal_fn_mask_index): Likewise.
12413 (internal_fn_stored_value_index): New function.
12414 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12415 for scatter stores.
12416 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12417 * optabs-query.c (supports_vec_scatter_store_p): New function.
12418 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12419 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12420 true for scatter stores.
12421 (vect_gather_scatter_fn_p): Handle scatter stores too.
12422 (vect_check_gather_scatter): Consider using scatter stores if
12423 supports_vec_scatter_store_p.
12424 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12425 scatter stores too.
12426 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12427 internal_fn_stored_value_index.
12428 (check_load_store_masking): Handle scatter stores too.
12429 (vect_get_store_rhs): Make public.
12430 (vectorizable_call): Use internal_store_fn_p.
12431 (vectorizable_store): Handle scatter store internal functions.
12432 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12433 when deciding whether the end of the group has been reached.
12434 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12435 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12436 (mask_scatter_store<mode>): New insns.
12438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12439 Alan Hayward <alan.hayward@arm.com>
12440 David Sherwood <david.sherwood@arm.com>
12442 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12443 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12444 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12446 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12447 Use vect_truncate_gather_scatter_offset if we can't treat the
12448 operation as a normal gather load or scatter store.
12449 (get_group_load_store_type): Take the gather_scatter_info
12450 as argument. Try using a gather load or scatter store for
12451 single-element groups.
12452 (get_load_store_type): Update calls to get_group_load_store_type
12453 and vect_use_strided_gather_scatters_p.
12455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12456 Alan Hayward <alan.hayward@arm.com>
12457 David Sherwood <david.sherwood@arm.com>
12459 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12460 optional tree argument.
12461 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12463 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12464 but continue to use the current value as a fallback.
12465 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12466 to compare the updates.
12467 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12468 (get_load_store_type): Use it when handling a strided access.
12469 (vect_get_strided_load_store_ops): New function.
12470 (vect_get_data_ptr_increment): Likewise.
12471 (vectorizable_load): Handle strided gather loads. Always pass
12472 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12475 Alan Hayward <alan.hayward@arm.com>
12476 David Sherwood <david.sherwood@arm.com>
12478 * doc/md.texi (gather_load@var{m}): Document.
12479 (mask_gather_load@var{m}): Likewise.
12480 * genopinit.c (main): Add supports_vec_gather_load and
12481 supports_vec_gather_load_cached to target_optabs.
12482 * optabs-tree.c (init_tree_optimization_optabs): Use
12483 ggc_cleared_alloc to allocate target_optabs.
12484 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12485 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12487 * internal-fn.h (internal_load_fn_p): Declare.
12488 (internal_gather_scatter_fn_p): Likewise.
12489 (internal_fn_mask_index): Likewise.
12490 (internal_gather_scatter_fn_supported_p): Likewise.
12491 * internal-fn.c (gather_load_direct): New macro.
12492 (expand_gather_load_optab_fn): New function.
12493 (direct_gather_load_optab_supported_p): New macro.
12494 (direct_internal_fn_optab): New function.
12495 (internal_load_fn_p): Likewise.
12496 (internal_gather_scatter_fn_p): Likewise.
12497 (internal_fn_mask_index): Likewise.
12498 (internal_gather_scatter_fn_supported_p): Likewise.
12499 * optabs-query.c (supports_at_least_one_mode_p): New function.
12500 (supports_vec_gather_load_p): Likewise.
12501 * optabs-query.h (supports_vec_gather_load_p): Declare.
12502 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12503 and memory_type field.
12504 (NUM_PATTERNS): Bump to 15.
12505 * tree-vect-data-refs.c: Include internal-fn.h.
12506 (vect_gather_scatter_fn_p): New function.
12507 (vect_describe_gather_scatter_call): Likewise.
12508 (vect_check_gather_scatter): Try using internal functions for
12509 gather loads. Recognize existing calls to a gather load function.
12510 (vect_analyze_data_refs): Consider using gather loads if
12511 supports_vec_gather_load_p.
12512 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12513 (vect_get_gather_scatter_offset_type): Likewise.
12514 (vect_convert_mask_for_vectype): Likewise.
12515 (vect_add_conversion_to_patterm): Likewise.
12516 (vect_try_gather_scatter_pattern): Likewise.
12517 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12518 (vect_vect_recog_func_ptrs): Add it.
12519 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12520 internal_fn_mask_index and internal_gather_scatter_fn_p.
12521 (check_load_store_masking): Take the gather_scatter_info as an
12522 argument and handle gather loads.
12523 (vect_get_gather_scatter_ops): New function.
12524 (vectorizable_call): Check internal_load_fn_p.
12525 (vectorizable_load): Likewise. Handle gather load internal
12527 (vectorizable_store): Update call to check_load_store_masking.
12528 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12529 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12530 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12531 (aarch64_gather_scale_operand_d): New predicates.
12532 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12533 (mask_gather_load<mode>): New insns.
12535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12536 Alan Hayward <alan.hayward@arm.com>
12537 David Sherwood <david.sherwood@arm.com>
12539 * optabs.def (fold_left_plus_optab): New optab.
12540 * doc/md.texi (fold_left_plus_@var{m}): Document.
12541 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12542 * internal-fn.c (fold_left_direct): Define.
12543 (expand_fold_left_optab_fn): Likewise.
12544 (direct_fold_left_optab_supported_p): Likewise.
12545 * fold-const-call.c (fold_const_fold_left): New function.
12546 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12547 * tree-parloops.c (valid_reduction_p): New function.
12548 (gather_scalar_reductions): Use it.
12549 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12550 (vect_finish_replace_stmt): Declare.
12551 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12552 (needs_fold_left_reduction_p): New function, split out from...
12553 (vect_is_simple_reduction): ...here. Accept reductions that
12554 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12555 (vect_force_simple_reduction): Also store the reduction type in
12556 the assignment's STMT_VINFO_REDUC_TYPE.
12557 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12558 (merge_with_identity): New function.
12559 (vect_expand_fold_left): Likewise.
12560 (vectorize_fold_left_reduction): Likewise.
12561 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12562 scalar phi in place for it. Check for target support and reject
12563 cases that would reassociate the operation. Defer the transform
12564 phase to vectorize_fold_left_reduction.
12565 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12566 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12567 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12571 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12572 call to ifc_temp_var.
12574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * target.def (legitimize_address_displacement): Take the original
12579 offset as a poly_int.
12580 * targhooks.h (default_legitimize_address_displacement): Update
12582 * targhooks.c (default_legitimize_address_displacement): Likewise.
12583 * doc/tm.texi: Regenerate.
12584 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12585 as an argument, moving assert of ad->disp == ad->disp_term to...
12586 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12587 Try calling targetm.legitimize_address_displacement before expanding
12588 the address rather than afterwards, and adjust for the new interface.
12589 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12590 Match the new hook interface. Handle SVE addresses.
12591 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12592 new hook interface.
12594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12596 * Makefile.in (OBJS): Add early-remat.o.
12597 * target.def (select_early_remat_modes): New hook.
12598 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12599 * doc/tm.texi: Regenerate.
12600 * targhooks.h (default_select_early_remat_modes): Declare.
12601 * targhooks.c (default_select_early_remat_modes): New function.
12602 * timevar.def (TV_EARLY_REMAT): New timevar.
12603 * passes.def (pass_early_remat): New pass.
12604 * tree-pass.h (make_pass_early_remat): Declare.
12605 * early-remat.c: New file.
12606 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12608 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12611 Alan Hayward <alan.hayward@arm.com>
12612 David Sherwood <david.sherwood@arm.com>
12614 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12615 vfm1 with a bound_epilog parameter.
12616 (vect_do_peeling): Update calls accordingly, and move the prologue
12617 call earlier in the function. Treat the base bound_epilog as 0 for
12618 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12619 this base when peeling for gaps.
12620 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12621 with fully-masked loops.
12622 (vect_estimate_min_profitable_iters): Handle the single peeled
12623 iteration in that case.
12625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12626 Alan Hayward <alan.hayward@arm.com>
12627 David Sherwood <david.sherwood@arm.com>
12629 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12630 single-element interleaving even if the size is not a power of 2.
12631 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12632 accesses for single-element interleaving if the group size is
12635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12636 Alan Hayward <alan.hayward@arm.com>
12637 David Sherwood <david.sherwood@arm.com>
12639 * doc/md.texi (fold_extract_last_@var{m}): Document.
12640 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12641 * optabs.def (fold_extract_last_optab): New optab.
12642 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12643 * internal-fn.c (fold_extract_direct): New macro.
12644 (expand_fold_extract_optab_fn): Likewise.
12645 (direct_fold_extract_optab_supported_p): Likewise.
12646 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12647 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12648 EXTRACT_LAST_REDUCTION.
12649 (get_initial_def_for_reduction): Do not create an initial vector
12650 for EXTRACT_LAST_REDUCTION reductions.
12651 (vectorizable_reduction): Leave the scalar phi in place for
12652 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12653 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12654 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12655 transform phase to vectorizable_condition.
12656 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12658 (vect_finish_stmt_generation): ...here.
12659 (vect_finish_replace_stmt): New function.
12660 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12661 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12663 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12666 Alan Hayward <alan.hayward@arm.com>
12667 David Sherwood <david.sherwood@arm.com>
12669 * doc/md.texi (extract_last_@var{m}): Document.
12670 * optabs.def (extract_last_optab): New optab.
12671 * internal-fn.def (EXTRACT_LAST): New internal function.
12672 * internal-fn.c (cond_unary_direct): New macro.
12673 (expand_cond_unary_optab_fn): Likewise.
12674 (direct_cond_unary_optab_supported_p): Likewise.
12675 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12676 loops using EXTRACT_LAST.
12677 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12678 (extract_last_<mode>): ...this optab.
12679 (vec_extract<mode><Vel>): Update accordingly.
12681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12682 Alan Hayward <alan.hayward@arm.com>
12683 David Sherwood <david.sherwood@arm.com>
12685 * target.def (empty_mask_is_expensive): New hook.
12686 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12687 * doc/tm.texi: Regenerate.
12688 * targhooks.h (default_empty_mask_is_expensive): Declare.
12689 * targhooks.c (default_empty_mask_is_expensive): New function.
12690 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12691 if the target says that empty masks are expensive.
12692 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12694 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12697 Alan Hayward <alan.hayward@arm.com>
12698 David Sherwood <david.sherwood@arm.com>
12700 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12701 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12702 (vect_use_loop_mask_for_alignment_p): New function.
12703 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12704 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12705 niters_skip argument. Make sure that the first niters_skip elements
12706 of the first iteration are inactive.
12707 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12708 Update call to vect_set_loop_masks_directly.
12709 (get_misalign_in_elems): New function, split out from...
12710 (vect_gen_prolog_loop_niters): ...here.
12711 (vect_update_init_of_dr): Take a code argument that specifies whether
12712 the adjustment should be added or subtracted.
12713 (vect_update_init_of_drs): Likewise.
12714 (vect_prepare_for_masked_peels): New function.
12715 (vect_do_peeling): Skip prologue peeling if we're using a mask
12716 instead. Update call to vect_update_inits_of_drs.
12717 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12719 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12720 alignment. Do not include the number of peeled iterations in
12721 the minimum threshold in that case.
12722 (vectorizable_induction): Adjust the start value down by
12723 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12724 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12725 Take the number of skipped iterations into account when calculating
12727 * tree-vect-stmts.c (vect_gen_while_not): New function.
12729 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12730 Alan Hayward <alan.hayward@arm.com>
12731 David Sherwood <david.sherwood@arm.com>
12733 * doc/sourcebuild.texi (vect_fully_masked): Document.
12734 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12735 default value to 0.
12736 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12738 (vect_analyze_loop_2): ...here. Don't check the vectorization
12739 factor against the number of loop iterations if the loop is
12742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12743 Alan Hayward <alan.hayward@arm.com>
12744 David Sherwood <david.sherwood@arm.com>
12746 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12747 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12748 (dump_groups): Update accordingly.
12749 (iv_use::mem_type): New member variable.
12750 (address_p): New function.
12751 (record_use): Add a mem_type argument and initialize the new
12753 (record_group_use): Add a mem_type argument. Use address_p.
12754 Remove obsolete null checks of base_object. Update call to record_use.
12755 (find_interesting_uses_op): Update call to record_group_use.
12756 (find_interesting_uses_cond): Likewise.
12757 (find_interesting_uses_address): Likewise.
12758 (get_mem_type_for_internal_fn): New function.
12759 (find_address_like_use): Likewise.
12760 (find_interesting_uses_stmt): Try find_address_like_use before
12761 calling find_interesting_uses_op.
12762 (addr_offset_valid_p): Use the iv mem_type field as the type
12763 of the addressed memory.
12764 (add_autoinc_candidates): Likewise.
12765 (get_address_cost): Likewise.
12766 (split_small_address_groups_p): Use address_p.
12767 (split_address_groups): Likewise.
12768 (add_iv_candidate_for_use): Likewise.
12769 (autoinc_possible_for_pair): Likewise.
12770 (rewrite_groups): Likewise.
12771 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12772 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12773 (get_alias_ptr_type_for_ptr_address): New function.
12774 (rewrite_use_address): Rewrite address uses in calls that were
12775 identified by find_address_like_use.
12777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12778 Alan Hayward <alan.hayward@arm.com>
12779 David Sherwood <david.sherwood@arm.com>
12781 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12783 * gimple-expr.h (is_gimple_addressable: Likewise.
12784 * gimple-expr.c (is_gimple_address): Likewise.
12785 * internal-fn.c (expand_call_mem_ref): New function.
12786 (expand_mask_load_optab_fn): Use it.
12787 (expand_mask_store_optab_fn): Likewise.
12789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12790 Alan Hayward <alan.hayward@arm.com>
12791 David Sherwood <david.sherwood@arm.com>
12793 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12794 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12795 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12796 (cond_umax@var{mode}): Document.
12797 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12798 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12799 (cond_umin_optab, cond_umax_optab): New optabs.
12800 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12801 (COND_IOR, COND_XOR): New internal functions.
12802 * internal-fn.h (get_conditional_internal_fn): Declare.
12803 * internal-fn.c (cond_binary_direct): New macro.
12804 (expand_cond_binary_optab_fn): Likewise.
12805 (direct_cond_binary_optab_supported_p): Likewise.
12806 (get_conditional_internal_fn): New function.
12807 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12808 Cope with reduction statements that are vectorized as calls rather
12810 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12811 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12812 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12813 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12814 (UNSPEC_COND_EOR): New unspecs.
12815 (optab): Add mappings for them.
12816 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12817 (sve_int_op, sve_fp_op): New int attributes.
12819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12820 Alan Hayward <alan.hayward@arm.com>
12821 David Sherwood <david.sherwood@arm.com>
12823 * optabs.def (while_ult_optab): New optab.
12824 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12825 * internal-fn.def (WHILE_ULT): New internal function.
12826 * internal-fn.h (direct_internal_fn_supported_p): New override
12827 that takes two types as argument.
12828 * internal-fn.c (while_direct): New macro.
12829 (expand_while_optab_fn): New function.
12830 (convert_optab_supported_p): Likewise.
12831 (direct_while_optab_supported_p): New macro.
12832 * wide-int.h (wi::udiv_ceil): New function.
12833 * tree-vectorizer.h (rgroup_masks): New structure.
12834 (vec_loop_masks): New typedef.
12835 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12836 and fully_masked_p.
12837 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12838 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12839 (vect_max_vf): New function.
12840 (slpeel_make_loop_iterate_ntimes): Delete.
12841 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12842 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12843 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12844 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12845 internal-fn.h, stor-layout.h and optabs-query.h.
12846 (vect_set_loop_mask): New function.
12847 (add_preheader_seq): Likewise.
12848 (add_header_seq): Likewise.
12849 (interleave_supported_p): Likewise.
12850 (vect_maybe_permute_loop_masks): Likewise.
12851 (vect_set_loop_masks_directly): Likewise.
12852 (vect_set_loop_condition_masked): Likewise.
12853 (vect_set_loop_condition_unmasked): New function, split out from
12854 slpeel_make_loop_iterate_ntimes.
12855 (slpeel_make_loop_iterate_ntimes): Rename to..
12856 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12857 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12858 (vect_do_peeling): Update call accordingly.
12859 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12861 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12862 mask_compare_type, can_fully_mask_p and fully_masked_p.
12863 (release_vec_loop_masks): New function.
12864 (_loop_vec_info): Use it to free the loop masks.
12865 (can_produce_all_loop_masks_p): New function.
12866 (vect_get_max_nscalars_per_iter): Likewise.
12867 (vect_verify_full_masking): Likewise.
12868 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12869 retries, and free the mask rgroups before retrying. Check loop-wide
12870 reasons for disallowing fully-masked loops. Make the final decision
12871 about whether use a fully-masked loop or not.
12872 (vect_estimate_min_profitable_iters): Do not assume that peeling
12873 for the number of iterations will be needed for fully-masked loops.
12874 (vectorizable_reduction): Disable fully-masked loops.
12875 (vectorizable_live_operation): Likewise.
12876 (vect_halve_mask_nunits): New function.
12877 (vect_double_mask_nunits): Likewise.
12878 (vect_record_loop_mask): Likewise.
12879 (vect_get_loop_mask): Likewise.
12880 (vect_transform_loop): Handle the case in which the final loop
12881 iteration might handle a partial vector. Call vect_set_loop_condition
12882 instead of slpeel_make_loop_iterate_ntimes.
12883 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12884 (check_load_store_masking): New function.
12885 (prepare_load_store_mask): Likewise.
12886 (vectorizable_store): Handle fully-masked loops.
12887 (vectorizable_load): Likewise.
12888 (supportable_widening_operation): Use vect_halve_mask_nunits for
12890 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12891 (vect_gen_while): New function.
12892 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12893 (aarch64_uqdec<mode>): New insn.
12895 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12896 Alan Hayward <alan.hayward@arm.com>
12897 David Sherwood <david.sherwood@arm.com>
12899 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12900 (reduc_xor_scal_optab): New optabs.
12901 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12902 (reduc_xor_scal_@var{m}): Document.
12903 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12904 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12905 internal functions.
12906 * fold-const-call.c (fold_const_call): Handle them.
12907 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12908 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12909 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12910 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12911 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
12912 (UNSPEC_XORV): New unspecs.
12913 (optab): Add entries for them.
12914 (BITWISEV): New int iterator.
12915 (bit_reduc_op): New int attributes.
12917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12918 Alan Hayward <alan.hayward@arm.com>
12919 David Sherwood <david.sherwood@arm.com>
12921 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
12922 * internal-fn.def (VEC_SHL_INSERT): New internal function.
12923 * optabs.def (vec_shl_insert_optab): New optab.
12924 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
12925 (duplicate_and_interleave): Likewise.
12926 * tree-vect-loop.c: Include internal-fn.h.
12927 (neutral_op_for_slp_reduction): New function, split out from
12928 get_initial_defs_for_reduction.
12929 (get_initial_def_for_reduction): Handle option 2 for variable-length
12930 vectors by loading the neutral value into a vector and then shifting
12931 the initial value into element 0.
12932 (get_initial_defs_for_reduction): Replace the code argument with
12933 the neutral value calculated by neutral_op_for_slp_reduction.
12934 Use gimple_build_vector for constant-length vectors.
12935 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
12936 but the first group_size elements have a neutral value.
12937 Use duplicate_and_interleave otherwise.
12938 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
12939 Update call to get_initial_defs_for_reduction. Handle SLP
12940 reductions for variable-length vectors by creating one vector
12941 result for each scalar result, with the elements associated
12942 with other scalar results stubbed out with the neutral value.
12943 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
12944 Require IFN_VEC_SHL_INSERT for double reductions on
12945 variable-length vectors, or SLP reductions that have
12946 a neutral value. Require can_duplicate_and_interleave_p
12947 support for variable-length unchained SLP reductions if there
12948 is no neutral value, such as for MIN/MAX reductions. Also require
12949 the number of vector elements to be a multiple of the number of
12950 SLP statements when doing variable-length unchained SLP reductions.
12951 Update call to vect_create_epilog_for_reduction.
12952 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
12953 and remove initial values.
12954 (duplicate_and_interleave): Make public.
12955 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
12956 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
12958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12959 Alan Hayward <alan.hayward@arm.com>
12960 David Sherwood <david.sherwood@arm.com>
12962 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
12963 (can_duplicate_and_interleave_p): New function.
12964 (vect_get_and_check_slp_defs): Take the vector of statements
12965 rather than just the current one. Remove excess parentheses.
12966 Restriction rejectinon of vect_constant_def and vect_external_def
12967 for variable-length vectors to boolean types, or types for which
12968 can_duplicate_and_interleave_p is false.
12969 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
12970 (duplicate_and_interleave): New function.
12971 (vect_get_constant_vectors): Use gimple_build_vector for
12972 constant-length vectors and suitable variable-length constant
12973 vectors. Use duplicate_and_interleave for other variable-length
12974 vectors. Don't defer the update when inserting new statements.
12976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12977 Alan Hayward <alan.hayward@arm.com>
12978 David Sherwood <david.sherwood@arm.com>
12980 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
12981 min_profitable_iters doesn't go negative.
12983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12984 Alan Hayward <alan.hayward@arm.com>
12985 David Sherwood <david.sherwood@arm.com>
12987 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
12988 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
12989 * optabs.def (vec_mask_load_lanes_optab): New optab.
12990 (vec_mask_store_lanes_optab): Likewise.
12991 * internal-fn.def (MASK_LOAD_LANES): New internal function.
12992 (MASK_STORE_LANES): Likewise.
12993 * internal-fn.c (mask_load_lanes_direct): New macro.
12994 (mask_store_lanes_direct): Likewise.
12995 (expand_mask_load_optab_fn): Handle masked operations.
12996 (expand_mask_load_lanes_optab_fn): New macro.
12997 (expand_mask_store_optab_fn): Handle masked operations.
12998 (expand_mask_store_lanes_optab_fn): New macro.
12999 (direct_mask_load_lanes_optab_supported_p): Likewise.
13000 (direct_mask_store_lanes_optab_supported_p): Likewise.
13001 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13003 (vect_load_lanes_supported): Likewise.
13004 * tree-vect-data-refs.c (strip_conversion): New function.
13005 (can_group_stmts_p): Likewise.
13006 (vect_analyze_data_ref_accesses): Use it instead of checking
13007 for a pair of assignments.
13008 (vect_store_lanes_supported): Take a masked_p parameter.
13009 (vect_load_lanes_supported): Likewise.
13010 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13011 vect_store_lanes_supported and vect_load_lanes_supported.
13012 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13013 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13014 parameter. Don't allow gaps for masked accesses.
13015 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13016 and vect_load_lanes_supported.
13017 (get_load_store_type): Take a masked_p parameter and update
13018 call to get_group_load_store_type.
13019 (vectorizable_store): Update call to get_load_store_type.
13020 Handle IFN_MASK_STORE_LANES.
13021 (vectorizable_load): Update call to get_load_store_type.
13022 Handle IFN_MASK_LOAD_LANES.
13024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13025 Alan Hayward <alan.hayward@arm.com>
13026 David Sherwood <david.sherwood@arm.com>
13028 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13030 * config/aarch64/aarch64-protos.h
13031 (aarch64_sve_struct_memory_operand_p): Declare.
13032 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13033 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13034 (VPRED, vpred): Handle SVE structure modes.
13035 * config/aarch64/constraints.md (Utx): New constraint.
13036 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13037 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13038 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13039 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13040 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13041 structure modes. Split into pieces after RA.
13042 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13043 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13045 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13046 SVE structure modes.
13047 (aarch64_classify_address): Likewise.
13048 (sizetochar): Move earlier in file.
13049 (aarch64_print_operand): Handle SVE register lists.
13050 (aarch64_array_mode): New function.
13051 (aarch64_sve_struct_memory_operand_p): Likewise.
13052 (TARGET_ARRAY_MODE): Redefine.
13054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13055 Alan Hayward <alan.hayward@arm.com>
13056 David Sherwood <david.sherwood@arm.com>
13058 * target.def (array_mode): New target hook.
13059 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13060 * doc/tm.texi: Regenerate.
13061 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13062 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13063 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13064 targetm.array_mode.
13065 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13069 Alan Hayward <alan.hayward@arm.com>
13070 David Sherwood <david.sherwood@arm.com>
13072 * fold-const.c (fold_binary_loc): Check the argument types
13073 rather than the result type when testing for a vector operation.
13075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13077 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13078 * doc/tm.texi: Regenerate.
13080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13081 Alan Hayward <alan.hayward@arm.com>
13082 David Sherwood <david.sherwood@arm.com>
13084 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13085 (sve): Document new AArch64 extension.
13086 * doc/md.texi (w): Extend the description of the AArch64
13087 constraint to include SVE vectors.
13088 (Upl, Upa): Document new AArch64 predicate constraints.
13089 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13091 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13092 (msve-vector-bits=): New option.
13093 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13094 SVE when these are disabled.
13095 (sve): New extension.
13096 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13097 modes. Adjust their number of units based on aarch64_sve_vg.
13098 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13099 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13100 aarch64_addr_query_type.
13101 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13102 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13103 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13104 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13105 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13106 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13107 (aarch64_simd_imm_zero_p): Delete.
13108 (aarch64_check_zero_based_sve_index_immediate): Declare.
13109 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13110 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13111 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13112 (aarch64_sve_float_mul_immediate_p): Likewise.
13113 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13114 rather than an rtx.
13115 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13116 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13117 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13118 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13119 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13120 (aarch64_regmode_natural_size): Likewise.
13121 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13122 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13124 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13125 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13126 for VG and the SVE predicate registers.
13127 (V_ALIASES): Add a "z"-prefixed alias.
13128 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13129 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13130 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13131 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13132 (REG_CLASS_NAMES): Add entries for them.
13133 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13134 and the predicate registers.
13135 (aarch64_sve_vg): Declare.
13136 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13137 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13138 (REGMODE_NATURAL_SIZE): Define.
13139 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13141 * config/aarch64/aarch64.c: Include cfgrtl.h.
13142 (simd_immediate_info): Add a constructor for series vectors,
13143 and an associated step field.
13144 (aarch64_sve_vg): New variable.
13145 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13146 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13147 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13148 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13149 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13150 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13151 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13152 (aarch64_get_mask_mode): New functions.
13153 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13154 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13155 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13156 predicate modes and predicate registers. Explicitly restrict
13157 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13158 to store a vector mode if it is recognized by
13159 aarch64_classify_vector_mode.
13160 (aarch64_regmode_natural_size): New function.
13161 (aarch64_hard_regno_caller_save_mode): Return the original mode
13163 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13164 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13165 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13166 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13168 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13169 does not overlap dest if the function is frame-related. Handle
13171 (aarch64_split_add_offset): New function.
13172 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13173 them aarch64_add_offset.
13174 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13175 and update call to aarch64_sub_sp.
13176 (aarch64_add_cfa_expression): New function.
13177 (aarch64_expand_prologue): Pass extra temporary registers to the
13178 functions above. Handle the case in which we need to emit new
13179 DW_CFA_expressions for registers that were originally saved
13180 relative to the stack pointer, but now have to be expressed
13181 relative to the frame pointer.
13182 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13184 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13185 IP0 and IP1 values for SVE frames.
13186 (aarch64_expand_vec_series): New function.
13187 (aarch64_expand_sve_widened_duplicate): Likewise.
13188 (aarch64_expand_sve_const_vector): Likewise.
13189 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13190 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13191 into the register, rather than emitting a SET directly.
13192 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13193 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13194 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13195 (offset_9bit_signed_scaled_p): New functions.
13196 (aarch64_replicate_bitmask_imm): New function.
13197 (aarch64_bitmask_imm): Use it.
13198 (aarch64_cannot_force_const_mem): Reject expressions involving
13199 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13200 (aarch64_classify_index): Handle SVE indices, by requiring
13201 a plain register index with a scale that matches the element size.
13202 (aarch64_classify_address): Handle SVE addresses. Assert that
13203 the mode of the address is VOIDmode or an integer mode.
13204 Update call to aarch64_classify_symbol.
13205 (aarch64_classify_symbolic_expression): Update call to
13206 aarch64_classify_symbol.
13207 (aarch64_const_vec_all_in_range_p): New function.
13208 (aarch64_print_vector_float_operand): Likewise.
13209 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13210 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13211 and the FP immediates 1.0 and 0.5.
13212 (aarch64_print_address_internal): Handle SVE addresses.
13213 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13214 (aarch64_regno_regclass): Handle predicate registers.
13215 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13217 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13218 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13219 (aarch64_convert_sve_vector_bits): New function.
13220 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13221 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13222 rather than an rtx.
13223 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13224 Handle SVE vector and predicate modes. Accept VL-based constants
13225 that need only one temporary register, and VL offsets that require
13226 no temporary registers.
13227 (aarch64_conditional_register_usage): Mark the predicate registers
13228 as fixed if SVE isn't available.
13229 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13230 Return true for SVE vector and predicate modes.
13231 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13232 rather than an unsigned int. Handle SVE modes.
13233 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13235 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13237 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13238 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13239 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13240 (aarch64_sve_float_mul_immediate_p): New functions.
13241 (aarch64_sve_valid_immediate): New function.
13242 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13243 Explicitly reject structure modes. Check for INDEX constants.
13244 Handle PTRUE and PFALSE constants.
13245 (aarch64_check_zero_based_sve_index_immediate): New function.
13246 (aarch64_simd_imm_zero_p): Delete.
13247 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13248 vector modes. Accept constants in the range of CNT[BHWD].
13249 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13250 ask for an Advanced SIMD mode.
13251 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13252 (aarch64_simd_vector_alignment): Handle SVE predicates.
13253 (aarch64_vectorize_preferred_vector_alignment): New function.
13254 (aarch64_simd_vector_alignment_reachable): Use it instead of
13256 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13257 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13259 (MAX_VECT_LEN): Delete.
13260 (expand_vec_perm_d): Add a vec_flags field.
13261 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13262 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13263 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13265 (aarch64_evpc_rev): Rename to...
13266 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13267 (aarch64_evpc_rev_global): New function.
13268 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13269 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13271 (aarch64_evpc_sve_tbl): New function.
13272 (aarch64_expand_vec_perm_const_1): Update after rename of
13273 aarch64_evpc_rev. Handle SVE permutes too, trying
13274 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13275 than aarch64_evpc_tbl.
13276 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13277 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13278 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13279 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13280 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13281 (aarch64_expand_sve_vcond): New functions.
13282 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13283 of aarch64_vector_mode_p.
13284 (aarch64_dwarf_poly_indeterminate_value): New function.
13285 (aarch64_compute_pressure_classes): Likewise.
13286 (aarch64_can_change_mode_class): Likewise.
13287 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13288 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13289 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13290 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13291 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13292 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13293 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13294 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13296 (Dn, Dl, Dr): Accept const as well as const_vector.
13297 (Dz): Likewise. Compare against CONST0_RTX.
13298 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13299 of "vector" where appropriate.
13300 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13301 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13302 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13303 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13304 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13305 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13306 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13307 (v_int_equiv): Extend to SVE modes.
13308 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13310 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13311 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13312 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13313 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13314 (SVE_COND_FP_CMP): New int iterators.
13315 (perm_hilo): Handle the new unpack unspecs.
13316 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13318 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13319 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13320 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13321 (aarch64_equality_operator, aarch64_constant_vector_operand)
13322 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13323 (aarch64_sve_nonimmediate_operand): Likewise.
13324 (aarch64_sve_general_operand): Likewise.
13325 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13326 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13327 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13328 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13329 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13330 (aarch64_sve_float_arith_immediate): Likewise.
13331 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13332 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13333 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13334 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13335 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13336 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13337 (aarch64_sve_float_arith_operand): Likewise.
13338 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13339 (aarch64_sve_float_mul_operand): Likewise.
13340 (aarch64_sve_vec_perm_operand): Likewise.
13341 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13342 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13343 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13344 as well as const_vector.
13345 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13346 in file. Use CONST0_RTX and CONSTM1_RTX.
13347 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13348 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13349 Use aarch64_simd_imm_zero.
13350 * config/aarch64/aarch64-sve.md: New file.
13351 * config/aarch64/aarch64.md: Include it.
13352 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13353 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13354 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13355 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13356 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13357 (sve): New attribute.
13358 (enabled): Disable instructions with the sve attribute unless
13360 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13361 aarch64_expand_mov_immediate.
13362 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13363 CNT[BHSD] immediates.
13364 (movti): Split CONST_POLY_INT moves into two halves.
13365 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13366 Split additions that need a temporary here if the destination
13367 is the stack pointer.
13368 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13369 (*add<mode>3_poly_1): New instruction.
13370 (set_clobber_cc): New expander.
13372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13374 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13375 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13376 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13377 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13378 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13379 Change innermode from fixed_mode_size to machine_mode.
13380 (simplify_subreg): Update call accordingly. Handle a constant-sized
13381 subreg of a variable-length CONST_VECTOR.
13383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13384 Alan Hayward <alan.hayward@arm.com>
13385 David Sherwood <david.sherwood@arm.com>
13387 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13388 (add_offset_to_base): New function, split out from...
13389 (create_mem_ref): ...here. When handling a scale other than 1,
13390 check first whether the address is valid without the offset.
13391 Add it into the base if so, leaving the index and scale as-is.
13393 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13396 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13397 fold_for_warn before checking if arg2 is INTEGER_CST.
13399 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13401 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13402 (store_multiple_operation): Delete.
13403 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13404 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13405 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13406 guarded by TARGET_STRING.
13407 (rs6000_output_load_multiple): Delete.
13408 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13409 OPTION_MASK_STRING / TARGET_STRING handling.
13410 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13411 (const rs6000_opt_masks) <"string">: Change mask to 0.
13412 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13413 (MASK_STRING): Delete.
13414 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13416 (load_multiple): Delete.
13423 (store_multiple): Delete.
13430 (movmemsi_8reg): Delete.
13431 (corresponding unnamed define_insn): Delete.
13432 (movmemsi_6reg): Delete.
13433 (corresponding unnamed define_insn): Delete.
13434 (movmemsi_4reg): Delete.
13435 (corresponding unnamed define_insn): Delete.
13436 (movmemsi_2reg): Delete.
13437 (corresponding unnamed define_insn): Delete.
13438 (movmemsi_1reg): Delete.
13439 (corresponding unnamed define_insn): Delete.
13440 * config/rs6000/rs6000.opt (mno-string): New.
13441 (mstring): Replace by deprecation warning stub.
13442 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13444 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13446 * regrename.c (regrename_do_replace): If replacing the same
13447 reg multiple times, try to reuse last created gen_raw_REG.
13450 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13451 main to workaround a bug in GDB.
13453 2018-01-12 Tom de Vries <tom@codesourcery.com>
13456 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13458 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13460 PR rtl-optimization/80481
13461 * ira-color.c (get_cap_member): New function.
13462 (allocnos_conflict_by_live_ranges_p): Use it.
13463 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13464 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13466 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13469 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13470 (*saddl_se_1): Ditto.
13471 (*ssubsi_1): Ditto.
13472 (*ssubl_se_1): Ditto.
13474 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13476 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13477 rather than wi::to_widest for DR_INITs.
13478 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13479 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13480 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13482 (vect_analyze_group_access_1): Note that here.
13484 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13486 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13487 polynomial type sizes.
13489 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13491 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13492 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13493 (gimple_add_tmp_var): Likewise.
13495 2018-01-12 Martin Liska <mliska@suse.cz>
13497 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13498 (gimple_alloc_sizes): Likewise.
13499 (dump_gimple_statistics): Use PRIu64 in printf format.
13500 * gimple.h: Change uint64_t to int.
13502 2018-01-12 Martin Liska <mliska@suse.cz>
13504 * tree-core.h: Use uint64_t instead of int.
13505 * tree.c (tree_node_counts): Likewise.
13506 (tree_node_sizes): Likewise.
13507 (dump_tree_statistics): Use PRIu64 in printf format.
13509 2018-01-12 Martin Liska <mliska@suse.cz>
13511 * Makefile.in: As qsort_chk is implemented in vec.c, add
13512 vec.o to linkage of gencfn-macros.
13513 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13514 passing the info to record_node_allocation_statistics.
13515 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13517 * ggc-common.c (struct ggc_usage): Add operator== and use
13518 it in operator< and compare function.
13519 * mem-stats.h (struct mem_usage): Likewise.
13520 * vec.c (struct vec_usage): Remove operator< and compare
13521 function. Can be simply inherited.
13523 2018-01-12 Martin Jambor <mjambor@suse.cz>
13526 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13527 * tree-ssa-math-opts.c: Include domwalk.h.
13528 (convert_mult_to_fma_1): New function.
13529 (fma_transformation_info): New type.
13530 (fma_deferring_state): Likewise.
13531 (cancel_fma_deferring): New function.
13532 (result_of_phi): Likewise.
13533 (last_fma_candidate_feeds_initial_phi): Likewise.
13534 (convert_mult_to_fma): Added deferring logic, split actual
13535 transformation to convert_mult_to_fma_1.
13536 (math_opts_dom_walker): New type.
13537 (math_opts_dom_walker::after_dom_children): New method, body moved
13538 here from pass_optimize_widening_mul::execute, added deferring logic
13540 (pass_optimize_widening_mul::execute): Moved most of code to
13541 math_opts_dom_walker::after_dom_children.
13542 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13543 * config/i386/i386.c (ix86_option_override_internal): Added
13544 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13546 2018-01-12 Richard Biener <rguenther@suse.de>
13549 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13550 inline instance vars.
13552 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13555 * config/rx/rx.c (rx_is_restricted_memory_address):
13556 Handle SUBREG case.
13558 2018-01-12 Richard Biener <rguenther@suse.de>
13560 PR tree-optimization/80846
13561 * target.def (split_reduction): New target hook.
13562 * targhooks.c (default_split_reduction): New function.
13563 * targhooks.h (default_split_reduction): Declare.
13564 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13565 target requests first reduce vectors by combining low and high
13567 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13568 (get_vectype_for_scalar_type_and_size): Export.
13569 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13570 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13571 * doc/tm.texi: Regenerate.
13572 * config/i386/i386.c (ix86_split_reduction): Implement
13573 TARGET_VECTORIZE_SPLIT_REDUCTION.
13575 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13578 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13579 in PIC mode except for TARGET_VXWORKS_RTP.
13580 * config/sparc/sparc.c: Include cfgrtl.h.
13581 (TARGET_INIT_PIC_REG): Define.
13582 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13583 (sparc_pic_register_p): New predicate.
13584 (sparc_legitimate_address_p): Use it.
13585 (sparc_legitimize_pic_address): Likewise.
13586 (sparc_delegitimize_address): Likewise.
13587 (sparc_mode_dependent_address_p): Likewise.
13588 (gen_load_pcrel_sym): Remove 4th parameter.
13589 (load_got_register): Adjust call to above. Remove obsolete stuff.
13590 (sparc_expand_prologue): Do not call load_got_register here.
13591 (sparc_flat_expand_prologue): Likewise.
13592 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13593 (sparc_use_pseudo_pic_reg): New function.
13594 (sparc_init_pic_reg): Likewise.
13595 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13596 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13598 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13600 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13601 Add item for branch_cost.
13603 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13605 PR rtl-optimization/83565
13606 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13607 not extend the result to a larger mode for rotate operations.
13608 (num_sign_bit_copies1): Likewise.
13610 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13613 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13615 Use values-Xc.o for -pedantic.
13616 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13618 2018-01-12 Martin Liska <mliska@suse.cz>
13621 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13623 (possible_polymorphic_call_targets): Use it.
13624 (ipa_devirt): Likewise.
13626 2018-01-12 Martin Liska <mliska@suse.cz>
13628 * profile-count.h (enum profile_quality): Use 0 as invalid
13629 enum value of profile_quality.
13631 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13633 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13634 -mext-string options.
13636 2018-01-12 Richard Biener <rguenther@suse.de>
13638 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13639 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13640 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13642 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13644 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13646 * configure.ac (--with-long-double-format): Add support for the
13647 configuration option to change the default long double format on
13649 * config.gcc (powerpc*-linux*-*): Likewise.
13650 * configure: Regenerate.
13651 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13652 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13653 used without modification.
13655 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13657 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13658 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13659 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13660 MISC_BUILTIN_SPEC_BARRIER.
13661 (rs6000_init_builtins): Likewise.
13662 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13664 (speculation_barrier): New define_insn.
13665 * doc/extend.texi: Document __builtin_speculation_barrier.
13667 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13670 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13671 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13672 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13674 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13675 integral modes instead of "ss" and "sd".
13676 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13677 vectors with 32-bit and 64-bit elements.
13678 (vecdupssescalarmodesuffix): New mode attribute.
13679 (vec_dup<mode>): Use it.
13681 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13684 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13685 frame if argument is passed on stack.
13687 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13690 * ree.c (combine_reaching_defs): Optimize also
13691 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13692 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13694 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13696 PR middle-end/83189
13697 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13699 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13701 PR middle-end/83718
13702 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13703 after they are computed.
13705 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13707 PR tree-optimization/83695
13708 * gimple-loop-linterchange.cc
13709 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13710 reset cached scev information after interchange.
13711 (pass_linterchange::execute): Remove call to scev_reset_htab.
13713 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13715 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13716 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13717 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13718 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13719 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13720 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13721 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13722 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13723 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13724 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13725 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13726 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13727 (V_lane_reg): Likewise.
13728 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13730 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13731 (vfmal_lane_low<mode>_intrinsic,
13732 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13733 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13734 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13735 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13736 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13737 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13739 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13741 * config/arm/arm-cpus.in (fp16fml): New feature.
13742 (ALL_SIMD): Add fp16fml.
13743 (armv8.2-a): Add fp16fml as an option.
13744 (armv8.3-a): Likewise.
13745 (armv8.4-a): Add fp16fml as part of fp16.
13746 * config/arm/arm.h (TARGET_FP16FML): Define.
13747 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13749 * config/arm/arm-modes.def (V2HF): Define.
13750 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13751 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13752 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13753 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13754 vfmsl_low, vfmsl_high): New set of builtins.
13755 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13756 (vfml_op): New code attribute.
13757 (VFMLHALVES): New int iterator.
13758 (VFML, VFMLSEL): New mode attributes.
13759 (V_reg): Define mapping for V2HF.
13760 (V_hi, V_lo): New mode attributes.
13761 (VF_constraint): Likewise.
13762 (vfml_half, vfml_half_selector): New int attributes.
13763 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13765 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13766 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13768 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13769 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13770 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13771 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13773 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13774 Document new effective target and option set.
13776 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13778 * config/arm/arm-cpus.in (armv8_4): New feature.
13779 (ARMv8_4a): New fgroup.
13780 (armv8.4-a): New arch.
13781 * config/arm/arm-tables.opt: Regenerate.
13782 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13783 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13784 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13785 Add matching rules for -march=armv8.4-a and extensions.
13786 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13788 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13791 * config/rx/rx.md (BW): New mode attribute.
13792 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13794 2018-01-11 Richard Biener <rguenther@suse.de>
13796 PR tree-optimization/83435
13797 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13798 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13799 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13801 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13802 Alan Hayward <alan.hayward@arm.com>
13803 David Sherwood <david.sherwood@arm.com>
13805 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13807 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13808 (aarch64_print_address_internal): Use it to check for a zero offset.
13810 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13811 Alan Hayward <alan.hayward@arm.com>
13812 David Sherwood <david.sherwood@arm.com>
13814 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13815 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13816 Return a poly_int64 rather than a HOST_WIDE_INT.
13817 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13818 rather than a HOST_WIDE_INT.
13819 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13820 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13821 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13822 final_offset from HOST_WIDE_INT to poly_int64.
13823 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13824 to_constant when getting the number of units in an Advanced SIMD
13826 (aarch64_builtin_vectorized_function): Check for a constant number
13828 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13830 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13831 attribute instead of GET_MODE_NUNITS.
13832 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13833 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13834 GET_MODE_SIZE for fixed-size registers.
13835 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13836 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13837 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13838 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13839 (aarch64_print_operand, aarch64_print_address_internal)
13840 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13841 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13842 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13843 Handle polynomial GET_MODE_SIZE.
13844 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13845 wider than SImode without modification.
13846 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13847 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13848 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13849 passing and returning SVE modes.
13850 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13851 rather than GEN_INT.
13852 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13853 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13854 (aarch64_allocate_and_probe_stack_space): Likewise.
13855 (aarch64_layout_frame): Cope with polynomial offsets.
13856 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13857 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13858 polynomial offsets.
13859 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13860 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13861 poly_int64 rather than a HOST_WIDE_INT.
13862 (aarch64_get_separate_components, aarch64_process_components)
13863 (aarch64_expand_prologue, aarch64_expand_epilogue)
13864 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13865 (aarch64_anchor_offset): New function, split out from...
13866 (aarch64_legitimize_address): ...here.
13867 (aarch64_builtin_vectorization_cost): Handle polynomial
13868 TYPE_VECTOR_SUBPARTS.
13869 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13871 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13872 number of elements from the PARALLEL rather than the mode.
13873 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13874 rather than GET_MODE_BITSIZE.
13875 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13876 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13877 (aarch64_expand_vec_perm_const_1): Handle polynomial
13878 d->perm.length () and d->perm elements.
13879 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13880 Apply to_constant to d->perm elements.
13881 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13882 polynomial CONST_VECTOR_NUNITS.
13883 (aarch64_move_pointer): Take amount as a poly_int64 rather
13885 (aarch64_progress_pointer): Avoid temporary variable.
13886 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13887 the mode attribute instead of GET_MODE.
13889 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13890 Alan Hayward <alan.hayward@arm.com>
13891 David Sherwood <david.sherwood@arm.com>
13893 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13894 x exists before using it.
13895 (aarch64_add_constant_internal): Rename to...
13896 (aarch64_add_offset_1): ...this. Replace regnum with separate
13897 src and dest rtxes. Handle the case in which they're different,
13898 including when the offset is zero. Replace scratchreg with an rtx.
13899 Use 2 additions if there is no spare register into which we can
13900 move a 16-bit constant.
13901 (aarch64_add_constant): Delete.
13902 (aarch64_add_offset): Replace reg with separate src and dest
13903 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13904 Use aarch64_add_offset_1.
13905 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13906 an rtx rather than an int. Take the delta as a poly_int64
13907 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13908 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13909 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13910 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13911 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
13912 and aarch64_add_sp.
13913 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
13914 aarch64_add_constant.
13916 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13918 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
13919 Use scalar_float_mode.
13921 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13923 * config/aarch64/aarch64-simd.md
13924 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
13925 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
13926 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
13927 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
13928 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
13929 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
13930 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
13931 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
13932 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
13933 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
13935 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13938 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
13939 targ_options->x_arm_arch_string is non NULL.
13941 2018-01-11 Tamar Christina <tamar.christina@arm.com>
13943 * config/aarch64/aarch64.h
13944 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
13946 2018-01-11 Sudakshina Das <sudi.das@arm.com>
13949 * expmed.c (emit_store_flag_force): Swap if const op0
13950 and change VOIDmode to mode of op0.
13952 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13954 PR rtl-optimization/83761
13955 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
13956 than bytes to mode_for_size.
13958 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13960 PR middle-end/83189
13961 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
13962 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
13965 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13967 PR middle-end/83575
13968 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
13969 when in layout mode.
13970 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
13971 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
13974 2018-01-10 Michael Collison <michael.collison@arm.com>
13976 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
13977 * config/aarch64/aarch64-option-extension.def: Add
13978 AARCH64_OPT_EXTENSION of 'fp16fml'.
13979 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
13980 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
13981 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
13982 * config/aarch64/constraints.md (Ui7): New constraint.
13983 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
13984 (VFMLA_SEL_W): Ditto.
13987 (VFMLA16_LOW): New int iterator.
13988 (VFMLA16_HIGH): Ditto.
13989 (UNSPEC_FMLAL): New unspec.
13990 (UNSPEC_FMLSL): Ditto.
13991 (UNSPEC_FMLAL2): Ditto.
13992 (UNSPEC_FMLSL2): Ditto.
13993 (f16mac): New code attribute.
13994 * config/aarch64/aarch64-simd-builtins.def
13995 (aarch64_fmlal_lowv2sf): Ditto.
13996 (aarch64_fmlsl_lowv2sf): Ditto.
13997 (aarch64_fmlalq_lowv4sf): Ditto.
13998 (aarch64_fmlslq_lowv4sf): Ditto.
13999 (aarch64_fmlal_highv2sf): Ditto.
14000 (aarch64_fmlsl_highv2sf): Ditto.
14001 (aarch64_fmlalq_highv4sf): Ditto.
14002 (aarch64_fmlslq_highv4sf): Ditto.
14003 (aarch64_fmlal_lane_lowv2sf): Ditto.
14004 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14005 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14006 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14007 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14008 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14009 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14010 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14011 (aarch64_fmlal_lane_highv2sf): Ditto.
14012 (aarch64_fmlsl_lane_highv2sf): Ditto.
14013 (aarch64_fmlal_laneq_highv2sf): Ditto.
14014 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14015 (aarch64_fmlalq_lane_highv4sf): Ditto.
14016 (aarch64_fmlsl_lane_highv4sf): Ditto.
14017 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14018 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14019 * config/aarch64/aarch64-simd.md:
14020 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14021 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14022 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14023 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14024 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14025 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14026 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14027 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14028 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14029 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14030 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14031 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14032 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14033 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14034 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14035 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14036 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14037 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14038 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14039 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14040 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14041 (vfmlsl_low_u32): Ditto.
14042 (vfmlalq_low_u32): Ditto.
14043 (vfmlslq_low_u32): Ditto.
14044 (vfmlal_high_u32): Ditto.
14045 (vfmlsl_high_u32): Ditto.
14046 (vfmlalq_high_u32): Ditto.
14047 (vfmlslq_high_u32): Ditto.
14048 (vfmlal_lane_low_u32): Ditto.
14049 (vfmlsl_lane_low_u32): Ditto.
14050 (vfmlal_laneq_low_u32): Ditto.
14051 (vfmlsl_laneq_low_u32): Ditto.
14052 (vfmlalq_lane_low_u32): Ditto.
14053 (vfmlslq_lane_low_u32): Ditto.
14054 (vfmlalq_laneq_low_u32): Ditto.
14055 (vfmlslq_laneq_low_u32): Ditto.
14056 (vfmlal_lane_high_u32): Ditto.
14057 (vfmlsl_lane_high_u32): Ditto.
14058 (vfmlal_laneq_high_u32): Ditto.
14059 (vfmlsl_laneq_high_u32): Ditto.
14060 (vfmlalq_lane_high_u32): Ditto.
14061 (vfmlslq_lane_high_u32): Ditto.
14062 (vfmlalq_laneq_high_u32): Ditto.
14063 (vfmlslq_laneq_high_u32): Ditto.
14064 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14065 (AARCH64_FL_FOR_ARCH8_4): New.
14066 (AARCH64_ISA_F16FML): New ISA flag.
14067 (TARGET_F16FML): New feature flag for fp16fml.
14068 (doc/invoke.texi): Document new fp16fml option.
14070 2018-01-10 Michael Collison <michael.collison@arm.com>
14072 * config/aarch64/aarch64-builtins.c:
14073 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14074 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14075 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14076 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14077 (AARCH64_ISA_SHA3): New ISA flag.
14078 (TARGET_SHA3): New feature flag for sha3.
14079 * config/aarch64/iterators.md (sha512_op): New int attribute.
14080 (CRYPTO_SHA512): New int iterator.
14081 (UNSPEC_SHA512H): New unspec.
14082 (UNSPEC_SHA512H2): Ditto.
14083 (UNSPEC_SHA512SU0): Ditto.
14084 (UNSPEC_SHA512SU1): Ditto.
14085 * config/aarch64/aarch64-simd-builtins.def
14086 (aarch64_crypto_sha512hqv2di): New builtin.
14087 (aarch64_crypto_sha512h2qv2di): Ditto.
14088 (aarch64_crypto_sha512su0qv2di): Ditto.
14089 (aarch64_crypto_sha512su1qv2di): Ditto.
14090 (aarch64_eor3qv8hi): Ditto.
14091 (aarch64_rax1qv2di): Ditto.
14092 (aarch64_xarqv2di): Ditto.
14093 (aarch64_bcaxqv8hi): Ditto.
14094 * config/aarch64/aarch64-simd.md:
14095 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14096 (aarch64_crypto_sha512su0qv2di): Ditto.
14097 (aarch64_crypto_sha512su1qv2di): Ditto.
14098 (aarch64_eor3qv8hi): Ditto.
14099 (aarch64_rax1qv2di): Ditto.
14100 (aarch64_xarqv2di): Ditto.
14101 (aarch64_bcaxqv8hi): Ditto.
14102 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14103 (vsha512h2q_u64): Ditto.
14104 (vsha512su0q_u64): Ditto.
14105 (vsha512su1q_u64): Ditto.
14106 (veor3q_u16): Ditto.
14107 (vrax1q_u64): Ditto.
14108 (vxarq_u64): Ditto.
14109 (vbcaxq_u16): Ditto.
14110 * config/arm/types.md (crypto_sha512): New type attribute.
14111 (crypto_sha3): Ditto.
14112 (doc/invoke.texi): Document new sha3 option.
14114 2018-01-10 Michael Collison <michael.collison@arm.com>
14116 * config/aarch64/aarch64-builtins.c:
14117 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14118 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14119 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14120 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14121 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14122 (AARCH64_ISA_SM4): New ISA flag.
14123 (TARGET_SM4): New feature flag for sm4.
14124 * config/aarch64/aarch64-simd-builtins.def
14125 (aarch64_sm3ss1qv4si): Ditto.
14126 (aarch64_sm3tt1aq4si): Ditto.
14127 (aarch64_sm3tt1bq4si): Ditto.
14128 (aarch64_sm3tt2aq4si): Ditto.
14129 (aarch64_sm3tt2bq4si): Ditto.
14130 (aarch64_sm3partw1qv4si): Ditto.
14131 (aarch64_sm3partw2qv4si): Ditto.
14132 (aarch64_sm4eqv4si): Ditto.
14133 (aarch64_sm4ekeyqv4si): Ditto.
14134 * config/aarch64/aarch64-simd.md:
14135 (aarch64_sm3ss1qv4si): Ditto.
14136 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14137 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14138 (aarch64_sm4eqv4si): Ditto.
14139 (aarch64_sm4ekeyqv4si): Ditto.
14140 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14141 (sm3part_op): Ditto.
14142 (CRYPTO_SM3TT): Ditto.
14143 (CRYPTO_SM3PART): Ditto.
14144 (UNSPEC_SM3SS1): New unspec.
14145 (UNSPEC_SM3TT1A): Ditto.
14146 (UNSPEC_SM3TT1B): Ditto.
14147 (UNSPEC_SM3TT2A): Ditto.
14148 (UNSPEC_SM3TT2B): Ditto.
14149 (UNSPEC_SM3PARTW1): Ditto.
14150 (UNSPEC_SM3PARTW2): Ditto.
14151 (UNSPEC_SM4E): Ditto.
14152 (UNSPEC_SM4EKEY): Ditto.
14153 * config/aarch64/constraints.md (Ui2): New constraint.
14154 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14155 * config/arm/types.md (crypto_sm3): New type attribute.
14156 (crypto_sm4): Ditto.
14157 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14158 (vsm3tt1aq_u32): Ditto.
14159 (vsm3tt1bq_u32): Ditto.
14160 (vsm3tt2aq_u32): Ditto.
14161 (vsm3tt2bq_u32): Ditto.
14162 (vsm3partw1q_u32): Ditto.
14163 (vsm3partw2q_u32): Ditto.
14164 (vsm4eq_u32): Ditto.
14165 (vsm4ekeyq_u32): Ditto.
14166 (doc/invoke.texi): Document new sm4 option.
14168 2018-01-10 Michael Collison <michael.collison@arm.com>
14170 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14171 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14172 (AARCH64_FL_FOR_ARCH8_4): New.
14173 (AARCH64_FL_V8_4): New flag.
14174 (doc/invoke.texi): Document new armv8.4-a option.
14176 2018-01-10 Michael Collison <michael.collison@arm.com>
14178 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14179 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14180 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14181 * config/aarch64/aarch64-option-extension.def: Add
14182 AARCH64_OPT_EXTENSION of 'sha2'.
14183 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14184 (crypto): Disable sha2 and aes if crypto disabled.
14185 (crypto): Enable aes and sha2 if enabled.
14186 (simd): Disable sha2 and aes if simd disabled.
14187 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14189 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14190 (TARGET_SHA2): New feature flag for sha2.
14191 (TARGET_AES): New feature flag for aes.
14192 * config/aarch64/aarch64-simd.md:
14193 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14194 conditional on TARGET_AES.
14195 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14196 (aarch64_crypto_sha1hsi): Make pattern conditional
14198 (aarch64_crypto_sha1hv4si): Ditto.
14199 (aarch64_be_crypto_sha1hv4si): Ditto.
14200 (aarch64_crypto_sha1su1v4si): Ditto.
14201 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14202 (aarch64_crypto_sha1su0v4si): Ditto.
14203 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14204 (aarch64_crypto_sha256su0v4si): Ditto.
14205 (aarch64_crypto_sha256su1v4si): Ditto.
14206 (doc/invoke.texi): Document new aes and sha2 options.
14208 2018-01-10 Martin Sebor <msebor@redhat.com>
14210 PR tree-optimization/83781
14211 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14214 2018-01-11 Martin Sebor <msebor@gmail.com>
14215 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14217 PR tree-optimization/83501
14218 PR tree-optimization/81703
14220 * tree-ssa-strlen.c (get_string_cst): Rename...
14221 (get_string_len): ...to this. Handle global constants.
14222 (handle_char_store): Adjust.
14224 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14225 Jim Wilson <jimw@sifive.com>
14227 * config/riscv/riscv-protos.h (riscv_output_return): New.
14228 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14229 (riscv_attribute_table, riscv_output_return),
14230 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14231 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14232 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14234 (riscv_expand_prologue): Add early return for naked function.
14235 (riscv_expand_epilogue): Likewise.
14236 (riscv_function_ok_for_sibcall): Return false for naked function.
14237 (riscv_set_current_function): New.
14238 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14239 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14240 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14241 * doc/extend.texi (RISC-V Function Attributes): New.
14243 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14245 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14246 check for 128-bit long double before checking TCmode.
14247 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14248 128-bit long doubles before checking TFmode or TCmode.
14249 (FLOAT128_IBM_P): Likewise.
14251 2018-01-10 Martin Sebor <msebor@redhat.com>
14253 PR tree-optimization/83671
14254 * builtins.c (c_strlen): Unconditionally return zero for the empty
14256 Use -Warray-bounds for warnings.
14257 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14258 for non-constant array indices with COMPONENT_REF, arrays of
14259 arrays, and pointers to arrays.
14260 (gimple_fold_builtin_strlen): Determine and set length range for
14261 non-constant character arrays.
14263 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14265 PR middle-end/81897
14266 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14269 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14271 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14273 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14276 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14277 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14278 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14279 indexed_or_indirect_operand predicate.
14280 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14281 (*vsx_le_perm_load_v8hi): Likewise.
14282 (*vsx_le_perm_load_v16qi): Likewise.
14283 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14284 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14285 (*vsx_le_perm_store_v8hi): Likewise.
14286 (*vsx_le_perm_store_v16qi): Likewise.
14287 (eight unnamed splitters): Likewise.
14289 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14291 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14292 * config/rs6000/emmintrin.h: Likewise.
14293 * config/rs6000/mmintrin.h: Likewise.
14294 * config/rs6000/xmmintrin.h: Likewise.
14296 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14299 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14301 * tree.c (tree_nop_conversion): Return true for location wrapper
14303 (maybe_wrap_with_location): New function.
14304 (selftest::check_strip_nops): New function.
14305 (selftest::test_location_wrappers): New function.
14306 (selftest::tree_c_tests): Call it.
14307 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14308 (maybe_wrap_with_location): New decl.
14309 (EXPR_LOCATION_WRAPPER_P): New macro.
14310 (location_wrapper_p): New inline function.
14311 (tree_strip_any_location_wrapper): New inline function.
14313 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14316 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14317 stack_realign_offset for the largest alignment of stack slot
14319 (ix86_find_max_used_stack_alignment): New function.
14320 (ix86_finalize_stack_frame_flags): Use it. Set
14321 max_used_stack_alignment if we don't realign stack.
14322 * config/i386/i386.h (machine_function): Add
14323 max_used_stack_alignment.
14325 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14327 * config/arm/arm.opt (-mbranch-cost): New option.
14328 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14331 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14334 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14335 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14337 2018-01-10 Richard Biener <rguenther@suse.de>
14340 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14341 early out so it also covers the case where we have a non-NULL
14344 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14346 PR tree-optimization/83753
14347 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14348 for non-strided grouped accesses if the number of elements is 1.
14350 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14353 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14354 * i386.h (TARGET_USE_GATHER): Define.
14355 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14357 2018-01-10 Martin Liska <mliska@suse.cz>
14360 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14361 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14363 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14364 CLEANUP_NO_PARTITIONING is not set.
14366 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14368 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14369 for vectors, as a partial revert of r254296.
14370 * rtl.h (const_vec_p): Delete.
14371 (const_vec_duplicate_p): Don't test for vector CONSTs.
14372 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14373 * expmed.c (make_tree): Likewise.
14376 * common.md (E, F): Use CONSTANT_P instead of checking for
14378 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14379 checking for CONST_VECTOR.
14381 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14383 PR middle-end/83575
14384 * predict.c (force_edge_cold): Handle in more sane way edges
14385 with no prediction.
14387 2018-01-09 Carl Love <cel@us.ibm.com>
14389 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14391 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14392 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14393 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14394 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14395 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14396 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14397 * config/rs6000/rs6000-protos.h: Add extern defition for
14398 rs6000_generate_float2_double_code.
14399 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14401 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14402 (float2_v2df): Add define_expand.
14404 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14407 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14408 op_mode in the force_to_mode call.
14410 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14412 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14413 instead of checking each element individually.
14414 (aarch64_evpc_uzp): Likewise.
14415 (aarch64_evpc_zip): Likewise.
14416 (aarch64_evpc_ext): Likewise.
14417 (aarch64_evpc_rev): Likewise.
14418 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14419 instead of checking each element individually. Return true without
14421 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14422 whether all selected elements come from the same input, instead of
14423 checking each element individually. Remove calls to gen_rtx_REG,
14424 start_sequence and end_sequence and instead assert that no rtl is
14427 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14429 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14430 order of HIGH and CONST checks.
14432 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14434 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14435 if the destination isn't an SSA_NAME.
14437 2018-01-09 Richard Biener <rguenther@suse.de>
14439 PR tree-optimization/83668
14440 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14442 (canonicalize_loop_form): ... here, renamed from ...
14443 (canonicalize_loop_closed_ssa_form): ... this and amended to
14444 swap successor edges for loop exit blocks to make us use
14445 the RPO order we need for initial schedule generation.
14447 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14449 PR tree-optimization/64811
14450 * match.pd: When optimizing comparisons with Inf, avoid
14451 introducing or losing exceptions from comparisons with NaN.
14453 2018-01-09 Martin Liska <mliska@suse.cz>
14456 * asan.c (shadow_mem_size): Add gcc_assert.
14458 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14460 Don't save registers in main().
14463 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14464 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14465 * config/avr/avr.c (avr_set_current_function): Don't error if
14466 naked, OS_task or OS_main are specified at the same time.
14467 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14469 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14471 * common/config/avr/avr-common.c (avr_option_optimization_table):
14472 Switch on -mmain-is-OS_task for optimizing compilations.
14474 2018-01-09 Richard Biener <rguenther@suse.de>
14476 PR tree-optimization/83572
14477 * graphite.c: Include cfganal.h.
14478 (graphite_transform_loops): Connect infinite loops to exit
14479 and remove fake edges at the end.
14481 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14483 * ipa-inline.c (edge_badness): Revert accidental checkin.
14485 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14488 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14489 symbols; not inline clones.
14491 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14494 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14495 hard registers. Formatting fixes.
14497 PR preprocessor/83722
14498 * gcc.c (try_generate_repro): Pass
14499 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14500 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14503 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14504 Kito Cheng <kito.cheng@gmail.com>
14506 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14507 (riscv_leaf_function_p): Delete.
14508 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14510 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14512 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14514 (do_ifelse): New function.
14515 (do_isel): New function.
14516 (do_sub3): New function.
14517 (do_add3): New function.
14518 (do_load_mask_compare): New function.
14519 (do_overlap_load_compare): New function.
14520 (expand_compare_loop): New function.
14521 (expand_block_compare): Call expand_compare_loop() when appropriate.
14522 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14523 option description.
14524 (-mblock-compare-inline-loop-limit): New option.
14526 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14529 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14530 Reverse order of second and third operands in first alternative.
14531 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14532 of first and second elements in UNSPEC_VPERMR vector.
14533 (altivec_expand_vec_perm_le): Likewise.
14535 2018-01-08 Jeff Law <law@redhat.com>
14537 PR rtl-optimizatin/81308
14538 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14539 (process_switch): If group_case_labels makes a change, then set
14541 (pass_convert_switch::execute): If a switch is converted, then
14542 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14544 PR rtl-optimization/81308
14545 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14548 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14550 PR target/83663 - Revert r255946
14551 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14552 generation for cases where splatting a value is not useful.
14553 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14554 across a vec_duplicate and a paradoxical subreg forming a vector
14555 mode to a vec_concat.
14557 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14559 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14560 -march=armv8.3-a variants.
14561 * config/arm/t-multilib: Likewise.
14562 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14564 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14566 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14568 (cceq_ior_compare_complement): Give it a name so I can use it, and
14569 change boolean_or_operator predicate to boolean_operator so it can
14570 be used to generate a crand.
14571 (eqne): New code iterator.
14572 (bd/bd_neg): New code_attrs.
14573 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14574 a single define_insn.
14575 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14576 decrement (bdnzt/bdnzf/bdzt/bdzf).
14577 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14578 with the new names of the branch decrement patterns, and added the
14579 names of the branch decrement conditional patterns.
14581 2018-01-08 Richard Biener <rguenther@suse.de>
14583 PR tree-optimization/83563
14584 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14587 2018-01-08 Richard Biener <rguenther@suse.de>
14589 PR middle-end/83713
14590 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14592 2018-01-08 Richard Biener <rguenther@suse.de>
14594 PR tree-optimization/83685
14595 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14596 references to abnormals.
14598 2018-01-08 Richard Biener <rguenther@suse.de>
14601 * dwarf2out.c (output_indirect_strings): Handle empty
14602 skeleton_debug_str_hash.
14603 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14605 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14607 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14608 (emit_store_direct): Likewise.
14609 (arc_trampoline_adjust_address): Likewise.
14610 (arc_asm_trampoline_template): New function.
14611 (arc_initialize_trampoline): Use asm_trampoline_template.
14612 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14613 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14614 * config/arc/arc.md (flush_icache): Delete pattern.
14616 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14618 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14619 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14622 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14625 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14626 by not USED_FOR_TARGET.
14627 (make_pass_resolve_sw_modes): Likewise.
14629 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14631 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14634 2018-01-08 Richard Biener <rguenther@suse.de>
14636 PR middle-end/83580
14637 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14639 2018-01-08 Richard Biener <rguenther@suse.de>
14641 PR middle-end/83517
14642 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14644 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14646 PR middle-end/81897
14647 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14648 basic blocks with a small number of successors.
14649 (convert_control_dep_chain_into_preds): Improve handling of
14651 (dump_predicates): Split apart into...
14652 (dump_pred_chain): ...here...
14653 (dump_pred_info): ...and here.
14654 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14655 (can_chain_union_be_invalidated_p): Improve check for invalidation
14657 (uninit_uses_cannot_happen): Avoid unnecessary if
14658 convert_control_dep_chain_into_preds yielded nothing.
14660 2018-01-06 Martin Sebor <msebor@redhat.com>
14662 PR tree-optimization/83640
14663 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14664 subtracting negative offset from size.
14665 (builtin_access::overlap): Adjust offset bounds of the access to fall
14666 within the size of the object if possible.
14668 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14670 PR rtl-optimization/83699
14671 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14672 extract_bit_field_as_subreg to cases in which the extracted
14673 value is also a vector.
14675 * lra-constraints.c (process_alt_operands): Test for the equivalence
14676 substitutions when detecting a possible reload cycle.
14678 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14681 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14682 by default if flag_selective_schedling{,2}. Formatting fixes.
14684 PR rtl-optimization/83682
14685 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14686 if it has non-VECTOR_MODE element mode.
14687 (vec_duplicate_p): Likewise.
14689 PR middle-end/83694
14690 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14691 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14693 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14696 * config/i386/i386-builtin.def
14697 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14698 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14699 Require also OPTION_MASK_ISA_AVX512F in addition to
14700 OPTION_MASK_ISA_GFNI.
14701 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14702 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14703 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14704 to OPTION_MASK_ISA_GFNI.
14705 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14706 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14707 OPTION_MASK_ISA_AVX512BW.
14708 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14709 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14710 addition to OPTION_MASK_ISA_GFNI.
14711 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14712 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14713 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14714 to OPTION_MASK_ISA_GFNI.
14715 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14716 a requirement for all ISAs rather than any of them with a few
14718 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14720 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14721 bitmasks to be enabled with 3 exceptions, instead of requiring any
14722 enabled ISA with lots of exceptions.
14723 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14724 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14725 Change avx512bw in isa attribute to avx512f.
14726 * config/i386/sgxintrin.h: Add license boilerplate.
14727 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14728 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14729 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14730 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14732 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14733 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14734 temporarily sse2 rather than sse if not enabled already.
14737 * config/i386/sse.md (VI248_VLBW): Rename to ...
14738 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14739 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14740 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14741 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14742 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14743 mode iterator instead of VI248_VLBW.
14745 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14747 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14748 (record_modified): Skip clobbers; add debug output.
14749 (param_change_prob): Use sreal frequencies.
14751 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14753 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14754 punt for user-aligned variables.
14756 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14758 * tree-chrec.c (chrec_contains_symbols): Return true for
14761 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14764 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14765 of (x|y) == x for BICS pattern.
14767 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14769 PR tree-optimization/83605
14770 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14771 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14774 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14776 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14777 * config/epiphany/rtems.h: New file.
14779 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14780 Uros Bizjak <ubizjak@gmail.com>
14783 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14784 QIreg_operand instead of register_operand predicate.
14785 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14786 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14787 comments instead of -fmitigate[-_]rop.
14789 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14792 * cgraphunit.c (symbol_table::compile): Switch to text_section
14793 before calling assembly_start debug hook.
14794 * run-rtl-passes.c (run_rtl_passes): Likewise.
14797 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14799 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14800 range_int_cst_p rather than !symbolic_range_p before calling
14801 extract_range_from_multiplicative_op_1.
14803 2018-01-04 Jeff Law <law@redhat.com>
14805 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14806 redundant test in assertion.
14808 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14810 * doc/rtl.texi: Document machine_mode wrapper classes.
14812 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14814 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14815 using tree_to_uhwi.
14817 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14819 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14820 the VEC_PERM_EXPR fold to fail.
14822 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14825 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14826 to switched_sections.
14828 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14831 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14832 test for d.testing.
14834 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14837 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14838 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14840 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14843 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14844 is BLKmode and bitpos not zero or mode change is needed.
14846 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14849 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14852 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14855 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14856 instead of MULT rtx. Update all corresponding splitters.
14857 (*saddl_se): Ditto.
14858 (*ssub<modesuffix>): Ditto.
14859 (*ssubl_se): Ditto.
14860 (*cmp_sadd_di): Update split patterns.
14861 (*cmp_sadd_si): Ditto.
14862 (*cmp_sadd_sidi): Ditto.
14863 (*cmp_ssub_di): Ditto.
14864 (*cmp_ssub_si): Ditto.
14865 (*cmp_ssub_sidi): Ditto.
14866 * config/alpha/predicates.md (const23_operand): New predicate.
14867 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14868 Look for ASHIFT, not MULT inner operand.
14869 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14871 2018-01-04 Martin Liska <mliska@suse.cz>
14873 PR gcov-profile/83669
14874 * gcov.c (output_intermediate_file): Add version to intermediate
14876 * doc/gcov.texi: Document new field 'version' in intermediate
14877 file format. Fix location of '-k' option of gcov command.
14879 2018-01-04 Martin Liska <mliska@suse.cz>
14882 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14884 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14886 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14888 2018-01-03 Martin Sebor <msebor@redhat.com>
14890 PR tree-optimization/83655
14891 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14892 checking calls with invalid arguments.
14894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14896 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14897 (vectorizable_mask_load_store): Delete.
14898 (vectorizable_call): Return false for masked loads and stores.
14899 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14900 instead of gimple_assign_rhs1.
14901 (vectorizable_load): Handle IFN_MASK_LOAD.
14902 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14906 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14908 (vectorizable_mask_load_store): ...here.
14909 (vectorizable_load): ...and here.
14911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14913 * tree-vect-stmts.c (vect_build_all_ones_mask)
14914 (vect_build_zero_merge_argument): New functions, split out from...
14915 (vectorizable_load): ...here.
14917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14919 * tree-vect-stmts.c (vect_check_store_rhs): New function,
14921 (vectorizable_mask_load_store): ...here.
14922 (vectorizable_store): ...and here.
14924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14926 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
14928 (vectorizable_mask_load_store): ...here.
14930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14932 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
14933 (vect_model_store_cost): Take a vec_load_store_type instead of a
14935 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
14936 (vect_model_store_cost): Take a vec_load_store_type instead of a
14938 (vectorizable_mask_load_store): Update accordingly.
14939 (vectorizable_store): Likewise.
14940 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
14942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14944 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
14945 IFN_MASK_LOAD calls here rather than...
14946 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
14948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14949 Alan Hayward <alan.hayward@arm.com>
14950 David Sherwood <david.sherwood@arm.com>
14952 * expmed.c (extract_bit_field_1): For vector extracts,
14953 fall back to extract_bit_field_as_subreg if vec_extract
14956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14957 Alan Hayward <alan.hayward@arm.com>
14958 David Sherwood <david.sherwood@arm.com>
14960 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
14961 they are variable or constant sized.
14962 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
14963 slots for constant-sized data.
14965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14966 Alan Hayward <alan.hayward@arm.com>
14967 David Sherwood <david.sherwood@arm.com>
14969 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
14970 handling COND_EXPRs with boolean comparisons, try to find a better
14971 basis for the mask type than the boolean itself.
14973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14975 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
14976 is calculated and how it can be overridden.
14977 * genmodes.c (max_bitsize_mode_any_mode): New variable.
14978 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
14980 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
14983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14984 Alan Hayward <alan.hayward@arm.com>
14985 David Sherwood <david.sherwood@arm.com>
14987 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
14988 Remove the mode argument.
14989 (aarch64_simd_valid_immediate): Remove the mode and inverse
14991 * config/aarch64/iterators.md (bitsize): New iterator.
14992 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
14993 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
14994 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
14995 aarch64_simd_valid_immediate.
14996 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
14997 (aarch64_reg_or_bic_imm): Likewise.
14998 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
14999 with an insn_type enum and msl with a modifier_type enum.
15000 Replace element_width with a scalar_mode. Change the shift
15001 to unsigned int. Add constructors for scalar_float_mode and
15002 scalar_int_mode elements.
15003 (aarch64_vect_float_const_representable_p): Delete.
15004 (aarch64_can_const_movi_rtx_p)
15005 (aarch64_simd_scalar_immediate_valid_for_move)
15006 (aarch64_simd_make_constant): Update call to
15007 aarch64_simd_valid_immediate.
15008 (aarch64_advsimd_valid_immediate_hs): New function.
15009 (aarch64_advsimd_valid_immediate): Likewise.
15010 (aarch64_simd_valid_immediate): Remove mode and inverse
15011 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15012 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15013 and aarch64_float_const_representable_p on the result.
15014 (aarch64_output_simd_mov_immediate): Remove mode argument.
15015 Update call to aarch64_simd_valid_immediate and use of
15016 simd_immediate_info.
15017 (aarch64_output_scalar_simd_mov_immediate): Update call
15020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15021 Alan Hayward <alan.hayward@arm.com>
15022 David Sherwood <david.sherwood@arm.com>
15024 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15025 (mode_nunits): Likewise CONST_MODE_NUNITS.
15026 * machmode.def (ADJUST_NUNITS): Document.
15027 * genmodes.c (mode_data::need_nunits_adj): New field.
15028 (blank_mode): Update accordingly.
15029 (adj_nunits): New variable.
15030 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15032 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15033 listed in adj_nunits.
15034 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15035 listed in adj_nunits. Don't emit case statements for such modes.
15036 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15037 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15038 nothing if adj_nunits is nonnull.
15039 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15040 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15041 (emit_mode_fbit): Update use of print_maybe_const_decl.
15042 (emit_move_size): Likewise. Treat the array as non-const
15044 (emit_mode_adjustments): Handle adj_nunits.
15046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15048 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15049 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15050 (VECTOR_MODES): Use it.
15051 (make_vector_modes): Take the prefix as an argument.
15053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15054 Alan Hayward <alan.hayward@arm.com>
15055 David Sherwood <david.sherwood@arm.com>
15057 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15058 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15059 for MODE_VECTOR_BOOL.
15060 * machmode.def (VECTOR_BOOL_MODE): Document.
15061 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15062 (make_vector_bool_mode): New function.
15063 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15065 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15066 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15068 * stor-layout.c (int_mode_for_mode): Likewise.
15069 * tree.c (build_vector_type_for_mode): Likewise.
15070 * varasm.c (output_constant_pool_2): Likewise.
15071 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15072 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15073 for MODE_VECTOR_BOOL.
15074 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15075 of mode class checks.
15076 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15077 instead of a list of mode class checks.
15078 (expand_vector_scalar_condition): Likewise.
15079 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15082 Alan Hayward <alan.hayward@arm.com>
15083 David Sherwood <david.sherwood@arm.com>
15085 * machmode.h (mode_size): Change from unsigned short to
15087 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15088 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15089 or if measurement_type is not polynomial.
15090 (fixed_size_mode::includes_p): Check for constant-sized modes.
15091 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15092 return a poly_uint16 rather than an unsigned short.
15093 (emit_mode_size): Change the type of mode_size from unsigned short
15094 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15095 (emit_mode_adjustments): Cope with polynomial vector sizes.
15096 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15098 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15100 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15101 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15102 * caller-save.c (setup_save_areas): Likewise.
15103 (replace_reg_with_saved_mem): Likewise.
15104 * calls.c (emit_library_call_value_1): Likewise.
15105 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15106 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15107 (gen_lowpart_for_combine): Likewise.
15108 * convert.c (convert_to_integer_1): Likewise.
15109 * cse.c (equiv_constant, cse_insn): Likewise.
15110 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15111 (cselib_subst_to_values): Likewise.
15112 * dce.c (word_dce_process_block): Likewise.
15113 * df-problems.c (df_word_lr_mark_ref): Likewise.
15114 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15115 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15116 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15117 (rtl_for_decl_location): Likewise.
15118 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15119 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15120 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15121 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15122 (expand_expr_real_1): Likewise.
15123 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15124 (pad_below): Likewise.
15125 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15126 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15127 * ira.c (get_subreg_tracking_sizes): Likewise.
15128 * ira-build.c (ira_create_allocno_objects): Likewise.
15129 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15130 (ira_sort_regnos_for_alter_reg): Likewise.
15131 * ira-costs.c (record_operand_costs): Likewise.
15132 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15133 (resolve_simple_move): Likewise.
15134 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15135 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15136 (lra_constraints): Likewise.
15137 (CONST_POOL_OK_P): Reject variable-sized modes.
15138 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15139 (add_pseudo_to_slot, lra_spill): Likewise.
15140 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15141 * optabs-query.c (get_best_extraction_insn): Likewise.
15142 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15143 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15144 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15145 * recog.c (offsettable_address_addr_space_p): Likewise.
15146 * regcprop.c (maybe_mode_change): Likewise.
15147 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15148 * regrename.c (build_def_use): Likewise.
15149 * regstat.c (dump_reg_info): Likewise.
15150 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15151 (find_reloads, find_reloads_subreg_address): Likewise.
15152 * reload1.c (eliminate_regs_1): Likewise.
15153 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15154 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15155 (simplify_binary_operation_1, simplify_subreg): Likewise.
15156 * targhooks.c (default_function_arg_padding): Likewise.
15157 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15158 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15159 (verify_gimple_assign_ternary): Likewise.
15160 * tree-inline.c (estimate_move_cost): Likewise.
15161 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15162 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15163 (get_address_cost_ainc): Likewise.
15164 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15165 (vect_supportable_dr_alignment): Likewise.
15166 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15167 (vectorizable_reduction): Likewise.
15168 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15169 (vectorizable_operation, vectorizable_load): Likewise.
15170 * tree.c (build_same_sized_truth_vector_type): Likewise.
15171 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15172 * var-tracking.c (emit_note_insn_var_location): Likewise.
15173 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15174 (ADDR_VEC_ALIGN): Likewise.
15176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15177 Alan Hayward <alan.hayward@arm.com>
15178 David Sherwood <david.sherwood@arm.com>
15180 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15182 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15183 or if measurement_type is polynomial.
15184 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15185 * combine.c (make_extraction): Likewise.
15186 * dse.c (find_shift_sequence): Likewise.
15187 * dwarf2out.c (mem_loc_descriptor): Likewise.
15188 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15189 (extract_bit_field, extract_low_bits): Likewise.
15190 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15191 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15192 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15193 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15194 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15195 * reload.c (find_reloads): Likewise.
15196 * reload1.c (alter_reg): Likewise.
15197 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15198 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15199 * tree-if-conv.c (predicate_mem_writes): Likewise.
15200 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15201 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15202 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15203 * valtrack.c (dead_debug_insert_temp): Likewise.
15204 * varasm.c (mergeable_constant_section): Likewise.
15205 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15208 Alan Hayward <alan.hayward@arm.com>
15209 David Sherwood <david.sherwood@arm.com>
15211 * expr.c (expand_assignment): Cope with polynomial mode sizes
15212 when assigning to a CONCAT.
15214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15215 Alan Hayward <alan.hayward@arm.com>
15216 David Sherwood <david.sherwood@arm.com>
15218 * machmode.h (mode_precision): Change from unsigned short to
15220 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15222 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15223 or if measurement_type is not polynomial.
15224 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15225 in which the mode is already known to be a scalar_int_mode.
15226 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15227 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15229 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15230 for GET_MODE_PRECISION.
15231 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15232 for GET_MODE_PRECISION.
15233 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15235 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15236 (expand_field_assignment, make_extraction): Likewise.
15237 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15238 (get_last_value): Likewise.
15239 * convert.c (convert_to_integer_1): Likewise.
15240 * cse.c (cse_insn): Likewise.
15241 * expr.c (expand_expr_real_1): Likewise.
15242 * lra-constraints.c (simplify_operand_subreg): Likewise.
15243 * optabs-query.c (can_atomic_load_p): Likewise.
15244 * optabs.c (expand_atomic_load): Likewise.
15245 (expand_atomic_store): Likewise.
15246 * ree.c (combine_reaching_defs): Likewise.
15247 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15248 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15249 * tree.h (type_has_mode_precision_p): Likewise.
15250 * ubsan.c (instrument_si_overflow): Likewise.
15252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15253 Alan Hayward <alan.hayward@arm.com>
15254 David Sherwood <david.sherwood@arm.com>
15256 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15257 polynomial numbers of units.
15258 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15259 (valid_vector_subparts_p): New function.
15260 (build_vector_type): Remove temporary shim and take the number
15261 of units as a poly_uint64 rather than an int.
15262 (build_opaque_vector_type): Take the number of units as a
15263 poly_uint64 rather than an int.
15264 * tree.c (build_vector_from_ctor): Handle polynomial
15265 TYPE_VECTOR_SUBPARTS.
15266 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15267 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15268 (build_vector_from_val): If the number of units is variable,
15269 use build_vec_duplicate_cst for constant operands and
15270 VEC_DUPLICATE_EXPR otherwise.
15271 (make_vector_type): Remove temporary is_constant ().
15272 (build_vector_type, build_opaque_vector_type): Take the number of
15273 units as a poly_uint64 rather than an int.
15274 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15276 * cfgexpand.c (expand_debug_expr): Likewise.
15277 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15278 (store_constructor, expand_expr_real_1): Likewise.
15279 (const_scalar_mask_from_tree): Likewise.
15280 * fold-const-call.c (fold_const_reduction): Likewise.
15281 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15282 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15283 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15284 (fold_relational_const): Likewise.
15285 (native_interpret_vector): Likewise. Change the size from an
15286 int to an unsigned int.
15287 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15288 TYPE_VECTOR_SUBPARTS.
15289 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15290 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15291 duplicating a non-constant operand into a variable-length vector.
15292 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15293 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15294 * ipa-icf.c (sem_variable::equals): Likewise.
15295 * match.pd: Likewise.
15296 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15297 * print-tree.c (print_node): Likewise.
15298 * stor-layout.c (layout_type): Likewise.
15299 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15300 * tree-cfg.c (verify_gimple_comparison): Likewise.
15301 (verify_gimple_assign_binary): Likewise.
15302 (verify_gimple_assign_ternary): Likewise.
15303 (verify_gimple_assign_single): Likewise.
15304 * tree-pretty-print.c (dump_generic_node): Likewise.
15305 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15306 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15307 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15308 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15309 (vect_shift_permute_load_chain): Likewise.
15310 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15311 (expand_vector_condition, optimize_vector_constructor): Likewise.
15312 (lower_vec_perm, get_compute_type): Likewise.
15313 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15314 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15315 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15316 (vect_recog_mask_conversion_pattern): Likewise.
15317 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15318 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15319 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15320 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15321 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15322 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15323 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15324 (supportable_widening_operation): Likewise.
15325 (supportable_narrowing_operation): Likewise.
15326 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15328 * varasm.c (output_constant): Likewise.
15330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15331 Alan Hayward <alan.hayward@arm.com>
15332 David Sherwood <david.sherwood@arm.com>
15334 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15335 so that both the length == 3 and length != 3 cases set up their
15336 own permute vectors. Add comments explaining why we know the
15337 number of elements is constant.
15338 (vect_permute_load_chain): Likewise.
15340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15341 Alan Hayward <alan.hayward@arm.com>
15342 David Sherwood <david.sherwood@arm.com>
15344 * machmode.h (mode_nunits): Change from unsigned char to
15346 (ONLY_FIXED_SIZE_MODES): New macro.
15347 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15348 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15349 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15351 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15352 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15353 or if measurement_type is not polynomial.
15354 * genmodes.c (ZERO_COEFFS): New macro.
15355 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15357 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15358 Use ZERO_COEFFS when emitting initializers.
15359 * data-streamer.h (bp_pack_poly_value): New function.
15360 (bp_unpack_poly_value): Likewise.
15361 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15362 for GET_MODE_NUNITS.
15363 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15364 for GET_MODE_NUNITS.
15365 * tree.c (make_vector_type): Remove temporary shim and make
15366 the real function take the number of units as a poly_uint64
15367 rather than an int.
15368 (build_vector_type_for_mode): Handle polynomial nunits.
15369 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15370 * emit-rtl.c (const_vec_series_p_1): Likewise.
15371 (gen_rtx_CONST_VECTOR): Likewise.
15372 * fold-const.c (test_vec_duplicate_folding): Likewise.
15373 * genrecog.c (validate_pattern): Likewise.
15374 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15375 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15376 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15377 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15378 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15379 * rtlanal.c (subreg_get_info): Likewise.
15380 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15381 (vect_grouped_load_supported): Likewise.
15382 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15383 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15384 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15385 (simplify_const_unary_operation, simplify_binary_operation_1)
15386 (simplify_const_binary_operation, simplify_ternary_operation)
15387 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15388 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15389 instead of CONST_VECTOR_NUNITS.
15390 * varasm.c (output_constant_pool_2): Likewise.
15391 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15392 explicit-encoded elements in the XVEC for variable-length vectors.
15394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15396 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15399 Alan Hayward <alan.hayward@arm.com>
15400 David Sherwood <david.sherwood@arm.com>
15402 * coretypes.h (fixed_size_mode): Declare.
15403 (fixed_size_mode_pod): New typedef.
15404 * builtins.h (target_builtins::x_apply_args_mode)
15405 (target_builtins::x_apply_result_mode): Change type to
15406 fixed_size_mode_pod.
15407 * builtins.c (apply_args_size, apply_result_size, result_vector)
15408 (expand_builtin_apply_args_1, expand_builtin_apply)
15409 (expand_builtin_return): Update accordingly.
15411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15413 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15414 * cselib.c (cselib_hash_rtx): Likewise.
15415 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15416 CONST_VECTOR encoding.
15418 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15419 Jeff Law <law@redhat.com>
15422 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15423 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15424 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15425 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15428 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15429 explicitly probe *sp in a noreturn function if there were any callee
15430 register saves or frame pointer is needed.
15432 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15435 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15436 BLKmode for ternary, binary or unary expressions.
15439 * var-tracking.c (delete_vta_debug_insn): New inline function.
15440 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15441 insns from get_insns () to NULL instead of each bb separately.
15442 Use delete_vta_debug_insn. No longer static.
15443 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15444 delete_vta_debug_insns callers.
15445 * rtl.h (delete_vta_debug_insns): Declare.
15446 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15447 instead of variable_tracking_main.
15449 2018-01-03 Martin Sebor <msebor@redhat.com>
15451 PR tree-optimization/83603
15452 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15453 arguments past the endof the argument list in functions declared
15454 without a prototype.
15455 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15456 Avoid checking when arguments are null.
15458 2018-01-03 Martin Sebor <msebor@redhat.com>
15461 * doc/extend.texi (attribute const): Fix a typo.
15462 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15463 issuing -Wsuggest-attribute for void functions.
15465 2018-01-03 Martin Sebor <msebor@redhat.com>
15467 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15468 offset_int::from instead of wide_int::to_shwi.
15469 (maybe_diag_overlap): Remove assertion.
15470 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15471 * gimple-ssa-sprintf.c (format_directive): Same.
15472 (parse_directive): Same.
15473 (sprintf_dom_walker::compute_format_length): Same.
15474 (try_substitute_return_value): Same.
15476 2018-01-03 Jeff Law <law@redhat.com>
15478 PR middle-end/83654
15479 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15480 non-constant residual for zero at runtime and avoid probing in
15481 that case. Reorganize code for trailing problem to mirror handling
15484 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15486 PR tree-optimization/83501
15487 * tree-ssa-strlen.c (get_string_cst): New.
15488 (handle_char_store): Call get_string_cst.
15490 2018-01-03 Martin Liska <mliska@suse.cz>
15492 PR tree-optimization/83593
15493 * tree-ssa-strlen.c: Include tree-cfg.h.
15494 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15495 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15496 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15498 (strlen_dom_walker::before_dom_children): Call
15499 gimple_purge_dead_eh_edges. Dump tranformation with details
15501 (strlen_dom_walker::before_dom_children): Update call by adding
15502 new argument cleanup_eh.
15503 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15505 2018-01-03 Martin Liska <mliska@suse.cz>
15508 * cif-code.def (VARIADIC_THUNK): New enum value.
15509 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15512 2018-01-03 Jan Beulich <jbeulich@suse.com>
15514 * sse.md (mov<mode>_internal): Tighten condition for when to use
15515 vmovdqu<ssescalarsize> for TI and OI modes.
15517 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15519 Update copyright years.
15521 2018-01-03 Martin Liska <mliska@suse.cz>
15524 * ipa-visibility.c (function_and_variable_visibility): Skip
15525 functions with noipa attribure.
15527 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15529 * gcc.c (process_command): Update copyright notice dates.
15530 * gcov-dump.c (print_version): Ditto.
15531 * gcov.c (print_version): Ditto.
15532 * gcov-tool.c (print_version): Ditto.
15533 * gengtype.c (create_file): Ditto.
15534 * doc/cpp.texi: Bump @copying's copyright year.
15535 * doc/cppinternals.texi: Ditto.
15536 * doc/gcc.texi: Ditto.
15537 * doc/gccint.texi: Ditto.
15538 * doc/gcov.texi: Ditto.
15539 * doc/install.texi: Ditto.
15540 * doc/invoke.texi: Ditto.
15542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15544 * vector-builder.h (vector_builder::m_full_nelts): Change from
15545 unsigned int to poly_uint64.
15546 (vector_builder::full_nelts): Update prototype accordingly.
15547 (vector_builder::new_vector): Likewise.
15548 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15549 (vector_builder::operator ==): Likewise.
15550 (vector_builder::finalize): Likewise.
15551 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15552 Take the number of elements as a poly_uint64 rather than an
15554 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15555 from unsigned int to poly_uint64.
15556 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15557 (vec_perm_indices::new_vector): Likewise.
15558 (vec_perm_indices::length): Likewise.
15559 (vec_perm_indices::nelts_per_input): Likewise.
15560 (vec_perm_indices::input_nelts): Likewise.
15561 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15562 number of elements per input as a poly_uint64 rather than an
15563 unsigned int. Use the original encoding for variable-length
15564 vectors, rather than clamping each individual element.
15565 For the second and subsequent elements in each pattern,
15566 clamp the step and base before clamping their sum.
15567 (vec_perm_indices::series_p): Handle polynomial element counts.
15568 (vec_perm_indices::all_in_range_p): Likewise.
15569 (vec_perm_indices_to_tree): Likewise.
15570 (vec_perm_indices_to_rtx): Likewise.
15571 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15572 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15573 (tree_vector_builder::new_binary_operation): Handle polynomial
15574 element counts. Return false if we need to know the number
15575 of elements at compile time.
15576 * fold-const.c (fold_vec_perm): Punt if the number of elements
15577 isn't known at compile time.
15579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15581 * vec-perm-indices.h (vec_perm_builder): Change element type
15582 from HOST_WIDE_INT to poly_int64.
15583 (vec_perm_indices::element_type): Update accordingly.
15584 (vec_perm_indices::clamp): Handle polynomial element_types.
15585 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15586 (vec_perm_indices::all_in_range_p): Likewise.
15587 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15589 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15590 polynomial vec_perm_indices element types.
15591 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15592 * fold-const.c (fold_vec_perm): Likewise.
15593 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15594 * tree-vect-generic.c (lower_vec_perm): Likewise.
15595 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15596 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15597 element type to HOST_WIDE_INT.
15599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15600 Alan Hayward <alan.hayward@arm.com>
15601 David Sherwood <david.sherwood@arm.com>
15603 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15604 rather than an int. Use plus_constant.
15605 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15606 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15609 Alan Hayward <alan.hayward@arm.com>
15610 David Sherwood <david.sherwood@arm.com>
15612 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15613 a HOST_WIDE_INT to a poly_int64.
15615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15616 Alan Hayward <alan.hayward@arm.com>
15617 David Sherwood <david.sherwood@arm.com>
15619 * calls.c (load_register_parameters): Cope with polynomial
15620 mode sizes. Require a constant size for BLKmode parameters
15621 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15622 forces a parameter to be padded at the lsb end in order to
15623 fill a complete number of words, require the parameter size
15624 to be ordered wrt UNITS_PER_WORD.
15626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15627 Alan Hayward <alan.hayward@arm.com>
15628 David Sherwood <david.sherwood@arm.com>
15630 * reload1.c (spill_stack_slot_width): Change element type
15631 from unsigned int to poly_uint64_pod.
15632 (alter_reg): Treat mode sizes as polynomial.
15634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15635 Alan Hayward <alan.hayward@arm.com>
15636 David Sherwood <david.sherwood@arm.com>
15638 * reload.c (complex_word_subreg_p): New function.
15639 (reload_inner_reg_of_subreg, push_reload): Use it.
15641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15642 Alan Hayward <alan.hayward@arm.com>
15643 David Sherwood <david.sherwood@arm.com>
15645 * lra-constraints.c (process_alt_operands): Reject matched
15646 operands whose sizes aren't ordered.
15647 (match_reload): Refer to this check here.
15649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15650 Alan Hayward <alan.hayward@arm.com>
15651 David Sherwood <david.sherwood@arm.com>
15653 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15654 that the mode size is in the set {1, 2, 4, 8, 16}.
15656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15657 Alan Hayward <alan.hayward@arm.com>
15658 David Sherwood <david.sherwood@arm.com>
15660 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15661 Use plus_constant instead of gen_rtx_PLUS.
15663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15664 Alan Hayward <alan.hayward@arm.com>
15665 David Sherwood <david.sherwood@arm.com>
15667 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15668 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15669 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15670 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15671 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15672 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15673 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15674 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15675 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15676 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15678 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15679 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15680 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15681 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15682 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15683 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15684 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15685 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15686 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15687 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15689 * expr.c (emit_move_resolve_push): Treat the input and result
15690 of PUSH_ROUNDING as a poly_int64.
15691 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15692 (emit_push_insn): Likewise.
15693 * lra-eliminations.c (mark_not_eliminable): Likewise.
15694 * recog.c (push_operand): Likewise.
15695 * reload1.c (elimination_effects): Likewise.
15696 * rtlanal.c (nonzero_bits1): Likewise.
15697 * calls.c (store_one_arg): Likewise. Require the padding to be
15698 known at compile time.
15700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15701 Alan Hayward <alan.hayward@arm.com>
15702 David Sherwood <david.sherwood@arm.com>
15704 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15705 Use plus_constant instead of gen_rtx_PLUS.
15707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15708 Alan Hayward <alan.hayward@arm.com>
15709 David Sherwood <david.sherwood@arm.com>
15711 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15712 rather than an int.
15714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15715 Alan Hayward <alan.hayward@arm.com>
15716 David Sherwood <david.sherwood@arm.com>
15718 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15719 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15720 via stack temporaries. Treat the mode size as polynomial too.
15722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15723 Alan Hayward <alan.hayward@arm.com>
15724 David Sherwood <david.sherwood@arm.com>
15726 * expr.c (expand_expr_real_2): When handling conversions involving
15727 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15728 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15729 as a poly_uint64 too.
15731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15732 Alan Hayward <alan.hayward@arm.com>
15733 David Sherwood <david.sherwood@arm.com>
15735 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15738 Alan Hayward <alan.hayward@arm.com>
15739 David Sherwood <david.sherwood@arm.com>
15741 * combine.c (can_change_dest_mode): Handle polynomial
15742 REGMODE_NATURAL_SIZE.
15743 * expmed.c (store_bit_field_1): Likewise.
15744 * expr.c (store_constructor): Likewise.
15745 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15746 and polynomial REGMODE_NATURAL_SIZE.
15747 (gen_lowpart_common): Likewise.
15748 * reginfo.c (record_subregs_of_mode): Likewise.
15749 * rtlanal.c (read_modify_subreg_p): Likewise.
15751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15752 Alan Hayward <alan.hayward@arm.com>
15753 David Sherwood <david.sherwood@arm.com>
15755 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15756 numbers of elements.
15758 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15759 Alan Hayward <alan.hayward@arm.com>
15760 David Sherwood <david.sherwood@arm.com>
15762 * match.pd: Cope with polynomial numbers of vector elements.
15764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15765 Alan Hayward <alan.hayward@arm.com>
15766 David Sherwood <david.sherwood@arm.com>
15768 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15769 in a POINTER_PLUS_EXPR.
15771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15772 Alan Hayward <alan.hayward@arm.com>
15773 David Sherwood <david.sherwood@arm.com>
15775 * omp-simd-clone.c (simd_clone_subparts): New function.
15776 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15777 (ipa_simd_modify_function_body): Likewise.
15779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15780 Alan Hayward <alan.hayward@arm.com>
15781 David Sherwood <david.sherwood@arm.com>
15783 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15784 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15785 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15786 (expand_vector_condition, vector_element): Likewise.
15787 (subparts_gt): New function.
15788 (get_compute_type): Use subparts_gt.
15789 (count_type_subparts): Delete.
15790 (expand_vector_operations_1): Use subparts_gt instead of
15791 count_type_subparts.
15793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15794 Alan Hayward <alan.hayward@arm.com>
15795 David Sherwood <david.sherwood@arm.com>
15797 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15798 (vect_compile_time_alias): ...this new function. Do the calculation
15799 on poly_ints rather than trees.
15800 (vect_prune_runtime_alias_test_list): Update call accordingly.
15802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15803 Alan Hayward <alan.hayward@arm.com>
15804 David Sherwood <david.sherwood@arm.com>
15806 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15808 (vect_schedule_slp_instance): Likewise.
15810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15811 Alan Hayward <alan.hayward@arm.com>
15812 David Sherwood <david.sherwood@arm.com>
15814 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15815 constant and extern definitions for variable-length vectors.
15816 (vect_get_constant_vectors): Note that the number of units
15817 is known to be constant.
15819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15820 Alan Hayward <alan.hayward@arm.com>
15821 David Sherwood <david.sherwood@arm.com>
15823 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15824 of units as polynomial. Choose between WIDE and NARROW based
15827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15828 Alan Hayward <alan.hayward@arm.com>
15829 David Sherwood <david.sherwood@arm.com>
15831 * tree-vect-stmts.c (simd_clone_subparts): New function.
15832 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15835 Alan Hayward <alan.hayward@arm.com>
15836 David Sherwood <david.sherwood@arm.com>
15838 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15839 vectors as polynomial. Use build_index_vector for
15840 IFN_GOMP_SIMD_LANE.
15842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15843 Alan Hayward <alan.hayward@arm.com>
15844 David Sherwood <david.sherwood@arm.com>
15846 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15847 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15848 for variable-length vectors.
15849 (vectorizable_mask_load_store): Treat the number of units as
15850 polynomial, asserting that it is constant if the condition has
15851 already been enforced.
15852 (vectorizable_store, vectorizable_load): Likewise.
15854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15855 Alan Hayward <alan.hayward@arm.com>
15856 David Sherwood <david.sherwood@arm.com>
15858 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15859 of units as polynomial. Punt if we can't tell at compile time
15860 which vector contains the final result.
15862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15863 Alan Hayward <alan.hayward@arm.com>
15864 David Sherwood <david.sherwood@arm.com>
15866 * tree-vect-loop.c (vectorizable_induction): Treat the number
15867 of units as polynomial. Punt on SLP inductions. Use an integer
15868 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15869 cast of such a series for variable-length floating-point
15872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15873 Alan Hayward <alan.hayward@arm.com>
15874 David Sherwood <david.sherwood@arm.com>
15876 * tree.h (build_index_vector): Declare.
15877 * tree.c (build_index_vector): New function.
15878 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15879 of units as polynomial, forcibly converting it to a constant if
15880 vectorizable_reduction has already enforced the condition.
15881 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15882 to create a {1,2,3,...} vector.
15883 (vectorizable_reduction): Treat the number of units as polynomial.
15884 Choose vectype_in based on the largest scalar element size rather
15885 than the smallest number of units. Enforce the restrictions
15888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15889 Alan Hayward <alan.hayward@arm.com>
15890 David Sherwood <david.sherwood@arm.com>
15892 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15893 number of units as polynomial.
15895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15896 Alan Hayward <alan.hayward@arm.com>
15897 David Sherwood <david.sherwood@arm.com>
15899 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15900 * target.def (autovectorize_vector_sizes): Return the vector sizes
15901 by pointer, using vector_sizes rather than a bitmask.
15902 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15903 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15904 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15906 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15907 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15908 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15909 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15910 * omp-general.c (omp_max_vf): Likewise.
15911 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15912 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
15913 * tree-vect-loop.c (vect_analyze_loop): Likewise.
15914 * tree-vect-slp.c (vect_slp_bb): Likewise.
15915 * doc/tm.texi: Regenerate.
15916 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
15918 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
15919 the vector size as a poly_uint64 rather than an unsigned int.
15920 (current_vector_size): Change from an unsigned int to a poly_uint64.
15921 (get_vectype_for_scalar_type): Update accordingly.
15922 * tree.h (build_truth_vector_type): Take the size and number of
15923 units as a poly_uint64 rather than an unsigned int.
15924 (build_vector_type): Add a temporary overload that takes
15925 the number of units as a poly_uint64 rather than an unsigned int.
15926 * tree.c (make_vector_type): Likewise.
15927 (build_truth_vector_type): Take the number of units as a poly_uint64
15928 rather than an unsigned int.
15930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15931 Alan Hayward <alan.hayward@arm.com>
15932 David Sherwood <david.sherwood@arm.com>
15934 * target.def (get_mask_mode): Take the number of units and length
15935 as poly_uint64s rather than unsigned ints.
15936 * targhooks.h (default_get_mask_mode): Update accordingly.
15937 * targhooks.c (default_get_mask_mode): Likewise.
15938 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
15939 * doc/tm.texi: Regenerate.
15941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15942 Alan Hayward <alan.hayward@arm.com>
15943 David Sherwood <david.sherwood@arm.com>
15945 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
15946 * omp-general.c (omp_max_vf): Likewise.
15947 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
15948 (expand_omp_simd): Handle polynomial safelen.
15949 * omp-low.c (omplow_simd_context): Add a default constructor.
15950 (omplow_simd_context::max_vf): Change from int to poly_uint64.
15951 (lower_rec_simd_input_clauses): Update accordingly.
15952 (lower_rec_input_clauses): Likewise.
15954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15955 Alan Hayward <alan.hayward@arm.com>
15956 David Sherwood <david.sherwood@arm.com>
15958 * tree-vectorizer.h (vect_nunits_for_cost): New function.
15959 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
15960 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
15961 (vect_analyze_slp_cost): Likewise.
15962 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
15963 (vect_model_load_cost): Likewise.
15965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15966 Alan Hayward <alan.hayward@arm.com>
15967 David Sherwood <david.sherwood@arm.com>
15969 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
15970 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
15971 from an unsigned int * to a poly_uint64_pod *.
15972 (calculate_unrolling_factor): New function.
15973 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
15975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15976 Alan Hayward <alan.hayward@arm.com>
15977 David Sherwood <david.sherwood@arm.com>
15979 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
15980 from an unsigned int to a poly_uint64.
15981 (_loop_vec_info::slp_unrolling_factor): Likewise.
15982 (_loop_vec_info::vectorization_factor): Change from an int
15984 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
15985 (vect_get_num_vectors): New function.
15986 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
15987 (vect_get_num_copies): Use vect_get_num_vectors.
15988 (vect_analyze_data_ref_dependences): Change max_vf from an int *
15989 to an unsigned int *.
15990 (vect_analyze_data_refs): Change min_vf from an int * to a
15992 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
15993 than an unsigned HOST_WIDE_INT.
15994 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
15995 (vect_analyze_data_ref_dependence): Change max_vf from an int *
15996 to an unsigned int *.
15997 (vect_analyze_data_ref_dependences): Likewise.
15998 (vect_compute_data_ref_alignment): Handle polynomial vf.
15999 (vect_enhance_data_refs_alignment): Likewise.
16000 (vect_prune_runtime_alias_test_list): Likewise.
16001 (vect_shift_permute_load_chain): Likewise.
16002 (vect_supportable_dr_alignment): Likewise.
16003 (dependence_distance_ge_vf): Take the vectorization factor as a
16004 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16005 (vect_analyze_data_refs): Change min_vf from an int * to a
16007 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16008 vfm1 as a poly_uint64 rather than an int. Make the same change
16009 for the returned bound_scalar.
16010 (vect_gen_vector_loop_niters): Handle polynomial vf.
16011 (vect_do_peeling): Likewise. Update call to
16012 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16013 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16015 * tree-vect-loop.c (vect_determine_vectorization_factor)
16016 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16017 (vect_get_known_peeling_cost): Likewise.
16018 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16019 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16020 (vect_transform_loop): Likewise. Use the lowest possible VF when
16021 updating the upper bounds of the loop.
16022 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16023 rather than an int.
16024 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16025 polynomial unroll factors.
16026 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16027 (vect_make_slp_decision): Likewise.
16028 (vect_supported_load_permutation_p): Likewise, and polynomial
16030 (vect_analyze_slp_cost): Handle polynomial vf.
16031 (vect_slp_analyze_node_operations): Likewise.
16032 (vect_slp_analyze_bb_1): Likewise.
16033 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16034 than an unsigned HOST_WIDE_INT.
16035 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16036 (vectorizable_load): Handle polynomial vf.
16037 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16039 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16042 Alan Hayward <alan.hayward@arm.com>
16043 David Sherwood <david.sherwood@arm.com>
16045 * match.pd: Handle bit operations involving three constants
16046 and try to fold one pair.
16048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16050 * tree-vect-loop-manip.c: Include gimple-fold.h.
16051 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16052 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16053 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16054 Add a path that uses a step of VF instead of 1, but disable it
16056 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16057 and niters_no_overflow parameters. Update calls to
16058 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16059 Create a new SSA name if the latter choses to use a ste other
16060 than zero, and return it via niters_vector_mult_vf_var.
16061 * tree-vect-loop.c (vect_transform_loop): Update calls to
16062 vect_do_peeling, vect_gen_vector_loop_niters and
16063 slpeel_make_loop_iterate_ntimes.
16064 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16065 (vect_gen_vector_loop_niters): Update declarations after above changes.
16067 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16069 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16070 128-bit round to integer instructions.
16071 (ceil<mode>2): Likewise.
16072 (btrunc<mode>2): Likewise.
16073 (round<mode>2): Likewise.
16075 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16077 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16078 unaligned VSX load/store on P8/P9.
16079 (expand_block_clear): Allow the use of unaligned VSX
16080 load/store on P8/P9.
16082 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16084 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16086 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16087 swap associated with both a load and a store.
16089 2018-01-02 Andrew Waterman <andrew@sifive.com>
16091 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16092 * config/riscv/riscv.md (clear_cache): Use it.
16094 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16096 * web.c: Remove out-of-date comment.
16098 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16100 * expr.c (fixup_args_size_notes): Check that any existing
16101 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16102 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16103 (emit_single_push_insn): ...here.
16105 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16107 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16108 (const_vector_encoded_nelts): New function.
16109 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16110 (const_vector_int_elt, const_vector_elt): Declare.
16111 * emit-rtl.c (const_vector_int_elt_1): New function.
16112 (const_vector_elt): Likewise.
16113 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16114 of CONST_VECTOR_ELT.
16116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16118 * expr.c: Include rtx-vector-builder.h.
16119 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16120 directly on the tree encoding.
16121 (const_vector_from_tree): Likewise.
16122 * optabs.c: Include rtx-vector-builder.h.
16123 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16124 sequence of "u" values.
16125 * vec-perm-indices.c: Include rtx-vector-builder.h.
16126 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16127 directly on the vec_perm_indices encoding.
16129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16131 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16132 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16133 * rtx-vector-builder.h: New file.
16134 * rtx-vector-builder.c: Likewise.
16135 * rtl.h (rtx_def::u2): Add a const_vector field.
16136 (CONST_VECTOR_NPATTERNS): New macro.
16137 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16138 (CONST_VECTOR_DUPLICATE_P): Likewise.
16139 (CONST_VECTOR_STEPPED_P): Likewise.
16140 (CONST_VECTOR_ENCODED_ELT): Likewise.
16141 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16142 (unwrap_const_vec_duplicate): Likewise.
16143 (const_vec_series_p): Check for a non-duplicated vector encoding.
16144 Say that the function only returns true for integer vectors.
16145 * emit-rtl.c: Include rtx-vector-builder.h.
16146 (gen_const_vec_duplicate_1): Delete.
16147 (gen_const_vector): Call gen_const_vec_duplicate instead of
16148 gen_const_vec_duplicate_1.
16149 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16150 (gen_const_vec_duplicate): Use rtx_vector_builder.
16151 (gen_const_vec_series): Likewise.
16152 (gen_rtx_CONST_VECTOR): Likewise.
16153 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16154 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16155 Build a new vector rather than modifying a CONST_VECTOR in-place.
16156 (handle_special_swappables): Update call accordingly.
16157 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16158 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16159 Build a new vector rather than modifying a CONST_VECTOR in-place.
16160 (handle_special_swappables): Update call accordingly.
16162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16164 * simplify-rtx.c (simplify_const_binary_operation): Use
16165 CONST_VECTOR_ELT instead of XVECEXP.
16167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16169 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16170 the selector elements to be different from the data elements
16171 if the selector is a VECTOR_CST.
16172 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16173 ssizetype for the selector.
16175 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16177 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16178 before testing each element individually.
16179 * tree-vect-generic.c (lower_vec_perm): Likewise.
16181 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16183 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16184 * selftest-run-tests.c (selftest::run_tests): Call it.
16185 * vector-builder.h (vector_builder::operator ==): New function.
16186 (vector_builder::operator !=): Likewise.
16187 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16188 (vec_perm_indices::all_from_input_p): New function.
16189 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16190 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16191 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16192 instead of reading the VECTOR_CST directly. Detect whether both
16193 vector inputs are the same before constructing the vec_perm_indices,
16194 and update the number of inputs argument accordingly. Use the
16195 utility functions added above. Only construct sel2 if we need to.
16197 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16199 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16200 the broadcast of the low byte.
16201 (expand_mult_highpart): Use an explicit encoding for the permutes.
16202 * optabs-query.c (can_mult_highpart_p): Likewise.
16203 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16204 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16205 (vectorizable_bswap): Likewise.
16206 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16207 explicit encoding for the power-of-2 permutes.
16208 (vect_permute_store_chain): Likewise.
16209 (vect_grouped_load_supported): Likewise.
16210 (vect_permute_load_chain): Likewise.
16212 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16214 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16215 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16216 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16217 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16218 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16219 (vect_gen_perm_mask_any): Likewise.
16221 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16223 * int-vector-builder.h: New file.
16224 * vec-perm-indices.h: Include int-vector-builder.h.
16225 (vec_perm_indices): Redefine as an int_vector_builder.
16226 (auto_vec_perm_indices): Delete.
16227 (vec_perm_builder): Redefine as a stand-alone class.
16228 (vec_perm_indices::vec_perm_indices): New function.
16229 (vec_perm_indices::clamp): Likewise.
16230 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16231 (vec_perm_indices::new_vector): New function.
16232 (vec_perm_indices::new_expanded_vector): Update for new
16233 vec_perm_indices class.
16234 (vec_perm_indices::rotate_inputs): New function.
16235 (vec_perm_indices::all_in_range_p): Operate directly on the
16236 encoded form, without computing elided elements.
16237 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16238 encoding. Update for new vec_perm_indices class.
16239 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16240 the given vec_perm_builder.
16241 (expand_vec_perm_var): Update vec_perm_builder constructor.
16242 (expand_mult_highpart): Use vec_perm_builder instead of
16243 auto_vec_perm_indices.
16244 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16245 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16246 or double series encoding as appropriate.
16247 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16248 vec_perm_indices instead of auto_vec_perm_indices.
16249 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16250 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16251 (vect_permute_store_chain): Likewise.
16252 (vect_grouped_load_supported): Likewise.
16253 (vect_permute_load_chain): Likewise.
16254 (vect_shift_permute_load_chain): Likewise.
16255 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16256 (vect_transform_slp_perm_load): Likewise.
16257 (vect_schedule_slp_instance): Likewise.
16258 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16259 (vectorizable_mask_load_store): Likewise.
16260 (vectorizable_bswap): Likewise.
16261 (vectorizable_store): Likewise.
16262 (vectorizable_load): Likewise.
16263 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16264 vec_perm_indices instead of auto_vec_perm_indices. Use
16265 tree_to_vec_perm_builder to read the vector from a tree.
16266 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16267 vec_perm_builder instead of a vec_perm_indices.
16268 (have_whole_vector_shift): Use vec_perm_builder and
16269 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16270 truncation to calc_vec_perm_mask_for_shift.
16271 (vect_create_epilog_for_reduction): Likewise.
16272 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16273 from auto_vec_perm_indices to vec_perm_indices.
16274 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16275 instead of changing individual elements.
16276 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16277 the vector in d.perm.
16278 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16279 from auto_vec_perm_indices to vec_perm_indices.
16280 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16281 instead of changing individual elements.
16282 (arm_vectorize_vec_perm_const): Use new_vector to install
16283 the vector in d.perm.
16284 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16285 Update vec_perm_builder constructor.
16286 (rs6000_expand_interleave): Likewise.
16287 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16288 (rs6000_expand_interleave): Likewise.
16290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16292 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16293 to qimode could truncate the indices.
16294 * optabs.c (expand_vec_perm_var): Likewise.
16296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16298 * Makefile.in (OBJS): Add vec-perm-indices.o.
16299 * vec-perm-indices.h: New file.
16300 * vec-perm-indices.c: Likewise.
16301 * target.h (vec_perm_indices): Replace with a forward class
16303 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16304 * optabs.h: Include vec-perm-indices.h.
16305 (expand_vec_perm): Delete.
16306 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16307 (expand_vec_perm_const): Declare.
16308 * target.def (vec_perm_const_ok): Replace with...
16309 (vec_perm_const): ...this new hook.
16310 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16311 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16312 * doc/tm.texi: Regenerate.
16313 * optabs.def (vec_perm_const): Delete.
16314 * doc/md.texi (vec_perm_const): Likewise.
16315 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16316 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16317 expand_vec_perm for constant permutation vectors. Assert that
16318 the mode of variable permutation vectors is the integer equivalent
16319 of the mode that is being permuted.
16320 * optabs-query.h (selector_fits_mode_p): Declare.
16321 * optabs-query.c: Include vec-perm-indices.h.
16322 (selector_fits_mode_p): New function.
16323 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16324 is defined, instead of checking whether the vec_perm_const_optab
16325 exists. Use targetm.vectorize.vec_perm_const instead of
16326 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16327 fit in the vector mode before using a variable permute.
16328 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16329 vec_perm_indices instead of an rtx.
16330 (expand_vec_perm): Replace with...
16331 (expand_vec_perm_const): ...this new function. Take the selector
16332 as a vec_perm_indices rather than an rtx. Also take the mode of
16333 the selector. Update call to shift_amt_for_vec_perm_mask.
16334 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16335 Use vec_perm_indices::new_expanded_vector to expand the original
16336 selector into bytes. Check whether the indices fit in the vector
16337 mode before using a variable permute.
16338 (expand_vec_perm_var): Make global.
16339 (expand_mult_highpart): Use expand_vec_perm_const.
16340 * fold-const.c: Includes vec-perm-indices.h.
16341 * tree-ssa-forwprop.c: Likewise.
16342 * tree-vect-data-refs.c: Likewise.
16343 * tree-vect-generic.c: Likewise.
16344 * tree-vect-loop.c: Likewise.
16345 * tree-vect-slp.c: Likewise.
16346 * tree-vect-stmts.c: Likewise.
16347 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16349 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16350 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16351 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16352 (aarch64_vectorize_vec_perm_const): ...this new function.
16353 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16354 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16355 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16356 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16357 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16358 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16359 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16361 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16362 check for NEON modes.
16363 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16364 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16365 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16366 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16368 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16369 the old VEC_PERM_CONST conditions.
16370 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16371 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16372 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16373 (ia64_vectorize_vec_perm_const_ok): Merge into...
16374 (ia64_vectorize_vec_perm_const): ...this new function.
16375 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16376 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16377 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16378 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16379 * config/mips/mips.c (mips_expand_vec_perm_const)
16380 (mips_vectorize_vec_perm_const_ok): Merge into...
16381 (mips_vectorize_vec_perm_const): ...this new function.
16382 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16383 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16384 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16385 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16386 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16387 (rs6000_expand_vec_perm_const): Delete.
16388 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16390 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16391 (altivec_expand_vec_perm_const_le): Take each operand individually.
16392 Operate on constant selectors rather than rtxes.
16393 (altivec_expand_vec_perm_const): Likewise. Update call to
16394 altivec_expand_vec_perm_const_le.
16395 (rs6000_expand_vec_perm_const): Delete.
16396 (rs6000_vectorize_vec_perm_const_ok): Delete.
16397 (rs6000_vectorize_vec_perm_const): New function.
16398 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16399 an element count and rtx array.
16400 (rs6000_expand_extract_even): Update call accordingly.
16401 (rs6000_expand_interleave): Likewise.
16402 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16403 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16404 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16405 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16406 (rs6000_expand_vec_perm_const): Delete.
16407 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16408 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16409 (altivec_expand_vec_perm_const_le): Take each operand individually.
16410 Operate on constant selectors rather than rtxes.
16411 (altivec_expand_vec_perm_const): Likewise. Update call to
16412 altivec_expand_vec_perm_const_le.
16413 (rs6000_expand_vec_perm_const): Delete.
16414 (rs6000_vectorize_vec_perm_const_ok): Delete.
16415 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16416 reference to the SPE evmerge intructions.
16417 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16418 an element count and rtx array.
16419 (rs6000_expand_extract_even): Update call accordingly.
16420 (rs6000_expand_interleave): Likewise.
16421 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16422 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16424 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16428 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16429 vector mode and that that mode matches the mode of the data
16431 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16432 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16433 directly using expand_vec_perm_1 when forcing selectors into
16435 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16439 * optabs-query.h (can_vec_perm_p): Delete.
16440 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16441 * optabs-query.c (can_vec_perm_p): Split into...
16442 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16443 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16444 particular selector is valid.
16445 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16446 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16447 (vect_grouped_load_supported): Likewise.
16448 (vect_shift_permute_load_chain): Likewise.
16449 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16450 (vect_transform_slp_perm_load): Likewise.
16451 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16452 (vectorizable_bswap): Likewise.
16453 (vect_gen_perm_mask_checked): Likewise.
16454 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16455 implementations of variable permutation vectors into account
16456 when deciding which selector to use.
16457 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16458 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16459 with a false third argument.
16460 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16461 to test whether the constant selector is valid and can_vec_perm_var_p
16462 to test whether a variable selector is valid.
16464 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16466 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16467 * optabs-query.c (can_vec_perm_p): Likewise.
16468 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16469 instead of vec_perm_indices.
16470 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16471 (vect_gen_perm_mask_checked): Likewise,
16472 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16473 (vect_gen_perm_mask_checked): Likewise,
16475 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16477 * optabs-query.h (qimode_for_vec_perm): Declare.
16478 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16479 (qimode_for_vec_perm): ...this new function.
16480 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16482 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16484 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16485 does not have a conditional at the top.
16487 2018-01-02 Richard Biener <rguenther@suse.de>
16489 * ipa-inline.c (big_speedup_p): Fix expression.
16491 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16494 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16497 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16501 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16502 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16503 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16504 cond_taken_branch_cost 3->4.
16506 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16508 PR tree-optimization/83581
16509 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16510 TODO_cleanup_cfg if any changes have been made.
16512 PR middle-end/83608
16513 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16514 convert_modes if target mode has the right side, but different mode
16517 PR middle-end/83609
16518 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16519 last argument when extracting from CONCAT. If either from_real or
16520 from_imag is NULL, use expansion through memory. If result is not
16521 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16522 the parts directly to inner mode, if even that fails, use expansion
16525 PR middle-end/83623
16526 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16527 check for bswap in mode rather than HImode and use that in expand_unop
16530 Copyright (C) 2018 Free Software Foundation, Inc.
16532 Copying and distribution of this file, with or without modification,
16533 are permitted in any medium without royalty provided the copyright
16534 notice and this notice are preserved.