1 2018-06-13 Jeff Law <law@redhat.com>
3 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4 integer 0 for argument to print_rtl_with_bb.
5 (rl78_reorg): Likewise.
7 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
10 from rtx to rtx_insn *.
11 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
13 (add_sched_insns_for_speculation): Likewise for local "target",
14 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
16 from rtx_insn ** to rtx_code_label **.
17 (reorg_emit_nops): Likewise.
18 (c6x_reorg): Likewise for local "call_labels".
19 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
21 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
22 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
23 the loops over LABEL_REFS.
24 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
26 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
27 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
28 (split_branches): Strengthen local "olabel" from rtx to
29 rtx_insn *, adding a safe_as_a cast.
30 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
32 (add_insn_after): Likewise for first two params.
33 (add_insn_before): Likewise.
34 (remove_insn): Likewise for param.
35 (emit_pattern_before_noloc): Likewise for second and third params.
36 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
37 (emit_call_insn_before_noloc): Likewise.
38 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
40 (emit_barrier_before): Likewise.
41 (emit_label_before): Strengthen "label" param from "rtx" to
42 "rtx_code_label *". Strengthen "before" param from "rtx" to
44 (emit_insn_after_1): Strengthen "after" param from "rtx" to
46 (emit_pattern_after_noloc): Likewise.
47 (emit_insn_after_noloc): Likewise.
48 (emit_jump_insn_after_noloc): Likewise.
49 (emit_call_insn_after_noloc): Likewise.
50 (emit_debug_insn_after_noloc): Likewise.
51 (emit_barrier_after): Likewise.
52 (emit_label_after): Likewise for both params.
53 (emit_pattern_after_setloc): Likewise for "after" param. Convert
54 "loc" param from "int" to "location_t".
55 (emit_insn_after_setloc): Likewise.
56 (emit_jump_insn_after_setloc): Likewise.
57 (emit_call_insn_after_setloc): Likewise.
58 (emit_debug_insn_after_setloc): Likewise.
59 (emit_pattern_before_setloc): Likewise for "before" param. Convert
60 "loc" param from "int" to "location_t".
61 (emit_pattern_before): Convert NULL_RTX to NULL.
62 (emit_insn_before_setloc): Convert "loc" param from "int" to
64 (emit_jump_insn_before_setloc): Likewise.
65 (emit_call_insn_before_setloc): Likewise.
66 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
67 rtx_insn *. Convert "loc" param from "int" to "location_t".
68 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
69 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
70 Convert 3rd param from "int" to "location_t".
71 (emit_barrier_before, emit_barrier_after, next_real_insn):
72 Strengthen param from rtx to rtx_insn *.
73 (emit_label_before): Strengthen 1st param from "rtx" to
74 "rtx_code_label *". Strengthen 2nd param from "rtx" to
76 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
77 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
78 Strengthen 2nd param from "rtx" to "rtx_insn *".
79 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
80 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
81 Likewise. Convert 3rd param from "int" to "location_t".
82 (emit_label_after): Strengthen 1st param from "rtx" to
84 (next_real_insn, remove_insn): Strengthen param from "rtx" to
86 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
87 from "rtx" to "rtx_insn *".
89 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
91 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
92 bodies streamed in with -Q.
93 * dumpfile.c (dump_files): Add lto-stream-out dump file.
94 * dumpfile.h (tree_dump_index): Add lto_stream_out.
95 * gimple-streamer-out.c: Include gimple-pretty-print.h
96 (output_bb): Dump stmts streamed.
97 * lto-section-out.c: Include print-tree.h
98 (lto_begin_section): Dump sections created.
99 (lto_output_decl_index): Dump decl encoded.
100 * lto-streamer-out.c: Include print-tree.h
101 (create_output_block): Dump output block created.
102 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
103 (output_function): Dump function output.
104 (output_constructor): Dump constructor streamed.
105 (write_global_stream): Output indexes encoded.
106 (produce_asm_for_decls): Dump streams encoded.
107 * lto-streamer.c (streamer_dump_file): New global var.
108 * lto-streamer.h (streamer_dump_file): Declare.
109 * passes.c (ipa_write_summaries): Initialize streamer dump.
110 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
113 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
116 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
117 offsets for register save directives. Emit a second batch of save
118 directives, if need be, when the function accesses prior frames.
120 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
122 * config/arc/fpu.md (fmasf4): Force operand to register.
125 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
127 * config/arc/arc-protos.h (arc_pad_return): Remove.
128 * config/arc/arc.c (machine_function): Remove force_short_suffix
130 (arc_print_operand): Adjust printing of '&'.
131 (arc_verify_short): Remove conditional printing of short suffix.
132 (arc_final_prescan_insn): Remove reference to size_reason.
133 (pad_return): New function.
134 (arc_reorg): Call pad_return.
135 (arc_pad_return): Remove.
136 (arc_init_machine_status): Remove reference to force_short_suffix.
137 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
138 (attr length): When attribute iscompact is true force to 2
139 regardless; in the case of maybe check if we want to force the
140 instruction to have 4 bytes length.
141 (nopv): Change it to generate 4 byte long nop as well.
142 (blockage): New pattern.
143 (simple_return): Remove call to arc_pad_return.
144 (p_return_i): Likewise.
146 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
148 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
150 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
152 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
155 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
157 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
158 for ARC700 and ARCv2.
160 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
163 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
164 operands[2] instead of operands[1].
167 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
169 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
170 case, check whether the outer register overlaps an unallocatable
171 register, not just whether it fits the required class.
173 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
175 * poly-int.h (can_div_trunc_p): Add new overload in which all values
177 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
178 (memrefs_conflict_p): Likewise.
179 (init_alias_analysis): Likewise.
180 * cfgexpand.c (expand_debug_expr): Likewise.
181 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
182 * cse.c (fold_rtx): Likewise.
183 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
184 * expr.c (emit_block_move_hints): Likewise.
185 (clear_storage_hints, push_block, emit_push_insn): Likewise.
186 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
187 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
188 (emit_group_store): Likewise.
189 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
190 to read the PRE/POST_MODIFY increment.
191 * calls.c (store_one_arg): Use strip_offset.
192 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
194 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
196 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
197 (simplify_binary_operation_1): Extend CONST_INT handling to
199 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
200 than a HOST_WIDE_INT.
201 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
203 (adjust_mems, add_stores): Update accodingly.
204 (vt_canonicalize_addr): Track polynomial offsets.
205 (emit_note_insn_var_location): Likewise.
206 (vt_add_function_parameter): Likewise.
207 (vt_initialize): Likewise.
209 2018-06-12 Jeff Law <law@redhat.com>
211 * config.gcc (alpha*-*-freebsd*): Remove.
212 * config/alpha/freebsd.h: Remove.
214 2018-06-12 David Malcolm <dmalcolm@redhat.com>
217 * spellcheck-tree.c (levenshtein_distance): Rename to...
218 (get_edit_distance): ...this, and update for underlying renaming.
219 * spellcheck-tree.h (levenshtein_distance): Rename to...
220 (get_edit_distance): ...this.
221 * spellcheck.c (levenshtein_distance): Rename to...
222 (get_edit_distance): ...this. Convert from Levenshtein distance
223 to Damerau-Levenshtein distance by supporting transpositions of
224 adjacent characters. Rename "v1" to "v_next" and "v0" to
226 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
227 (selftest::test_edit_distance_unit_test_oneway): ...this, and
228 update for underlying renaming.
229 (selftest::levenshtein_distance_unit_test): Rename to...
230 (selftest::test_get_edit_distance_unit): ...this, and update for
232 (selftest::test_find_closest_string): Add example from PR 69968
233 where transposition helps
234 (selftest::test_metric_conditions): Update for renaming.
235 (selftest::test_metric_conditions): Likewise.
236 (selftest::spellcheck_c_tests): Likewise.
237 * spellcheck.h (levenshtein_distance): Rename both overloads to...
238 (get_edit_distance): ...this.
239 (best_match::consider): Update for renaming.
241 2018-06-12 Martin Sebor <msebor@redhat.com>
243 PR tree-optimization/85259
244 * builtins.c (compute_objsize): Handle constant offsets.
245 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
246 true iff a warning has been issued.
247 * gimple.h (gimple_nonartificial_location): New function.
248 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
249 gimple_nonartificial_location and handle -Wno-system-headers.
250 (handle_builtin_stxncpy): Same.
252 2018-06-12 Martin Sebor <msebor@redhat.com>
255 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
257 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
259 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
260 BUILTIN_VEC_XST entries for pointer to double and long long.
262 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
265 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
267 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
270 2018-06-12 Martin Liska <mliska@suse.cz>
272 * doc/options.texi: Document IntegerRange.
274 2018-06-12 Martin Liska <mliska@suse.cz>
276 * config/i386/i386.opt: Make MPX-related options as Deprecated.
277 * opt-functions.awk: Handle Deprecated flag.
278 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
280 (read_cmdline_option): Report warning for a deprecated option.
281 * opts.h (struct cl_option): Add new field cl_deprecated.
282 (CL_ERR_DEPRECATED): New.
284 2018-06-12 Martin Liska <mliska@suse.cz>
286 * doc/options.texi: Document Deprecated option flag.
288 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
290 * config/arc/arc-arch.h (arc_extras): New enum.
291 (arc_cpu_t):Add field extra.
292 (arc_cpu_types): Consider the extras.
293 * config/arc/arc-cpus.def: Add extras info.
294 * config/arc/arc-opts.h (processor_type): Consider extra field.
295 * config/arc/arc.c (arc_override_options): Handle extra field.
297 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
299 * config/arc/arc-arch.h: Update ARC_OPTX macro.
300 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
302 * config/arc/arc.c (arc_init): Update pic warning.
303 (irq_range): Update irq range parsing warnings.
304 (arc_override_options): Update various warning messages.
305 (arc_handle_aux_attribute): Likewise.
307 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
309 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
311 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
313 * doc/sourcebuild.texi: Document usage of line number 0 in verify
314 compiler messages directives.
316 2018-06-12 Matthew Fortune <mfortune@gmail.com>
318 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
319 * config/mips/mips-tables.opt: Regenerate.
320 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
322 * doc/invoke.texi: Document -march=i6500.
324 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
326 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
327 (i6400_gpmul): Add cpu_unit.
328 (i6400_gpdiv): Likewise.
329 (i6400_msa_add_d): Update reservations.
330 (i6400_msa_int_add) Likewise.
331 (i6400_msa_short_logic3) Likewise.
332 (i6400_msa_short_logic2) Likewise.
333 (i6400_msa_short_logic) Likewise.
334 (i6400_msa_move) Likewise.
335 (i6400_msa_cmp) Likewise.
336 (i6400_msa_short_float2) Likewise.
337 (i6400_msa_div_d) Likewise.
338 (i6400_msa_long_logic1) Likewise.
339 (i6400_msa_long_logic2) Likewise.
340 (i6400_msa_mult) Likewise.
341 (i6400_msa_long_float2) Likewise.
342 (i6400_msa_long_float4) Likewise.
343 (i6400_msa_long_float5) Likewise.
344 (i6400_msa_long_float8) Likewise.
345 (i6400_fpu_fadd): Include frint type.
346 (i6400_fpu_store): New define_insn_reservation.
347 (i6400_fpu_load): Likewise.
348 (i6400_fpu_move): Likewise.
349 (i6400_fpu_fcmp): Likewise.
350 (i6400_fpu_fmadd): Likewise.
351 (i6400_int_mult): Include imul3nc type and update reservation.
352 (i6400_int_div): Include idiv3 type and update reservation.
353 (i6400_int_load): Update to check type not move_type.
354 (i6400_int_store): Likewise.
355 (i6400_int_prefetch): Set zero latency.
357 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
359 * gcc.c: Document new %@{...} sequence.
360 (LINK_COMMAND_SPEC): Use it for the -L switches.
361 (cpp_unique_options): Use it for the -I switches.
362 (at_file_argbuf): New global variable.
363 (in_at_file): Likewise.
364 (alloc_args): Create at_file_argbuf.
365 (clear_args): Truncate at_file_argbuf.
366 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
367 (open_at_file): New function.
368 (close_at_file): Likewise.
369 (create_at_file): Delete.
370 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
373 (validate_switches_from_spec): Deal with %@{...} sequence.
374 (validate_switches): Likewise.
375 (driver::finalize): Call clear_args.
377 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
379 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
381 2018-06-11 Martin Sebor <msebor@redhat.com>
383 * doc/invoke.texi (-Wall): List -Wc++17-compat.
384 (Wno-class-memaccess): Add @opindex.
385 (Wno-templates, Wno-multiple-inheritance): Same.
386 (Wno-virtual-inheritance, Wno-namespaces): Same.
387 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
388 (Wno-format-overflow, Wno-format-truncation): Same.
389 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
390 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
391 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
392 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
393 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
394 (Wno-misspelled-isr): Same.
396 2018-06-11 Martin Sebor <msebor@redhat.com>
398 * PR tree-optimization/86083
399 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
401 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
403 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
405 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
408 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
409 on the correct operand.
410 (*movdi_internal64): Ditto.
412 2018-06-11 Martin Liska <mliska@suse.cz>
414 PR tree-optimization/86089
415 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
417 2018-06-11 Julia Koval <julia.koval@intel.com>
419 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
420 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
421 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
423 2018-06-11 Olivier Hainque <hainque@adacore.com>
425 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
426 for Ada with strict dwarf2.
428 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
431 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
434 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
436 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
438 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
440 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
443 2018-06-08 Martin Liska <mliska@suse.cz>
445 * tree-cfg.h (debug_function): Fix argument type to match
448 2018-06-08 Martin Liska <mliska@suse.cz>
450 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
451 Remove usage of MPX-related (and removed) fields.
452 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
454 2018-06-08 David Malcolm <dmalcolm@redhat.com>
456 * cfg.c (debug): Use TDF_NONE rather than 0.
457 * cfghooks.c (debug): Likewise.
458 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
459 (struct dump_option_value_info): Convert to...
460 (struct kv_pair): ...this template type.
461 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
463 (optinfo_verbosity_options): Likewise.
464 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
466 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
467 than int for "optgroup_flags" param.
468 (dump_generic_expr_loc): Use dump_flags_t rather than int for
470 (dump_dec): Likewise.
471 (dump_finish): Use TDF_NONE rather than 0.
472 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
473 rather than int for "optgroup_flags" param. Use TDF_NONE rather
474 than 0. Update for change to option_ptr.
475 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
476 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
477 0. Update for changes to optinfo_verbosity_options and
479 (opt_info_switch_p): Convert optgroup_flags from int to
481 (dump_basic_block): Use dump_flags_t rather than int
482 for "dump_kind" param.
483 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
484 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
485 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
486 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
487 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
488 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
489 TDF_NONE): Convert from macros to...
490 (enum dump_flag): ...this new enum.
491 (dump_flags_t): Update to use enum.
492 (operator|, operator&, operator~, operator|=, operator&=):
493 Implement for dump_flags_t.
494 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
495 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
496 Convert from macros to...
497 (enum optgroup_flag): ...this new enum.
498 (optgroup_flags_t): New typedef.
499 (operator|, operator|=): Implement for optgroup_flags_t.
500 (struct dump_file_info): Convert field "alt_flags" to
501 dump_flags_t. Convert field "optgroup_flags" to
503 (dump_basic_block): Use dump_flags_t rather than int for param.
504 (dump_generic_expr_loc): Likewise.
505 (dump_dec): Likewise.
506 (dump_register): Convert param "optgroup_flags" to
508 (opt_info_enable_passes): Likewise.
509 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
511 * gimple-pretty-print.c (debug): Likewise.
512 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
513 (merged_store_group::apply_stores): Likewise.
514 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
515 * gimple.c (verify_gimple_pp): Likewise.
516 * graphite-poly.c (print_pbb_body): Likewise.
517 * passes.c (pass_manager::register_one_dump_file): Convert
518 local "optgroup_flags" to optgroup_flags_t.
519 * print-tree.c (print_node): Use TDF_NONE rather than 0.
521 (debug_body): Likewise.
522 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
524 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
526 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
527 (convert_mult_to_fma): Likewise.
528 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
529 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
530 * tree-vect-data-refs.c (dump_lower_bound): Convert param
531 "dump_kind" to dump_flags_t.
533 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
535 * config/rs6000/rs6000.c (min, max): Delete.
537 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
539 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
540 -mabi=spe and -mabi=no-spe.
542 2018-06-08 Martin Liska <mliska@suse.cz>
544 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
545 where we expect an existing summary.
547 2018-06-08 Martin Liska <mliska@suse.cz>
549 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
550 * ipa-inline.h (estimate_edge_growth): Likewise.
552 2018-06-08 Martin Liska <mliska@suse.cz>
554 * cgraph.c (function_version_hasher::hash): Use
555 cgraph_node::get_uid ().
556 (function_version_hasher::equal):
557 * cgraph.h (cgraph_node::get_uid): New method.
558 * ipa-inline.c (update_caller_keys): Use
559 cgraph_node::get_uid ().
560 (update_callee_keys): Likewise.
561 * ipa-utils.c (searchc): Likewise.
562 (ipa_reduced_postorder): Likewise.
563 * lto-cgraph.c (input_node): Likewise.
564 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
565 * symbol-summary.h (symtab_insertion): Likewise.
566 (symtab_removal): Likewise.
567 (symtab_duplication): Likewise.
568 * tree-pretty-print.c (dump_function_header): Likewise.
569 * tree-sra.c (convert_callers_for_node): Likewise.
571 2018-06-08 Martin Liska <mliska@suse.cz>
573 * cgraph.c (symbol_table::create_edge): Always assign a new
575 (symbol_table::free_edge): Do not recycle numbers.
576 * cgraph.h (cgraph_edge::get): New method.
577 * symbol-summary.h (symtab_removal): Use it.
578 (symtab_duplication): Likewise.
579 (call_summary::hashable_uid): Remove.
581 2018-06-08 Martin Liska <mliska@suse.cz>
583 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
584 (initialize_growth_caches): Remove.
585 (free_growth_caches): Likewise.
586 (do_estimate_edge_time): Use edge_growth_cache.
587 (do_estimate_edge_size): Likewise.
588 (do_estimate_edge_hints): Likewise.
589 * ipa-inline.c (reset_edge_caches): Likewise.
590 (recursive_inlining): Likewise.
591 (inline_small_functions): Likewise.
592 * ipa-inline.h (initialize_growth_caches): Remove.
593 (estimate_edge_size): Likewise.
594 (estimate_edge_time): Likewise.
595 (estimate_edge_hints): Likewise.
596 (reset_edge_growth_cache): Likewise.
597 * symbol-summary.h (call_summary::remove): New method.
599 2018-06-08 Martin Liska <mliska@suse.cz>
601 * ipa-cp.c (class edge_clone_summary): New summary.
602 (grow_edge_clone_vectors): Remove.
603 (ipcp_edge_duplication_hook): Remove.
604 (class edge_clone_summary_t): New call_summary class.
605 (ipcp_edge_removal_hook): Remove.
606 (edge_clone_summary_t::duplicate): New function.
607 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
608 (create_specialized_node): Likewise.
609 (ipcp_driver): Initialize edge_clone_summaries and do not
612 2018-06-08 Martin Liska <mliska@suse.cz>
614 * symbol-summary.h (get): New function.
615 (call_summary::m_initialize_when_cloning): New class member.
617 2018-06-08 Martin Liska <mliska@suse.cz>
619 * cgraph.c (cgraph_node::remove): Do not recycle uid.
620 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
621 (symbol_table::allocate_cgraph_symbol): Do not set uid.
622 * passes.c (uid_hash_t): Record removed_nodes by their uids.
623 (remove_cgraph_node_from_order): Use the removed_nodes set.
624 (do_per_function_toporder): Likwise.
625 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
626 instead of summary_uid.
627 (symtab_removal): Likewise.
628 (symtab_duplication): Likewise.
630 2018-06-08 Martin Liska <mliska@suse.cz>
632 * ipa-cp.c (ipcp_store_bits_results): Use
633 ipcp_transformation_sum.
634 (ipcp_store_vr_results): Likewise.
635 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
637 (ipcp_transformation_initialize): ... this.
638 (ipa_set_node_agg_value_chain):
639 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
640 (write_ipcp_transformation_info): Likewise.
641 (read_ipcp_transformation_info): Likewise.
642 (ipcp_update_bits): Likewise.
643 (ipcp_update_vr): Likewise.
644 (ipcp_transform_function): Likewise.
645 * ipa-prop.h: Rename ipcp_transformation_summary to
647 (class ipcp_transformation_t): New function summary.
648 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
649 (ipa_get_agg_replacements_for_node): Likewise.
651 2018-06-08 Martin Liska <mliska@suse.cz>
653 * ipa-pure-const.c (struct funct_state_d): Do it class instead
655 (class funct_state_summary_t): New function_summary class.
656 (has_function_state): Remove.
657 (get_function_state): Likewise.
658 (set_function_state): Likewise.
659 (add_new_function): Likewise.
660 (funct_state_summary_t::insert): New function.
661 (duplicate_node_data): Remove.
662 (remove_node_data): Remove.
663 (funct_state_summary_t::duplicate): New function.
664 (register_hooks): Create new funct_state_summaries.
665 (pure_const_generate_summary): Use it.
666 (pure_const_write_summary): Likewise.
667 (pure_const_read_summary): Likewise.
668 (propagate_pure_const): Likewise.
669 (propagate_nothrow): Likewise.
670 (dump_malloc_lattice): Likewise.
671 (propagate_malloc): Likewise.
672 (execute): Do not register hooks, just remove summary
674 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
677 2018-06-08 Martin Liska <mliska@suse.cz>
679 * ipa-reference.c (remove_node_data): Remove.
680 (duplicate_node_data): Likewise.
681 (class ipa_ref_var_info_summary_t): New class.
682 (class ipa_ref_opt_summary_t): Likewise.
683 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
684 (get_reference_optimization_summary): Use
685 ipa_ref_opt_sum_summaries.
686 (set_reference_vars_info): Remove.
687 (set_reference_optimization_summary): Likewise.
688 (ipa_init): Create summaries.
689 (init_function_info): Use function summary.
690 (ipa_ref_opt_summary_t::duplicate): New function.
691 (ipa_ref_opt_summary_t::remove): New function.
692 (get_read_write_all_from_node): Fix GNU coding style.
693 (propagate): Use function summary.
694 (write_node_summary_p): Fix GNU coding style.
695 (stream_out_bitmap): Likewise.
696 (ipa_reference_read_optimization_summary): Use function summary.
697 (ipa_reference_c_finalize): Do not release hooks.
699 2018-06-08 Martin Liska <mliska@suse.cz>
701 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
702 (analyze_function_body): Extract multiple calls of get_create.
703 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
704 * ipa-inline.c (recursive_inlining): Use ::get method.
705 * ipa-inline.h (estimate_edge_growth): Likewise.
707 2018-06-08 Martin Liska <mliska@suse.cz>
709 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
711 (hsa_function_summary::hsa_function_summary): Use the new enum
713 (hsa_gpu_implementation_p): Use hsa_summaries::get.
714 * hsa-gen.c (hsa_get_host_function): Likewise.
715 (get_brig_function_name): Likewise.
716 * ipa-hsa.c (process_hsa_functions): Likewise.
717 (ipa_hsa_write_summary): Likewise.
718 * symbol-summary.h (symtab_duplication): Use ::get function/
721 2018-06-08 Martin Liska <mliska@suse.cz>
723 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
725 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
726 (hsa_register_kernel): Likewise.
727 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
728 * hsa-gen.c (hsa_get_host_function): Likewise.
729 (get_brig_function_name): Likewise.
730 (generate_hsa): Likewise.
731 (pass_gen_hsail::execute): Likewise.
732 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
733 (devirtualization_time_bonus): Likewise.
734 (ipcp_propagate_stage): Likewise.
735 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
736 (edge_set_predicate): Likewise.
737 (evaluate_conditions_for_known_args): Likewise.
738 (evaluate_properties_for_edge): Likewise.
739 (ipa_fn_summary::reset): Likewise.
740 (ipa_fn_summary_t::duplicate): Likewise.
741 (dump_ipa_call_summary): Likewise.
742 (ipa_dump_fn_summary): Likewise.
743 (analyze_function_body): Likewise.
744 (compute_fn_summary): Likewise.
745 (estimate_edge_devirt_benefit): Likewise.
746 (estimate_edge_size_and_time): Likewise.
747 (estimate_calls_size_and_time): Likewise.
748 (estimate_node_size_and_time): Likewise.
749 (inline_update_callee_summaries): Likewise.
750 (remap_edge_change_prob): Likewise.
751 (remap_edge_summaries): Likewise.
752 (ipa_merge_fn_summary_after_inlining): Likewise.
753 (ipa_update_overall_fn_summary): Likewise.
754 (read_ipa_call_summary): Likewise.
755 (inline_read_section): Likewise.
756 (write_ipa_call_summary): Likewise.
757 (ipa_fn_summary_write): Likewise.
758 (ipa_free_fn_summary): Likewise.
759 * ipa-hsa.c (process_hsa_functions): Likewise.
760 (ipa_hsa_write_summary): Likewise.
761 (ipa_hsa_read_section): Likewise.
762 * ipa-icf.c (sem_function::merge): Likewise.
763 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
764 (do_estimate_edge_time): Likewise.
765 (estimate_size_after_inlining): Likewise.
766 (estimate_growth): Likewise.
767 (growth_likely_positive): Likewise.
768 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
769 (inline_call): Likewise.
770 * ipa-inline.c (caller_growth_limits): Likewise.
771 (can_inline_edge_p): Likewise.
772 (can_inline_edge_by_limits_p): Likewise.
773 (compute_uninlined_call_time): Likewise.
774 (compute_inlined_call_time): Likewise.
775 (want_inline_small_function_p): Likewise.
776 (edge_badness): Likewise.
777 (update_caller_keys): Likewise.
778 (update_callee_keys): Likewise.
779 (recursive_inlining): Likewise.
780 (inline_small_functions): Likewise.
781 (inline_to_all_callers_1): Likewise.
782 (dump_overall_stats): Likewise.
783 (early_inline_small_functions): Likewise.
784 (early_inliner): Likewise.
785 * ipa-inline.h (estimate_edge_growth): Likewise.
786 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
787 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
788 * ipa-prop.h (IPA_NODE_REF): Likewise.
789 (IPA_EDGE_REF): Likewise.
790 * ipa-pure-const.c (malloc_candidate_p): Likewise.
791 (propagate_malloc): Likewise.
792 * ipa-split.c (execute_split_functions): Likewise.
793 * symbol-summary.h: Rename get to get_create.
795 (get_create): Likewise.
796 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
798 2018-06-08 Martin Liska <mliska@suse.cz>
800 * symbol-summary.h (release): Move definition out of class
802 (symtab_removal): Likewise.
803 (symtab_duplication): Likewise.
805 2018-06-08 Martin Liska <mliska@suse.cz>
807 * symbol-summary.h (function_summary): Move constructor
808 implementation out of class declaration.
810 (symtab_insertion): Likewise.
811 (symtab_removal): Likewise.
812 (symtab_duplication): Likewise.
815 2018-06-08 Martin Liska <mliska@suse.cz>
817 * Makefile.in: Remove support for MPX (macros, related functions,
818 fields in cgraph_node, ...).
819 * builtin-types.def (BT_BND): Likewise.
820 (BT_FN_BND_CONST_PTR): Likewise.
821 (BT_FN_CONST_PTR_BND): Likewise.
822 (BT_FN_VOID_PTR_BND): Likewise.
823 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
824 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
825 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
826 (expand_builtin_mempcpy_with_bounds): Likewise.
827 (expand_builtin_memset_with_bounds): Likewise.
828 (expand_builtin_memset_args): Likewise.
829 (std_expand_builtin_va_start): Likewise.
830 (expand_builtin): Likewise.
831 (expand_builtin_with_bounds): Likewise.
832 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
833 (DEF_LIB_BUILTIN_CHKP): Likewise.
834 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
835 (DEF_CHKP_BUILTIN): Likewise.
836 (BUILT_IN_MEMCPY): Likewise.
837 (BUILT_IN_MEMMOVE): Likewise.
838 (BUILT_IN_MEMPCPY): Likewise.
839 (BUILT_IN_MEMSET): Likewise.
840 (BUILT_IN_STPCPY): Likewise.
841 (BUILT_IN_STRCAT): Likewise.
842 (BUILT_IN_STRCHR): Likewise.
843 (BUILT_IN_STRCPY): Likewise.
844 (BUILT_IN_STRLEN): Likewise.
845 (BUILT_IN_MEMCPY_CHK): Likewise.
846 (BUILT_IN_MEMMOVE_CHK): Likewise.
847 (BUILT_IN_MEMPCPY_CHK): Likewise.
848 (BUILT_IN_MEMSET_CHK): Likewise.
849 (BUILT_IN_STPCPY_CHK): Likewise.
850 (BUILT_IN_STRCAT_CHK): Likewise.
851 (BUILT_IN_STRCPY_CHK): Likewise.
852 * calls.c (store_bounds): Likewise.
853 (emit_call_1): Likewise.
854 (special_function_p): Likewise.
855 (maybe_warn_nonstring_arg): Likewise.
856 (initialize_argument_information): Likewise.
857 (finalize_must_preallocate): Likewise.
858 (compute_argument_addresses): Likewise.
859 (expand_call): Likewise.
860 * cfgexpand.c (expand_call_stmt): Likewise.
861 (expand_return): Likewise.
862 (expand_gimple_stmt_1): Likewise.
863 (pass_expand::execute): Likewise.
864 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
865 (cgraph_node::remove): Likewise.
866 (cgraph_node::dump): Likewise.
867 (cgraph_node::verify_node): Likewise.
868 * cgraph.h (chkp_function_instrumented_p): Likewise.
869 (symtab_node::get_alias_target): Likewise.
870 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
871 (cgraph_local_p): Likewise.
872 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
873 (cgraph_edge::rebuild_references): Likewise.
874 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
875 (walk_polymorphic_call_targets): Likewise.
876 (cgraph_node::expand_thunk): Likewise.
877 (symbol_table::output_weakrefs): Likewise.
878 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
879 (ix86_handle_option): Likewise.
880 * config/i386/constraints.md: Likewise.
881 * config/i386/i386-builtin-types.def (BND): Likewise.
885 * config/i386/i386-builtin.def (BDESC_END): Likewise.
886 (BDESC_FIRST): Likewise.
888 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
889 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
890 * config/i386/i386.c (enum reg_class): Likewise.
891 (ix86_target_string): Likewise.
892 (ix86_option_override_internal): Likewise.
893 (ix86_conditional_register_usage): Likewise.
894 (ix86_valid_target_attribute_inner_p): Likewise.
895 (ix86_set_indirect_branch_type): Likewise.
896 (ix86_set_current_function): Likewise.
897 (ix86_function_arg_regno_p): Likewise.
898 (init_cumulative_args): Likewise.
899 (ix86_function_arg_advance): Likewise.
900 (ix86_function_arg): Likewise.
901 (ix86_pass_by_reference): Likewise.
902 (ix86_function_value_regno_p): Likewise.
903 (ix86_function_value_1): Likewise.
904 (ix86_function_value_bounds): Likewise.
905 (ix86_return_in_memory): Likewise.
906 (ix86_setup_incoming_vararg_bounds): Likewise.
907 (ix86_va_start): Likewise.
908 (indirect_thunk_need_prefix): Likewise.
909 (print_reg): Likewise.
910 (ix86_print_operand): Likewise.
911 (ix86_expand_call): Likewise.
912 (ix86_output_function_return): Likewise.
913 (reg_encoded_number): Likewise.
914 (BDESC_VERIFYS): Likewise.
915 (ix86_init_mpx_builtins): Likewise.
916 (ix86_init_builtins): Likewise.
917 (ix86_emit_cmove): Likewise.
918 (ix86_emit_move_max): Likewise.
919 (ix86_expand_builtin): Likewise.
920 (ix86_builtin_mpx_function): Likewise.
921 (ix86_get_arg_address_for_bt): Likewise.
922 (ix86_load_bounds): Likewise.
923 (ix86_store_bounds): Likewise.
924 (ix86_load_returned_bounds): Likewise.
925 (ix86_store_returned_bounds): Likewise.
926 (ix86_class_likely_spilled_p): Likewise.
927 (ix86_hard_regno_mode_ok): Likewise.
928 (x86_order_regs_for_local_alloc): Likewise.
929 (ix86_mitigate_rop): Likewise.
930 (ix86_bnd_prefixed_insn_p): Likewise.
931 (ix86_mpx_bound_mode): Likewise.
932 (ix86_make_bounds_constant): Likewise.
933 (ix86_initialize_bounds): Likewise.
934 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
935 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
936 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
937 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
938 (TARGET_CHKP_BOUND_MODE): Likewise.
939 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
940 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
941 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
942 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
943 * config/i386/i386.h (TARGET_MPX): Likewise.
944 (TARGET_MPX_P): Likewise.
945 (VALID_BND_REG_MODE): Likewise.
946 (FIRST_BND_REG): Likewise.
947 (LAST_BND_REG): Likewise.
948 (enum reg_class): Likewise.
949 (BND_REG_P): Likewise.
950 (BND_REGNO_P): Likewise.
952 (ADJUST_INSN_LENGTH): Likewise.
953 * config/i386/i386.md: Likewise.
954 * config/i386/i386.opt: Likewise.
955 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
957 (LINK_MPX): Likewise.
958 (MPX_SPEC): Likewise.
959 (LIBMPX_SPEC): Likewise.
960 (LIBMPXWRAPPERS_SPEC): Likewise.
961 (CHKP_SPEC): Likewise.
962 * config/i386/predicates.md: Likewise.
963 * dbxout.c (dbxout_type): Likewise.
964 * doc/extend.texi: Likewise.
965 * doc/invoke.texi: Likewise.
966 * doc/md.texi: Likewise.
967 * doc/tm.texi: Likewise.
968 * doc/tm.texi.in: Likewise.
969 * dwarf2out.c (is_base_type): Likewise.
970 (gen_formal_types_die): Likewise.
971 (gen_subprogram_die): Likewise.
972 (gen_type_die_with_usage): Likewise.
973 (gen_decl_die): Likewise.
974 (dwarf2out_late_global_decl): Likewise.
975 * expr.c (expand_assignment): Likewise.
976 (emit_storent_insn): Likewise.
977 (store_expr_with_bounds): Likewise.
978 (store_expr): Likewise.
979 (expand_expr_real_1): Likewise.
980 * expr.h (store_expr_with_bounds): Likewise.
981 * function.c (use_register_for_decl): Likewise.
982 (struct bounds_parm_data): Likewise.
983 (assign_parms_augmented_arg_list): Likewise.
984 (assign_parm_find_entry_rtl): Likewise.
985 (assign_parm_is_stack_parm): Likewise.
986 (assign_parm_load_bounds): Likewise.
987 (assign_bounds): Likewise.
988 (assign_parms): Likewise.
989 (expand_function_start): Likewise.
990 * gcc.c (CHKP_SPEC): Likewise.
991 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
992 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
993 (wrestrict_dom_walker::check_call): Likewise.
994 * gimple.c (gimple_build_call_from_tree): Likewise.
995 * gimple.h (enum gf_mask): Likewise.
996 (gimple_call_with_bounds_p): Likewise.
997 (gimple_call_set_with_bounds): Likewise.
998 * gimplify.c (gimplify_init_constructor): Likewise.
999 * ipa-cp.c (initialize_node_lattices): Likewise.
1000 (propagate_constants_across_call): Likewise.
1001 (find_more_scalar_values_for_callers_subset): Likewise.
1002 * ipa-hsa.c (process_hsa_functions): Likewise.
1003 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1004 * ipa-icf.c (sem_function::merge): Likewise.
1005 * ipa-inline.c (early_inliner): Likewise.
1006 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1007 (warn_function_cold): Likewise.
1008 (propagate_pure_const): Likewise.
1009 * ipa-ref.h (enum GTY): Likewise.
1010 * ipa-split.c (find_retbnd): Likewise.
1011 (consider_split): Likewise.
1012 (split_function): Likewise.
1013 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1014 * ipa.c (walk_polymorphic_call_targets): Likewise.
1015 (symbol_table::remove_unreachable_nodes): Likewise.
1016 (process_references): Likewise.
1017 (cgraph_build_static_cdtor_1): Likewise.
1018 * lto-cgraph.c (lto_output_node): Likewise.
1019 (output_refs): Likewise.
1020 (compute_ltrans_boundary): Likewise.
1021 (input_overwrite_node): Likewise.
1022 (input_node): Likewise.
1023 (input_cgraph_1): Likewise.
1024 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1025 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1026 (class pass_chkp_instrumentation_passes): Likewise.
1027 (make_pass_chkp_instrumentation_passes): Likewise.
1028 * passes.def: Likewise.
1029 * rtl.h (struct GTY): Likewise.
1030 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1031 * stor-layout.c (layout_type): Likewise.
1032 * symtab.c: Likewise.
1033 * target.def: Likewise.
1034 * targhooks.c (default_chkp_bound_type): Likewise.
1035 (default_chkp_bound_mode): Likewise.
1036 (default_builtin_chkp_function): Likewise.
1037 (default_chkp_function_value_bounds): Likewise.
1038 (default_chkp_make_bounds_constant): Likewise.
1039 (default_chkp_initialize_bounds): Likewise.
1040 * targhooks.h (default_chkp_bound_type): Likewise.
1041 (default_chkp_bound_mode): Likewise.
1042 (default_builtin_chkp_function): Likewise.
1043 (default_chkp_function_value_bounds): Likewise.
1044 (default_chkp_make_bounds_constant): Likewise.
1045 (default_chkp_initialize_bounds): Likewise.
1046 * toplev.c (compile_file): Likewise.
1047 (process_options): Likewise.
1048 * tree-core.h (DEF_BUILTIN): Likewise.
1049 (DEF_BUILTIN_CHKP): Likewise.
1050 * tree-inline.c (declare_return_variable): Likewise.
1051 (remap_gimple_stmt): Likewise.
1052 (copy_bb): Likewise.
1053 (initialize_inlined_parameters): Likewise.
1054 (expand_call_inline): Likewise.
1055 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1056 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1057 (make_pass_ipa_chkp_produce_thunks): Likewise.
1058 (make_pass_chkp): Likewise.
1059 (make_pass_chkp_opt): Likewise.
1060 (make_pass_chkp_instrumentation_passes): Likewise.
1061 * tree-pretty-print.c (dump_generic_node): Likewise.
1062 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1063 * tree-ssa-dce.c (propagate_necessity): Likewise.
1064 (eliminate_unnecessary_stmts): Likewise.
1065 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1066 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1067 * tree-ssa-sccvn.h: Likewise.
1068 * tree-ssa-strlen.c (get_string_length): Likewise.
1069 (valid_builtin_call): Likewise.
1070 (adjust_last_stmt): Likewise.
1071 (handle_builtin_strchr): Likewise.
1072 (handle_builtin_strcpy): Likewise.
1073 (handle_builtin_stxncpy): Likewise.
1074 (handle_builtin_memcpy): Likewise.
1075 (handle_builtin_strcat): Likewise.
1076 (strlen_check_and_optimize_stmt): Likewise.
1077 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1078 * tree-streamer-in.c: Likewise.
1079 * tree-streamer.c (record_common_node): Likewise.
1080 * tree.c (tree_code_size): Likewise.
1081 (wide_int_to_tree_1): Likewise.
1082 (type_contains_placeholder_1): Likewise.
1083 (build_common_tree_nodes): Likewise.
1084 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1085 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1086 (POINTER_BOUNDS_P): Likewise.
1087 (BOUNDED_TYPE_P): Likewise.
1088 (BOUNDED_P): Likewise.
1089 (CALL_WITH_BOUNDS_P): Likewise.
1090 (pointer_bounds_type_node): Likewise.
1091 * value-prof.c (gimple_ic): Likewise.
1092 * var-tracking.c (vt_add_function_parameters): Likewise.
1093 * varasm.c (make_decl_rtl): Likewise.
1094 (assemble_start_function): Likewise.
1095 (output_constant): Likewise.
1096 (maybe_assemble_visibility): Likewise.
1097 * varpool.c (ctor_for_folding): Likewise.
1098 * chkp-builtins.def: Remove.
1099 * ipa-chkp.c: Remove.
1100 * ipa-chkp.h: Remove.
1101 * rtl-chkp.c: Remove.
1102 * rtl-chkp.h: Remove.
1103 * tree-chkp-opt.c: Remove.
1104 * tree-chkp.c: Remove.
1105 * tree-chkp.h: Remove.
1107 2018-06-07 Carl Love <cel@us.ibm.com>
1109 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1110 vextract_fp_from_shortl): Add BE support.
1112 2018-06-07 Paul Koning <ni1d@arrl.net>
1114 * compare-elim.c (try_merge_compare): Don't merge compare if
1115 address contains a side effect.
1116 (try_eliminate_compare): Likewise.
1118 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1120 * config.gcc: Support "tremont".
1121 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1122 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1124 * config/i386/i386.c (m_TREMONT): Define.
1125 (processor_target_table): Add "tremont".
1126 (PTA_TREMONT): Define.
1127 (ix86_lea_outperforms): Add TARGET_TREMONT.
1128 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1129 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1130 and M_INTEL_GOLDMONT_PLUS.
1131 (fold_builtin_cpu): Add "tremont".
1132 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1133 (ix86_option_override_internal): Add "tremont".
1134 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1135 (processor_type): Add PROCESSOR_TREMONT.
1136 * config/i386/x86-tune.def: Add m_TREMONT.
1137 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1139 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1141 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1142 symbol defined for msp430i* devices to be lower case.
1144 gcc/testsuite/gcc.target/msp430/
1145 * msp430i-device-symbol.c: New test.
1146 * msp430f-device-symbol.c: New test.
1147 * msp430.h: New test header file.
1149 2018-06-07 Richard Biener <rguenther@suse.de>
1151 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1152 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1153 Properly wrap signed arithmetic if overflow wraps.
1155 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1157 PR tree-optimization/69615
1158 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1159 of a cast from a same precision integral SSA_NAME in a bb dominated
1160 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1161 cast to utype if rhs2 has already a compatible type.
1163 2018-06-07 Richard Biener <rguenther@suse.de>
1165 PR tree-optimization/85935
1166 * graphite-scop-detection.c (find_params_in_bb): Analyze
1167 condition operands with respect to the correct loop. Assert
1168 the analysis doesn't fail.
1170 2018-06-04 Carl Love <cel@us.ibm.com>
1172 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1173 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1174 as it is slightly cheaper.
1175 (first_match_or_eos_index_<mode>):
1176 Calculate index using natural element order.
1177 (first_match_index_<mode>):
1178 Calculate index using natural element order.
1179 (first_match_or_eos_index_<mode>):
1180 Calculate index using natural order.
1181 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1182 for BE and LE modes.
1183 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1184 P9V_BUILTIN_VCLZLSBB_V16QI.
1185 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1188 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1190 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1191 indentation and line wrap for many prototypes. Add missing
1192 @smallexample directives around block of prototypes for vec_xl and
1195 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1197 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1198 track if we pass or return IEEE 128-bit floating point.
1199 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1200 C++ mangling that is compatible with GCC 8.1.
1201 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1202 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1203 floating point types.
1204 (rs6000_function_arg_advance_1): Likewise.
1205 (rs6000_mangle_type): Optionally generate mangled names that match
1206 what GCC 8.1 generated for IEEE 128-bit floating point types.
1207 (rs6000_globalize_decl_name): If we have an external function that
1208 passes or returns IEEE 128-bit types, generate a weak reference
1209 from the mangled name used in GCC 8.1 to the current mangled
1211 (rs6000_init_builtins): Make __ibm128 use the long double type if
1212 long double is IBM extended double. Make __float128 use the long
1213 double type if long double is IEEE 128-bit.
1216 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1217 macro for __ibm128 built-in functions.
1218 (PACK_IF): Add __ibm128 pack/unpack functions.
1219 (UNPACK_IF): Likewise.
1220 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1221 enable long double built-in functions if long double is IEEE
1222 128-bit floating point.
1223 (rs6000_invalid_builtin): Update long double built-in function
1225 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1226 functions, adjust the built-in function to use the long double
1227 built-in function if __ibm128 and long double are the same type.
1228 * doc/extend.texi (PowerPC builtins): Update documention for
1229 __builtin_{,un}pack_longdouble. Add documentation for
1230 __builtin_{,un}pack_ibm128.
1232 2018-06-06 Jim Wilson <jimw@sifive.com>
1234 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1235 (struct machine_function): New field interrupt_mode.
1236 (riscv_handle_type_attribute): New function. Add forward declaration.
1237 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1238 (riscv_expand_epilogue): Check interrupt_mode field.
1239 (riscv_set_current_function): Check interrupt attribute args and
1240 set interrupt_mode field.
1241 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1242 (riscv_sret, riscv_uret): New.
1243 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1244 new arguments to interrupt attribute.
1246 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1249 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1250 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1252 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1254 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1255 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1256 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1257 several redundant entries.
1259 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1261 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1262 type from "rtx" to "rtx_insn *".
1263 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1264 for local "call_insn", removing cast.
1265 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1267 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1269 PR tree-optimization/86066
1270 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1271 for BIT_INSERT_EXPR stores.
1273 2018-06-06 Richard Biener <rguenther@suse.de>
1275 PR tree-optimization/86062
1276 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1277 component refs ontop
1278 of to be offsetted base.
1280 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1282 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1283 to be static and remove check on interrupt attribute name.
1285 gcc/testsuite/gcc.target/msp430/
1286 * function-attributes-4.c: New test.
1287 * static-interrupts.c: New test.
1289 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1291 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1292 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1294 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1297 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1299 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1300 Remove second argument, change how error is called.
1301 (aarch64_layout_arg): Remove second argument from
1302 aarch64_err_no_fpadvsimd call.
1303 (aarch64_init_cumulative_args): Ditto.
1304 (aarch64_gimplify_va_arg_expr): Ditto.
1305 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1307 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1309 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1310 (*simple_return_indirect_internal<mode>): Rename from
1311 simple_return_indirect_internal. Use W mode iterator.
1312 (rstorssp): New expander.
1313 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1314 (clrssbsy): New expander.
1315 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1317 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1319 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1321 (cmse_check_pointed_object): Likewise.
1323 2018-06-05 Martin Liska <mliska@suse.cz>
1325 PR gcov-profile/47618
1326 * doc/invoke.texi: Document how -fprofile-dir format
1329 2018-06-05 Richard Biener <rguenther@suse.de>
1331 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1332 removal pretend DOM info isn't available so we do not update
1333 it and only remove edges, not dominated blocks. Actually free
1334 DOM info in case we removed something. Remove unreachable blocks.
1335 (mfb_keep_latches): Work with either DOM info or marked backedges.
1336 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1337 first. Mark backedges if DOM info isn't available.
1338 (Re-)compute DOM info after cleanup_control_flow_pre.
1340 2018-06-05 Richard Biener <rguenther@suse.de>
1342 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1343 (locus_discrim_hasher::hash): Adjust.
1344 (locus_discrim_hasher::equal): Likewise.
1345 (next_discriminator_for_locus): Work on line directly.
1346 (same_line_p): Pass in expanded locus1 as well.
1347 (assign_discriminators): Avoid redundant location expansions.
1349 2018-06-05 Richard Biener <rguenther@suse.de>
1351 PR tree-optimization/86046
1352 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1353 if required after clearing TREE_ADDRESSABLE.
1355 2018-06-05 Richard Biener <rguenther@suse.de>
1357 PR tree-optimization/86047
1358 * tree-ssa-loop.c (for_each_index): Glob handling of all
1359 decls and constants and really handle all of them.
1361 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1364 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1365 qualifier_void_pointer and qualifier_const_void_pointer.
1366 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1367 (arm_init_builtins): Handle the above.
1368 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1369 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1370 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1373 2018-06-05 Martin Liska <mliska@suse.cz>
1375 * auto-profile.c (read_autofdo_file): Do not use
1376 gcov_ctr_summary struct.
1377 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1378 * coverage.c (struct counts_entry): Likewise.
1379 (read_counts_file): Read just single summary entry.
1380 (get_coverage_counts): Use gcov_summary struct.
1381 * coverage.h (get_coverage_counts): Likewise.
1382 * gcov-dump.c (dump_working_sets): Likewise.
1383 (tag_summary): Dump just single summary.
1384 * gcov-io.c (gcov_write_summary): Write just histogram
1386 (gcov_read_summary): Read just single summary.
1387 (compute_working_sets): Use gcov_summary struct.
1388 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1389 of GCOV_COUNTERS_SUMMABLE.
1390 (GCOV_COUNTERS_SUMMABLE): Remove.
1391 (GCOV_FIRST_VALUE_COUNTER): Replace with
1392 GCOV_COUNTER_V_INTERVAL.
1393 (struct gcov_ctr_summary): Remove.
1394 (struct gcov_summary): Directly use fields of former
1396 (compute_working_sets): Use gcov_summary struct.
1397 * gcov.c (read_count_file): Do not use ctrs fields.
1398 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1400 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1402 * profile.c: Likewise.
1403 * profile.h: Likewise.
1405 2018-06-05 Martin Liska <mliska@suse.cz>
1407 PR gcov-profile/84846
1408 * gcov.c (output_lines): Print working directory only
1409 in intermediate format.
1411 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1413 * config/s390/s390-builtin-types.def: Add void function type.
1414 * config/s390/s390-builtins.def: Use the function type for the
1417 2018-06-04 Jim Wilson <jimw@sifive.com>
1419 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1421 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1422 and maybe_eh_return. Change regno to unsigned int. Use new args to
1423 handle EH_RETURN_DATA_REGNO registers properly.
1424 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1425 (riscv_expand_epilogue): Update comment. Change argument name and
1426 type. Update code to use new name and type. Pass new args to
1427 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1429 * config/riscv/riscv.md (NORMAL_RETURN): New.
1430 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1431 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1432 (eh_return): Call gen_eh_return_internal and emit barrier.
1433 (eh_return_internal): Call riscv_expand_epilogue.
1435 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1437 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1438 bit_insertion field and declare can_be_merged_into method.
1439 (merged_store_group::can_be_merged_into): New method.
1440 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1441 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1442 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1444 2018-06-04 Richard Biener <rguenther@suse.de>
1446 PR tree-optimization/85955
1447 * builtins.c (fold_builtin_sincos): Convert pointers to
1448 destination to appropriate type before dereferencing.
1450 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1452 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1454 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1456 * expr.c (expand_expr_real_1): Force the operand into memory if
1457 its TYPE_MODE is BLKmode and if there is no integer mode for
1458 the number of bits being extracted.
1460 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1464 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1465 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1467 2018-06-04 Richard Biener <rguenther@suse.de>
1469 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1470 (cleanup_tree_cfg_noloop): ... single caller. Do
1471 start_recording_case_labels later.
1473 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1475 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1476 to _IMMINTRIN_H_INCLUDED.
1477 * config/i386/pconfigintrin.h: Ditto.
1478 * config/i386/waitpkgintrin.h: Ditto.
1479 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1480 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1481 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1482 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1483 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1484 waitpkgintrin.h and cldemoteintrin.h.
1486 2018-06-04 Richard Biener <rguenther@suse.de>
1488 PR tree-optimization/86038
1489 * tracer.c (find_best_successor): Check probability for
1490 being initialized, bail out if not.
1492 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1495 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1496 of bits to ignore when comparing architectures.
1498 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1500 PR tree-optimization/69615
1501 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1502 maximum or minimum of the type, try to merge it also as if
1503 range1 is + [-, x - 1] or + [x + 1, -].
1506 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1508 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1510 PR tree-optimization/86034
1511 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1512 the unsigned bitfield type in a bit insertion sequence if it does not
1513 have a larger precision than the bitfield size.
1514 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1516 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1518 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1520 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1522 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1523 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1524 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1525 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1527 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1529 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1530 Disable -fdelete-null-pointer-checks for ELF toolchain.
1532 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1533 Kito Cheng <kito.cheng@gmail.com>
1535 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1536 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1537 (nds32*-*-*): Add glibc and uclibc conditions.
1538 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1539 (TARGET_EXCEPT_UNWIND_INFO): Define.
1540 * config/nds32/elf.h: New file.
1541 * config/nds32/linux.h: New file.
1542 * config/nds32/nds32-elf.opt: New file.
1543 * config/nds32/nds32-linux.opt: New file.
1544 * config/nds32/nds32-fp-as-gp.c
1545 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1546 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1548 (nds32_asm_file_end): Ditto.
1549 (nds32_print_operand): Ditto.
1550 (nds32_insert_attributes): Ditto.
1551 (nds32_init_libfuncs): New function.
1552 (TARGET_HAVE_TLS): Define.
1553 (TARGET_INIT_LIBFUNCS): Define.
1554 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1556 (TARGET_ELF): Apply different mcmodel setting.
1557 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1558 been migrated into elf.h and linux.h files.
1559 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1560 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1561 (mcmodel): The content has been migrated into nds32-elf.opt and
1562 nds32-linux.opt files.
1563 * config/nds32/t-elf: New file.
1564 * config/nds32/t-linux: New file.
1566 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1567 Shiva Chen <shiva0217@gmail.com>
1569 * config/nds32/constants.md (unspec_volatile_element): Add
1570 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1571 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1573 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1574 (make_pass_nds32_fp_as_gp): Declare.
1575 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1577 (nds32_asm_function_end_prologue): Remove unused asm output.
1578 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1579 (nds32_asm_file_start): Output necessary fp_as_gp information.
1580 (nds32_option_override): Adjust register usage.
1581 (nds32_expand_prologue): Consider fp_as_gp situation.
1582 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1583 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1586 (simple_return): Ditto.
1587 (omit_fp_begin): Output special directive for fp_as_gp.
1588 (omit_fp_end): Output special directive for fp_as_gp.
1589 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1590 mforbid-fp-as-gp): New options.
1592 2018-06-01 Mark Wielaard <mark@klomp.org>
1594 * dwarf2out.c (dwarf2out_finish): Remove generation of
1595 DW_AT_loclists_base.
1597 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1599 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1600 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1601 (struct merged_store_group): Add bit_insertion field.
1602 (dump_char_array): Use standard hexadecimal format.
1603 (merged_store_group::merged_store_group): Set bit_insertion to false.
1604 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1605 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1606 also print the mask in the dump file.
1607 (pass_store_merging::gate): Minor tweak.
1608 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1609 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1610 stores with INTEGER_CST stores.
1611 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1612 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1613 and use it throughout. Generate bit insertion sequences if need be.
1614 (pass_store_merging::process_store): Remove redundant condition.
1615 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1617 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1619 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1620 the 128-bit floating point types. Fix function comment.
1622 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1624 * config/aarch64/aarch64-simd.md
1625 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1627 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1630 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1632 PR tree-optimization/85989
1633 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1635 (backprop::intersect_uses): Check it when deciding whether this
1636 is a backedge reference.
1637 (backprop::process_block): Add each phi to m_visited_phis
1638 after visiting it, then clear it at the end.
1640 2018-06-01 Richard Biener <rguenther@suse.de>
1642 * tree-vectorizer.h (vect_dr_stmt): New function.
1643 (vect_get_load_cost): Adjust.
1644 (vect_get_store_cost): Likewise.
1645 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1646 Use vect_dr_stmt instead of DR_SMTT.
1647 (vect_record_base_alignments): Likewise.
1648 (vect_calculate_target_alignment): Likewise.
1649 (vect_compute_data_ref_alignment): Likewise and make static.
1650 (vect_update_misalignment_for_peel): Likewise.
1651 (vect_verify_datarefs_alignment): Likewise.
1652 (vector_alignment_reachable_p): Likewise.
1653 (vect_get_data_access_cost): Likewise. Pass down
1654 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1655 (vect_get_peeling_costs_all_drs): Likewise.
1656 (vect_peeling_hash_get_lowest_cost): Likewise.
1657 (vect_enhance_data_refs_alignment): Likewise.
1658 (vect_find_same_alignment_drs): Likewise.
1659 (vect_analyze_data_refs_alignment): Likewise.
1660 (vect_analyze_group_access_1): Likewise.
1661 (vect_analyze_group_access): Likewise.
1662 (vect_analyze_data_ref_access): Likewise.
1663 (vect_analyze_data_ref_accesses): Likewise.
1664 (vect_vfa_segment_size): Likewise.
1665 (vect_small_gap_p): Likewise.
1666 (vectorizable_with_step_bound_p): Likewise.
1667 (vect_prune_runtime_alias_test_list): Likewise.
1668 (vect_analyze_data_refs): Likewise.
1669 (vect_supportable_dr_alignment): Likewise.
1670 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1671 (vect_gen_prolog_loop_niters): Likewise.
1672 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1673 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1675 (vect_recog_mask_conversion_pattern): Likewise.
1676 (vect_try_gather_scatter_pattern): Likewise.
1677 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1678 to vect_get_store_cost.
1679 (vect_get_store_cost): Get stmt_info instead of DR.
1680 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1681 (vect_get_load_cost): Get stmt_info instead of DR.
1683 2018-06-01 Richard Biener <rguenther@suse.de>
1686 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1687 vars wrapped in MEM_REFs.
1689 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1691 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1692 Fix subreg tests so that we only return a choice between
1693 GENERAL_REGS and FP_REGS if the original classes included both.
1695 2018-06-01 Richard Biener <rguenther@suse.de>
1698 * tree-ssa-structalias.c (get_function_part_constraint):
1699 Handle NULL fi->decl.
1700 (find_func_aliases_for_call): Properly handle indirect
1701 fi from direct call.
1702 (find_func_clobbers): Likewise.
1703 (ipa_pta_execute): Likewise.
1704 (create_variable_info_for): For functions that are ifunc_resolver
1705 resolve to a varinfo that contains the result of the resolver call.
1706 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1709 2018-05-31 Michael Collison <michael.collison@arm.com>
1711 * config/aarch64/aarch64.md:
1712 (*fix_to_zero_extenddfdi2): New pattern.
1713 * gcc.target/aarch64/fix_extend1.c: New testcase.
1715 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1719 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1720 and BUILT_IN_STRNCMP_EQ.
1721 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1722 BUILT_IN_STRNCMP_EQ.
1723 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1724 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1725 (gimple_fold_builtin): Likewise.
1726 * tree-ssa-strlen.c (compute_string_length): New function.
1727 (determine_min_obsize): New function.
1728 (handle_builtin_string_cmp): New function to handle calls to
1729 string compare functions.
1730 (strlen_optimize_stmt): Add handling to builtin string compare
1732 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1733 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1734 * tree.c (build_common_builtin_nodes): Add new defines of
1735 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1737 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1740 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1741 functions with naked attribute.
1743 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1745 * config/i386/sse.md (avx_vec_concat<mode>):
1746 Substitute concat_tg_mode mode attribute with xtg_mode.
1747 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1748 (concat_tg_mode): Remove mode attribute.
1750 2018-05-31 Martin Sebor <msebor@redhat.com>
1753 * calls.c (alloc_max_size): Correct a logic error/typo.
1754 Treat excessive arguments as infinite. Warn for invalid arguments.
1755 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1757 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1760 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1761 and movx for Haswell.
1763 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1764 Cesar Philippidis <cesar@codesourcery.com>
1767 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1768 when emitting error on private/firstprivate reductions.
1769 * omp-low.c (lower_omp_target): Avoid reference-type processing
1770 on pointers for firstprivate clause.
1772 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1774 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1777 * config/aarch64/aarch64-simd.md
1778 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1779 (aarch64_ld1_x3_<mode>): Likewise
1780 (aarch64_st1x2<VALLDIF:mode>): Likewise
1781 (aarch64_st1_x2_<mode>): Likewise
1782 (aarch64_st1x3<VALLDIF:mode>): Likewise
1783 (aarch64_st1_x3_<mode>): Likewise
1784 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1785 (vld1_s8_x3): Likewise.
1786 (vld1_u16_x3): Likewise.
1787 (vld1_s16_x3): Likewise.
1788 (vld1_u32_x3): Likewise.
1789 (vld1_s32_x3): Likewise.
1790 (vld1_u64_x3): Likewise.
1791 (vld1_s64_x3): Likewise.
1792 (vld1_f16_x3): Likewise.
1793 (vld1_f32_x3): Likewise.
1794 (vld1_f64_x3): Likewise.
1795 (vld1_p8_x3): Likewise.
1796 (vld1_p16_x3): Likewise.
1797 (vld1_p64_x3): Likewise.
1798 (vld1q_u8_x3): Likewise.
1799 (vld1q_s8_x3): Likewise.
1800 (vld1q_u16_x3): Likewise.
1801 (vld1q_s16_x3): Likewise.
1802 (vld1q_u32_x3): Likewise.
1803 (vld1q_s32_x3): Likewise.
1804 (vld1q_u64_x3): Likewise.
1805 (vld1q_s64_x3): Likewise.
1806 (vld1q_f16_x3): Likewise.
1807 (vld1q_f32_x3): Likewise.
1808 (vld1q_f64_x3): Likewise.
1809 (vld1q_p8_x3): Likewise.
1810 (vld1q_p16_x3): Likewise.
1811 (vld1q_p64_x3): Likewise.
1812 (vst1_s64_x2): Likewise.
1813 (vst1_u64_x2): Likewise.
1814 (vst1_f64_x2): Likewise.
1815 (vst1_s8_x2): Likewise.
1816 (vst1_p8_x2): Likewise.
1817 (vst1_s16_x2): Likewise.
1818 (vst1_p16_x2): Likewise.
1819 (vst1_s32_x2): Likewise.
1820 (vst1_u8_x2): Likewise.
1821 (vst1_u16_x2): Likewise.
1822 (vst1_u32_x2): Likewise.
1823 (vst1_f16_x2): Likewise.
1824 (vst1_f32_x2): Likewise.
1825 (vst1_p64_x2): Likewise.
1826 (vst1q_s8_x2): Likewise.
1827 (vst1q_p8_x2): Likewise.
1828 (vst1q_s16_x2): Likewise.
1829 (vst1q_p16_x2): Likewise.
1830 (vst1q_s32_x2): Likewise.
1831 (vst1q_s64_x2): Likewise.
1832 (vst1q_u8_x2): Likewise.
1833 (vst1q_u16_x2): Likewise.
1834 (vst1q_u32_x2): Likewise.
1835 (vst1q_u64_x2): Likewise.
1836 (vst1q_f16_x2): Likewise.
1837 (vst1q_f32_x2): Likewise.
1838 (vst1q_f64_x2): Likewise.
1839 (vst1q_p64_x2): Likewise.
1840 (vst1_s64_x3): Likewise.
1841 (vst1_u64_x3): Likewise.
1842 (vst1_f64_x3): Likewise.
1843 (vst1_s8_x3): Likewise.
1844 (vst1_p8_x3): Likewise.
1845 (vst1_s16_x3): Likewise.
1846 (vst1_p16_x3): Likewise.
1847 (vst1_s32_x3): Likewise.
1848 (vst1_u8_x3): Likewise.
1849 (vst1_u16_x3): Likewise.
1850 (vst1_u32_x3): Likewise.
1851 (vst1_f16_x3): Likewise.
1852 (vst1_f32_x3): Likewise.
1853 (vst1_p64_x3): Likewise.
1854 (vst1q_s8_x3): Likewise.
1855 (vst1q_p8_x3): Likewise.
1856 (vst1q_s16_x3): Likewise.
1857 (vst1q_p16_x3): Likewise.
1858 (vst1q_s32_x3): Likewise.
1859 (vst1q_s64_x3): Likewise.
1860 (vst1q_u8_x3): Likewise.
1861 (vst1q_u16_x3): Likewise.
1862 (vst1q_u32_x3): Likewise.
1863 (vst1q_u64_x3): Likewise.
1864 (vst1q_f16_x3): Likewise.
1865 (vst1q_f32_x3): Likewise.
1866 (vst1q_f64_x3): Likewise.
1867 (vst1q_p64_x3): Likewise.
1869 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1871 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1872 user_label_prefix to name.
1874 * tree-core.h: Update comment about the format of NAME string
1875 passed to handler in attribute_spec.
1877 * config/msp430/msp430.md: Remove erroneous subreg expression from
1878 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1879 zero_extend{q,h}isi2.
1881 2018-05-30 Borislav Petkov <bp@suse.de>
1883 * doc/extend.texi: Document some architecture specific
1884 constraints and sort entries.
1886 2018-05-30 Martin Sebor <msebor@redhat.com>
1889 * builtins.c (expand_builtin_stpcpy_1): New function.
1890 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1891 only if the former succeeds.
1893 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1895 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1898 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1900 * doc/invoke.texi (-flinker-output): Document
1902 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1904 * passes.c (ipa_write_summaries): Only modify statements if body
1906 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1907 incrementally linking.
1908 (ipa_passes): Likewise.
1909 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1910 pass down resolution info.
1911 * common.opt (flag_incremental_link): Update info.
1912 * gcc.c (plugin specs): Turn flinker-output=* to
1913 -plugin-opt=-linker-output-known
1914 * toplev.c (compile_file): Also cut compilation when doing incremental
1916 * flag-types. (enum lto_partition_model): Add
1917 LTO_LINKER_OUTPUT_NOLTOREL.
1918 (invoke.texi): Add -flinker-output docs.
1919 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1920 link same way as WPA; do not stream in dead initializers.
1922 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1923 darf2out_register_external_decl): Support incremental link.
1925 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1927 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
1929 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1931 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
1932 it down to simple_object_copy_lto_debug_sections.
1933 (run_gcc): Determine incremental LTO link time and configure
1934 lto1 into non-wpa mode, disable renaming of debug sections.
1936 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
1938 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1939 descriptions of various incorrectly documented functions.
1941 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
1944 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1945 address check not strict.
1947 2018-05-30 Richard Biener <rguenther@suse.de>
1949 PR tree-optimization/85964
1950 * tracer.c (better_p): Drop initialized count check, we only
1951 call the function with initialized counts now.
1952 (find_best_successor): Do find a best edge if one
1953 has uninitialized count.
1954 (find_best_predecessor): Likewise. Do BB frequency check only
1955 if count is initialized.
1957 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
1959 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
1960 (aarch64_ldrstr_offset_compare): New.
1961 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
1962 load/store orderings.
1963 (aarch64_gen_adjusted_ldpstp): Likewise.
1965 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
1967 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1968 Check for subset of GENERAL_REGS and FP_REGS.
1969 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
1972 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
1974 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
1975 and wi::to_poly_offset. Add the current offset and then check
1976 whether the sum fits, rather than using an unchecked addition of
1977 a checked term. Check for a shwi rather than a uhwi.
1978 * expr.c (get_bit_range): Use tree_to_poly_uint64.
1979 (store_constructor): Use poly_int_tree_p.
1980 (expand_expr_real_1): Likewise.
1981 * function.c (assign_temp): Likewise.
1982 * fold-const.c (const_binop): Use poly_int_tree_p and
1984 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
1986 * ipa-icf-gimple.c (func_checker::compare_operand): Use
1987 to_poly_offset for MEM offsets.
1988 * ipa-icf.c (sem_variable::equals): Likewise.
1989 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
1990 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
1991 wi::to_poly_offset for BIT_FIELD_REF offsets.
1992 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
1994 * var-tracking.c (emit_note_insn_var_location): Use
1995 tree_to_poly_uint64.
1997 2018-05-29 Jim Wilson <jimw@sifive.com>
1999 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2001 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2004 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2005 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2007 (sse4_1_round<mode>2): Use nonimmediate_operand
2008 for operand 1 predicate.
2010 2018-05-29 Martin Sebor <msebor@redhat.com>
2011 Richard Biener <rguenther@suse.de>
2014 * calls.c (get_size_range): Call determine_value_range instead
2015 of get_value_range..
2016 * tree-vrp.h (determine_value_range): Declared new function.
2017 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2019 2018-05-29 Richard Biener <rguenther@suse.de>
2021 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2022 sure to use non-pattern stmts for get_earlier_stmt arguments.
2023 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2024 called on pattern stmts.
2025 (get_later_stmt): Likewise.
2027 2018-05-29 Martin Liska <mliska@suse.cz>
2029 PR gcov-profile/85759
2030 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2033 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2035 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2037 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2041 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2042 VEC_PACK_FLOAT_EXPR): New tree codes.
2043 * tree-pretty-print.c (op_code_prio): Handle
2044 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2045 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2046 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2047 * tree-inline.c (estimate_operator_cost): Likewise.
2048 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2049 * fold-const.c (const_binop): Likewise.
2050 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2051 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2052 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2053 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2054 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2055 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2056 * expr.c (expand_expr_real_2): Likewise.
2057 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2058 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2059 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2061 * optabs.c (expand_widen_pattern_expr): For
2062 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2063 sign from result type rather than operand's type.
2064 (expand_binop_directly): For vec_packu_float_optab and
2065 vec_packs_float_optab allow result type to be different from operand's
2067 * optabs-tree.c (optab_for_tree_code): Handle
2068 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2069 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2070 * tree-vect-generic.c (expand_vector_operations_1): Handle
2071 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2072 VEC_PACK_FLOAT_EXPR.
2073 * tree-vect-stmts.c (supportable_widening_operation): Handle
2075 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2076 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2077 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2078 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2079 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2081 (vec_pack<floatprefix>_float_<mode>): New expander.
2082 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2084 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2085 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2086 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2087 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2088 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2090 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2091 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2092 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2093 VEC_PACK_FLOAT_EXPR): Document.
2095 2018-05-29 Richard Biener <rguenther@suse.de>
2097 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2099 (stmt_vec_info_vec): Make pointer.
2100 (init_stmt_vec_info_vec): Remove.
2101 (free_stmt_vec_info_vec): Likewise.
2102 (set_stmt_vec_info_vec): New function.
2103 (free_stmt_vec_infos): Likewise.
2104 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2105 (set_vinfo_for_stmt): Likewise.
2106 (get_earlier_stmt): Likewise.
2107 (get_later_stmt): Likewise.
2108 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2109 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2110 (vec_info::~vec_info): Free stmt_vec_infos.
2111 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2112 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2113 (pass_slp_vectorize::execute): Likewise.
2114 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2115 (free_stmt_vec_info_vec): Likewise.
2116 (set_stmt_vec_info_vec): New function.
2117 (free_stmt_vec_infos): Likewise.
2118 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2119 the global stmt_vec_info_vec.
2120 * tree-parloops.c (gather_scalar_reductions): Use
2121 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2124 2018-05-29 Richard Biener <rguenther@suse.de>
2126 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2128 2018-05-29 Martin Liska <mliska@suse.cz>
2129 David Malcolm <dmalcolm@redhat.com>
2131 * vec.c (test_reverse): New.
2132 (vec_c_tests): Add new test.
2133 * vec.h (vl_ptr>::reverse): New function.
2135 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2137 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2139 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2142 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2144 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2146 2018-05-28 Richard Biener <rguenther@suse.de>
2148 PR tree-optimization/85933
2149 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2150 look at stmts marked as vectorizable.
2152 2018-05-28 Richard Biener <rguenther@suse.de>
2154 PR tree-optimization/85934
2155 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2156 vector boolean check before scalar optimization.
2158 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2160 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2163 2018-05-28 Mark Wielaard <mark@klomp.org>
2165 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2166 if it is an expression containing a minus sign.
2168 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2170 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2172 2018-05-27 Paul Koning <ni1d@arrl.net>
2174 * config/pdp11/pdp11.md (truncsihi2): Remove.
2176 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2177 Chung-Ju Wu <jasonwucj@gmail.com>
2179 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2181 (unaligned_store_dw): Ditto.
2182 * config/nds32/nds32-memory-manipulation.c
2183 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2184 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2185 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2186 (emit_setmem_word_loop): Rename to ...
2187 (emit_setmem_doubleword_loop): ... this.
2188 (nds32_gen_dup_4_byte_to_word_value): New function.
2189 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2190 (nds32_expand_setmem_loop): Refine implementation.
2191 (nds32_expand_setmem_loop_v3m): Ditto.
2192 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2195 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2197 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2199 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2201 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2202 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2203 machine->attr_no_prologue_p.
2204 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2205 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2206 (nds32_expand_epilogue): Consider attr_naked_p.
2207 (nds32_expand_epilogue_v3pop): Likewise.
2208 (nds32_can_use_return_insn): Likewise.
2209 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2210 attr_no_prologue_p fields.
2211 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2213 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2216 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2218 * config/i386/sse.md
2219 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2221 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2223 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2225 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2227 (*<floatsuffix>floatv2div2sf2): Rename to ...
2228 (*float<floatunssuffix>v2div2sf2): ... this.
2229 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2230 (float<floatunssuffix>v2div2sf2_mask): ... this.
2231 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2232 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2233 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2235 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2237 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2239 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2241 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2243 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2245 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2246 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2247 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2248 gen_ufix_truncv8dfv8si2.
2249 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2250 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2251 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2252 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2253 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2254 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2255 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2256 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2258 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2262 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2264 2018-05-25 Jim Wilson <jimw@sifive.com>
2266 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2267 * config/riscv/riscv.c (struct machine_function): Add
2268 interrupt_handler_p and attribute_checked_p fields.
2269 (riscv_attribute_table): Add interrupt.
2270 (riscv_interrupt_type_p): New.
2271 (riscv_save_reg_p): Save extra regs for interrupt handler.
2272 (riscv_use_save_libcall): Return false for interrupt handler.
2273 (riscv_first_stack_step): Add forward declaration.
2274 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2275 for interrupt handler with large frame. Use it for saved reg list.
2276 (riscv_expand_prologue): Move flag_stack_usage_info support to
2277 eliminate duplication.
2278 (riscv_expand_epilogue): Generate mret for interrupt handler.
2279 (riscv_epilogue_uses): New.
2280 (riscv_can_use_return_insn): Return false for interrupt handler.
2281 (riscv_function_ok_for_sibcall): Likewise.
2282 (riscv_set_current_function): Add interrupt handler support.
2283 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2284 * config/riscv/riscv.md (UNSPECV_MRET): New.
2286 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2288 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2290 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2292 PR tree-optimization/85712
2293 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2294 this candidate has already been replaced in-situ by a copy.
2296 2018-05-25 Jason Merrill <jason@redhat.com>
2298 PR c++/80485 - inline function non-zero address.
2299 * symtab.c (nonzero_address): Check DECL_COMDAT.
2301 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2304 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2305 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2306 extension of SImode operation. Use const123_operand predicate.
2307 (*saddsi_1): Remove.
2308 (*saddl_se_1): Ditto.
2310 (*ssubl_se_1): Ditto.
2311 * config/alpha/predicates.md (const123_operand): New predicate.
2312 * config/alpha/constraints.md (P): Use IN_RANGE.
2314 2018-05-25 Richard Biener <rguenther@suse.de>
2316 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2318 (ref_maybe_used_by_stmt_p): Likewise.
2319 (stmt_may_clobber_ref_p): Likewise.
2320 (stmt_may_clobber_ref_p_1): Likewise.
2321 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2323 (ref_maybe_used_by_stmt_p): Likewise.
2324 (stmt_may_clobber_ref_p): Likewise.
2325 (stmt_may_clobber_ref_p_1): Likewise.
2326 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2327 the alias oracle to disambiguate DRs with stmts DR analysis
2329 (vect_analyze_data_refs): Do not give up on not analyzable
2330 DRs for BB vectorization. Remove code truncating the dataref
2333 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2336 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2337 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2338 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2340 2018-05-25 Richard Biener <rguenther@suse.de>
2342 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2343 function, combining stmt data ref gathering and fatal analysis
2345 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2346 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2347 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2348 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2349 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2351 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2353 PR tree-optimization/85720
2354 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2355 SCC if all partitions are builtins.
2356 (version_loop_by_alias_check): New parameter. Generate cancelable
2357 runtime alias check if all partitions are builtins.
2358 (distribute_loop): Update call to above function.
2360 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2362 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2363 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2364 (parm_default_def_partition_arg): Ditto.
2365 (set_parm_default_def_partition): Ditto.
2366 (get_parm_default_def_partitions): Ditto and make it static.
2367 (get_undefined_value_partitions): Ditto and make it static.
2368 (remove_ssa_form): Refactor call to init_var_map here.
2369 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2370 computation for loop region.
2371 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2372 (register_default_def): Delete.
2373 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2374 (parm_default_def_partition_arg): Ditto.
2375 (set_parm_default_def_partition): Ditto.
2376 (get_parm_default_def_partitions): Ditto and make it static.
2377 (get_undefined_value_partitions): Ditto and make it static.
2378 (coalesce_with_default, coalesce_with_default): Update comment.
2379 (create_coalesce_list_for_region): New func factored out from
2380 create_outofssa_var_map.
2381 (populate_coalesce_list_for_outofssa): New func factored out from
2382 create_outofssa_var_map and coalesce_ssa_name.
2383 (create_outofssa_var_map): Delete.
2384 (coalesce_ssa_name): Refactor to support live range computation.
2385 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2386 (get_parm_default_def_partitions): Delete.
2387 (get_undefined_value_partitions): Ditto.
2388 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2389 computation for loop region.
2390 (new_tree_live_info, loe_visit_block): Ditto.
2391 (live_worklist, set_var_live_on_entry): Ditto.
2392 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2393 * tree-ssa-live.h (struct _var_map): New fields.
2394 (init_var_map): Change decl.
2395 (region_contains_p): New.
2397 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2399 * tree-ssa-live.h (live_merge_and_clear): Delete.
2401 2018-05-25 Richard Biener <rguenther@suse.de>
2404 * tree-dump.c (dequeue_and_dump): Remove access to removed
2405 operand 2 of a SWITCH_EXPR.
2407 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2409 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2410 multiplication and division.
2411 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2412 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2413 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2414 (cond_udiv_optab, cond_umod_optab): New optabs.
2415 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2416 (IFN_COND_RDIV): New internal functions.
2417 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2418 TRUNC_MOD_EXPR and RDIV_EXPR.
2419 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2420 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2422 (SVE_INT_BINARY): Include mult.
2423 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2424 (optab, sve_int_op): Handle mult.
2425 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2427 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2428 for SVE_INT_BINARY_SD.
2430 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2432 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2433 (optab, sve_int_op): Handle div and udiv.
2434 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2435 for SVE_INT_BINARY_SD.
2436 (*<optab><mode>3): New insn for the same.
2438 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2440 * tree-vect-patterns.c: Include predict.h.
2441 (vect_recog_divmod_pattern): Restrict check for division support
2442 to when optimizing for size.
2444 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2446 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2447 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2448 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2449 (gimple_match_op::set_op): Likewise.
2450 (gimple_resimplify4): Declare.
2451 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2452 (expr::gen_transform): Likewise.
2453 (decision_tree::gen): Generate a simplification routine for 4 operands.
2454 * gimple-match-head.c (gimple_simplify): Add an overload for
2455 4 operands. In the top-level function, handle up to 4 call
2456 arguments and call gimple_resimplify4.
2457 (gimple_resimplify4): New function.
2458 (build_call_internal): Pass a fourth operand.
2459 (maybe_push_to_seq): Likewise.
2460 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2461 Fold VEC_COND_EXPRs of an operation and a default value into
2462 an IFN_COND_* function if possible.
2463 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2465 (SVE_COND_FP_BINARY): Include them.
2466 (optab, sve_fp_op): Handle them.
2467 (SVE_INT_BINARY_REV): New code iterator.
2468 (SVE_COND_FP_BINARY_REV): New int iterator.
2469 (commutative): New int attribute.
2470 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2472 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2474 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2475 (*cond_<optab><mode>): New patterns for reversed operands.
2477 2018-05-25 Richard Biener <rguenther@suse.de>
2479 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2480 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2481 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2482 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2483 * tree-vect-data-refs.c (everywhere): Adjust users.
2484 * tree-vect-loop.c (everywhere): Likewise.
2485 * tree-vect-slp.c (everywhere): Likewise.
2486 * tree-vect-stmts.c (everywhere): Likewise.
2487 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2489 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2491 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2493 (gcc_cv_as_section_exclude): ... this.
2494 Try Solaris as #exclude syntax.
2495 * configure: Regenerate.
2496 * config.in: Regenerate.
2497 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2499 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2500 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2502 * varasm.c (default_elf_asm_named_section): Don't check if
2503 HAVE_GAS_SECTION_EXCLUDE is defined.
2505 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2507 * doc/md.texi: Update the documentation of the cond_* optabs
2508 to mention the new final operand. Fix GET_MODE_NUNITS call.
2509 Describe the scalar case too.
2510 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2511 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2513 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2514 (get_conditional_internal_fn): Update comment.
2515 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2516 accumulator value as a final argument to conditional functions.
2517 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2518 a define_expand and add an "else" operand. Assert for now that
2519 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2520 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2521 (*cond_<optab><mode>): New patterns.
2522 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2523 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2524 (UNSPEC_COND_EOR): Delete.
2525 (optab): Remove associated mappings.
2526 (SVE_INT_BINARY): New code iterator.
2527 (sve_int_op): Remove int attribute and add "minus" to the code
2529 (SVE_COND_INT_OP): Delete.
2530 (SVE_COND_FP_OP): Rename to...
2531 (SVE_COND_FP_BINARY): ...this.
2533 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2535 * optabs.c (can_reuse_operands_p): New function.
2536 (maybe_legitimize_operands): Try to reuse the results for
2539 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2541 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2542 Add {q} suffix to insn mnemonic.
2544 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2546 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2547 (msp430_warn_func_return): New.
2549 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2551 * fold-const.c (tree_nonzero_bits): New function.
2552 * fold-const.h (tree_nonzero_bits): Likewise.
2553 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2554 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2556 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2560 * varasm.c (assemble_alias): Check ifunc_resolver only on
2563 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2566 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2567 when memory input operand is handled.
2569 2018-05-24 Luis Machado <luis.machado@linaro.org>
2571 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2573 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2575 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2577 * match.pd: Delay FMA folds until after vectorization.
2579 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2582 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2583 address check not strict.
2585 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2587 * gimple-match.h (gimple_match_op): New class.
2588 (mprts_hook): Replace parameters with a gimple_match_op *.
2589 (maybe_build_generic_op): Likewise.
2590 (gimple_simplified_result_is_gimple_val): Replace parameters with
2591 a const gimple_match_op *.
2592 (gimple_simplify): Replace code_helper * and tree * parameters with
2593 a gimple_match_op * parameter.
2594 (gimple_resimplify1): Replace code_helper *, tree and tree *
2595 parameters with a gimple_match_op * parameter.
2596 (gimple_resimplify2): Likewise.
2597 (gimple_resimplify3): Likewise.
2598 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2599 parameters with a gimple_match_op * parameter.
2600 * gimple-match-head.c (gimple_simplify): Change prototypes of
2601 auto-generated functions to take a gimple_match_op * instead of
2602 separate code_helper * and tree * parameters. Make the same
2603 change in the top-level overload and update calls to the
2604 gimple_resimplify routines. Update calls to the auto-generated
2605 functions and to maybe_push_res_to_seq in the publicly-facing
2606 operation-specific gimple_simplify overloads.
2607 (gimple_match_op::MAX_NUM_OPS): Define.
2608 (gimple_resimplify1): Replace rcode and ops with a single res_op
2609 parameter. Update call to gimple_simplify.
2610 (gimple_resimplify2): Likewise.
2611 (gimple_resimplify3): Likewise.
2612 (mprts_hook): Replace parameters with a gimple_match_op *.
2613 (maybe_build_generic_op): Likewise.
2614 (build_call_internal): Replace type, nargs and ops with
2615 a gimple_match_op *.
2616 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2617 with a single gimple_match_op *. Update calls to mprts_hook,
2618 build_call_internal and gimple_simplified_result_is_gimple_val.
2619 Factor out code that is common to the tree_code and combined_fn cases.
2620 * genmatch.c (expr::gen_transform): Replace tem_code and
2621 tem_ops with a gimple_match_op called tem_op. Update calls
2622 to the gimple_resimplify functions and maybe_push_res_to_seq.
2623 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2624 res_ops. Update call to the gimple_resimplify functions.
2625 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2626 (decision_tree::gen): Make the functions take a gimple_match_op *
2627 called res_op instead of separate res_code and res_ops parameters.
2628 Update call accordingly.
2629 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2630 and ops with a single res_op parameter. Update calls to
2631 maybe_build_generic_op and maybe_push_res_to_seq.
2632 (fold_stmt_1): Update calls to gimple_simplify and
2633 replace_stmt_with_simplification.
2634 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2635 and gimple_simplified_result_is_gimple_val.
2636 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2638 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2639 with a gimple_match_op *.
2640 (vn_nary_build_or_lookup): Likewise. Update call to
2641 vn_nary_build_or_lookup_1.
2642 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2643 gimple_match_op *. Update calls to the gimple_resimplify routines
2644 and to gimple_simplified_result_is_gimple_val.
2645 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2646 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2647 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2648 (visit_nary_op): Likewise.
2649 (visit_reference_op_load): Likewise.
2651 2018-05-23 Luis Machado <luis.machado@linaro.org>
2653 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2654 modifier for printing the step amount.
2656 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2659 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2662 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2664 * doc/sourcebuild.texi (Endianness): New subsubsection.
2666 2018-05-23 Luis Machado <luis.machado@linaro.org>
2668 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2669 <prefetch_dynamic_strides>: New const bool field.
2670 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2671 prefetch_dynamic_strides.
2672 (exynosm1_prefetch_tune): Likewise.
2673 (thunderxt88_prefetch_tune): Likewise.
2674 (thunderx_prefetch_tune): Likewise.
2675 (thunderx2t99_prefetch_tune): Likewise.
2676 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2678 (aarch64_override_options_internal): Update to set
2679 PARAM_PREFETCH_DYNAMIC_STRIDES.
2680 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2681 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2682 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2683 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2684 prefetch-dynamic-strides setting.
2686 2018-05-23 Luis Machado <luis.machado@linaro.org>
2688 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2689 <minimum_stride>: New const int field.
2690 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2691 minimum_stride field defaulting to -1.
2692 (exynosm1_prefetch_tune): Likewise.
2693 (thunderxt88_prefetch_tune): Likewise.
2694 (thunderx_prefetch_tune): Likewise.
2695 (thunderx2t99_prefetch_tune): Likewise.
2696 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2697 <default_opt_level>: Set to 3.
2698 (aarch64_override_options_internal): Update to set
2699 PARAM_PREFETCH_MINIMUM_STRIDE.
2700 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2701 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2702 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2703 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2704 stride is constant and is below the minimum stride threshold.
2706 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2708 * config/arm/arm-cpus.in (mode26): Delete.
2709 (armv4): Delete mode26 reference.
2710 * config/arm/arm.c (arm_configure_build_target): Delete use of
2713 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2715 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2717 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2718 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2720 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2721 (floatunsdidf2): Ditto.
2723 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2725 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2726 (fixuns_trunc<mode>si2_avx512f): Ditto.
2727 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2728 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2729 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2731 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2733 PR rtl-optimization/79985
2734 * df-scan.c (df_insn_refs_collect): Remove special case for
2735 global registers and asm statements.
2737 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2739 * extend.texi (Global Register Variables): Rewrite the bullet list.
2740 Note that the register is available for allocation. Note that access
2741 via inline asm must use constraints. Add note about async-signal
2742 handlers. Remove paragraph about automagic register selection.
2744 2018-05-23 Richard Biener <rguenther@suse.de>
2746 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2747 of fixed offset from memset VN.
2749 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2751 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2753 (alloc_cand_and_find_basis): Initialize first_interp field.
2754 (slsr_process_mul): Modify first_interp field.
2755 (slsr_process_add): Likewise.
2756 (slsr_process_cast): Modify first_interp field for each new
2758 (slsr_process_copy): Likewise.
2759 (dump_candidate): Dump first_interp field.
2760 (replace_mult_candidate): Process all interpretations, not just
2762 (replace_rhs_if_not_dup): Likewise.
2763 (replace_one_candidate): Likewise.
2765 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2767 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2769 (aarch64_needs_frame_chain): New function.
2770 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2772 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2775 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2776 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2777 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2778 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2779 as true for strict-align.
2780 (aarch64_can_inline_p): Perform checks even when callee has no
2781 attributes to check for strict alignment.
2782 * doc/extend.texi (AArch64 Function Attributes): Document
2784 * doc/invoke.texi: (AArch64 Options): Likewise.
2786 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2788 PR tree-optimization/85853
2789 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2790 the handling of the root of the node to...
2791 (vect_slp_analyze_node_operations_1): ...this new function,
2792 and run the whole thing with the child nodes' def types
2793 set according to their SLP node's def type.
2795 2018-05-23 Richard Biener <rguenther@suse.de>
2798 * tree-data-ref.c (create_runtime_alias_checks): Defer
2799 and ignore overflow warnings.
2801 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2803 PR tree-optimization/85822
2804 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2807 2018-05-23 Richard Biener <rguenther@suse.de>
2809 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2810 memset constants via native_interpret_expr.
2812 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2815 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2817 (cgraph_node::create_alias): Likewise.
2818 (cgraph_node::get_availability): Check ifunc_resolver instead
2819 of looking up ifunc attribute.
2820 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2821 * varasm.c (do_assemble_alias): Likewise.
2822 (assemble_alias): Likewise.
2823 (default_binds_local_p_3): Likewise.
2824 * cgraph.h (cgraph_node): Add ifunc_resolver.
2825 (cgraph_node::only_called_directly_or_aliased_p): Return false
2827 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2829 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2830 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2831 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2832 instead of looking up ifunc attribute.
2834 2018-05-22 Luis Machado <luis.machado@linaro.org>
2836 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2838 2018-05-22 Martin Sebor <msebor@redhat.com>
2841 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2842 only when expasion succeeds.
2843 (expand_builtin_strcmp): Same.
2844 (expand_builtin_strncmp): Same.
2846 2018-05-22 Martin Sebor <msebor@redhat.com>
2848 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2850 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2851 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2853 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2854 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2856 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2857 Add check for legitimate_address.
2858 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2859 (aarch64_swap_ldrstr_operands): New.
2860 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2863 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2864 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2866 * config/aarch64/aarch64.md: New patterns to generate stp
2868 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2869 single words and double words.
2870 (load_pair_sw, load_pair_dw): Likewise.
2871 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2873 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2875 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2876 for different mode ldpstp and add peephole for merged zero stores.
2878 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2880 (aarch64_gen_store_pair): Rename calls to match new patterns.
2881 (aarch64_gen_load_pair): Rename calls to match new patterns.
2882 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2883 (load_pair<DREG:mode><DREG2:mode>): ... This.
2884 (store_pair<mode>): Rename to...
2885 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2886 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2888 (V_INT_EQUIV): Handle SImode.
2889 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2892 2018-05-22 Martin Sebor <msebor@redhat.com>
2895 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2896 or ajust the presumed bound on an operation to avoid unnecessary
2899 2018-05-22 Martin Sebor <msebor@redhat.com>
2901 PR tree-optimization/85826
2902 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2903 assuming that a DECL necesarily has a constant size.
2905 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2908 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2910 2018-05-22 Richard Biener <rguenther@suse.de>
2912 PR tree-optimization/85834
2913 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2914 non-constant and non-zero memset arguments.
2916 2018-05-22 Martin Liska <mliska@suse.cz>
2919 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2921 2018-05-22 Richard Biener <rguenther@suse.de>
2923 PR tree-optimization/85863
2924 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2925 comparisons when vectype is specified.
2926 (vectorizable_condition): Do not specify vectype for
2927 vect_is_simple_cond when SLP vectorizing.
2929 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
2932 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
2933 define __ibm128 as long double.
2934 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
2935 as a distinct type when IEEE 128-bit support is enabled.
2936 (init_float128_ieee): Fix up conversions between IFmode and IEEE
2937 128-bit types to use the correct functions.
2938 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
2939 convert between 128-bit floating point types that have different
2940 modes but the same representation, instead of using gen_lowpart to
2942 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
2944 (IFKF_reg): New attributes to give the register constraints for
2946 (extend<mode>tf2_internal): New insns to mark an explicit
2947 conversion between 128-bit floating point types that have a
2948 different mode but share the same representation.
2950 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
2952 PR tree-optimization/85814
2953 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
2954 a null return from get_strinfo when unsharing the next
2955 strinfo in the chain.
2957 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2960 * varasm.c (weak_finish): Clean up weak_decls.
2962 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2964 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
2965 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
2966 UNSPEC_UADALP values.
2967 * config/aarch64/iterators.md (ABAL): New int iterator.
2970 (sur): Add mappings for the above.
2971 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
2973 (aarch64_<sur>abal<mode>_4): Likewise.
2974 (aarch64_<sur>adalp<mode>_3): Likewise.
2975 (<sur>sadv16qi): New define_expand.
2977 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
2979 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
2980 (*movdf_internal): Ditto.
2981 (*rcpsf2_sse): Ditto.
2982 (*rsqrtsf2_sse): Ditto.
2983 (*sqrt<mode>2_sse): Ditto.
2985 2018-05-21 Tamar Christina <tamar.christina@arm.com>
2987 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
2989 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
2990 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
2991 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
2992 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
2994 * config/aarch64/arm_neon.h: Likewise.
2995 * config/aarch64/iterators.md (VQ_I): New.
2997 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
2999 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3001 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3004 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3006 * config/nds32/constraints.md (S): New constraint.
3007 * config/nds32/nds32.md (call_internal): Use constraint S.
3008 (call_value_internal): Likewise.
3009 (sibcall_internal): Likewise.
3010 (sibcall_value_internal): Likewise.
3012 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3013 Chung-Ju Wu <jasonwucj@gmail.com>
3015 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3018 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3019 Chung-Ju Wu <jasonwucj@gmail.com>
3021 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3022 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3023 (nds32_rtx_costs_impl): Simplify.
3024 (nds32_address_cost_impl): Simplify.
3025 (nds32_init_rtx_costs): New function.
3026 (nds32_rtx_costs_speed_prefer): Likewise.
3027 (nds32_rtx_costs_size_prefer): Likewise.
3028 (nds32_address_cost_speed_prefer): Likewise.
3029 (nds32_address_cost_speed_fwprop): Likewise.
3030 (nds32_address_cost_size_prefer): Likewise.
3031 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3032 * config/nds32/nds32.c (nds32_option_override): Use
3033 nds32_init_rtx_costs function.
3035 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3037 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3038 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3039 (TARGET_PIPELINE_N8): Likewise.
3040 (TARGET_PIPELINE_N10): Likewise.
3041 (TARGET_PIPELINE_N13): Likewise.
3042 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3044 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3046 * config/nds32/nds32-fpu.md: Update copyright year.
3048 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3050 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3052 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3054 * config/nds32/nds32.c
3055 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3056 * config/nds32/nds32.opt (minline-asm-r15): New option.
3058 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3060 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3062 * config/nds32/nds32.md (abssi2): New pattern.
3064 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3066 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3067 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3068 Merge insn pattern from sse_cvtsi2ss<round_name> and
3069 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3070 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3071 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3072 using SWI48 mode iterator.
3073 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3074 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3075 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3076 pattern from sse_cvttss2si<round_saeonly_name>
3077 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3078 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3079 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3080 using SWI48 mode iterator.
3081 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3082 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3083 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3084 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3085 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3086 using SWI48 mode iterator.
3087 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3088 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3089 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3090 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3091 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3092 SWI48 mode iterator.
3093 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3094 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3095 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3096 pattern from sse_cvttsd2si<round_saeonly_name>
3097 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3099 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3101 * config/nds32/nds32-md-auxiliary.c
3102 (nds32_valid_smw_lwm_base_p): Refine.
3103 (nds32_output_smw_single_word): Refine.
3104 (nds32_output_smw_double_word): New.
3105 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3107 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3109 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3110 (nds32_output_stack_pop): Refine.
3111 (nds32_expand_unaligned_load): Refine.
3112 (nds32_expand_unaligned_store): Refine.
3114 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3115 Chung-Ju Wu <jasonwucj@gmail.com>
3117 * config/nds32/constants.md: Add TP_REGNUM constant.
3118 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3119 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3121 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3122 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3123 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3124 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3125 and PIC code generation.
3126 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3128 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3130 * config/nds32/nds32.md: Support TLS and PIC.
3131 * config/nds32/nds32.c: Support TLS and PIC.
3132 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3133 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3136 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3138 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3139 mode with E_ prefix.
3141 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3142 Chung-Ju Wu <jasonwucj@gmail.com>
3144 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3145 * config/nds32/nds32-md-auxiliary.c
3146 (symbolic_reference_mentioned_p): New.
3147 (nds32_legitimize_ict_address): New.
3148 (nds32_expand_ict_move): New.
3149 (nds32_indirect_call_referenced_p): New.
3150 (nds32_symbol_binds_local_p): Delete.
3151 (nds32_long_call_p): Modify.
3152 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3153 * config/nds32/nds32-protos.h
3154 (symbolic_reference_mentioned_p): Declare.
3155 (nds32_legitimize_ict_address): Declare.
3156 (nds32_expand_ict_move): Declare.
3157 (nds32_indirect_call_referenced_p): Declare.
3158 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3159 (nds32_relax_group): Use nds32_ict_const_p as condition.
3160 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3161 (nds32_asm_file_start): Output ict_model directive in asm code.
3162 (nds32_legitimate_address_p): Consider indirect call.
3163 (nds32_print_operand): Consider indirect call.
3164 (nds32_print_operand_address): Consider indirect call.
3165 (nds32_insert_attributes): Handle "indirect_call" attribute.
3166 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3167 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3168 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3169 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3170 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3171 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3172 (TARGET_ICT_MODEL_SMALL): Define.
3173 (TARGET_ICT_MODEL_LARGE): Define.
3174 * config/nds32/nds32.md (movsi): Consider ict model.
3175 (call, call_value): Consider ict model.
3176 (sibcall, sibcall_value): Consider ict model.
3177 * config/nds32/nds32.opt (mict-model): New option.
3178 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3181 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3182 Monk Chiang <sh.chiang04@gmail.com>
3183 Jim Wilson <jimw@sifive.com>
3185 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3186 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3187 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3188 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3189 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3190 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3191 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3192 compute save_libcall_adjustment properly.
3193 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3194 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3195 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3196 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3197 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3198 (ABI_SPEC): Handle mabi=ilp32e.
3199 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3200 (RVE): Add RVE mask.
3201 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3202 <-march>: Add rv32e as an example.
3204 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3207 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3208 (intra_create_variable_infos): Handle C++ constructors.
3210 2018-05-18 Martin Liska <mliska@suse.cz>
3212 * passes.def: Remove a redundant pass.
3214 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3217 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3219 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3221 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3223 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3225 (armv2, armv2a, armv3, armv3m): Delete architectures.
3226 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3227 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3228 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3230 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3231 (*mulsidi3adddi): Likewise.
3232 (mulsidi3): Likewise.
3233 (*mulsidi3_nov6): Likewise.
3234 (umulsidi3): Likewise.
3235 (umulsidi3_nov6): Likewise.
3236 (umaddsidi4): Likewise.
3237 (*umulsidi3adddi): Likewise.
3238 (smulsi3_highpart): Likewise.
3239 (*smulsi3_highpart_nov6): Likewise.
3240 (umulsi3_highpart): Likewise.
3241 (*umulsi3_highpart_nov6): Likewise.
3242 * config/arm/arm.h (arm_arch3m): Delete.
3243 * config/arm/arm.c (arm_arch3m): Delete.
3244 (arm_option_override_internal): Update armv3-related comment.
3245 (arm_configure_build_target): Delete use of isa_bit_mode32.
3246 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3247 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3248 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3250 (mulusa3): Likewise.
3251 * config/arm/arm-protos.h (arm_arch3m): Delete.
3252 * config/arm/arm-tables.opt: Regenerate.
3253 * config/arm/arm-tune.md: Likewise.
3254 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3255 deleted architectures.
3257 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3259 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3260 (armv5t, armv5te): New features.
3261 (ARMv5, ARMv5e): Delete fgroups.
3262 (ARMv5t, ARMv5te): Adjust for above changes.
3264 (armv5, armv5e): Delete arches.
3265 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3267 (*call_reg_arm): Likewise.
3268 (*call_value_reg_armv5): Likewise.
3269 (*call_value_reg_arm): Likewise.
3270 (*call_symbol): Likewise.
3271 (*call_value_symbol): Likewise.
3272 (*sibcall_insn): Likewise.
3273 (*sibcall_value_insn): Likewise.
3275 (prefetch): Likewise.
3276 (define_split and define_peephole2 dependent on arm_arch5):
3278 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3280 (TARGET_ARM_QBIT): Likewise.
3281 (TARGET_DSP_MULTIPLY): Likewise.
3282 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3283 (arm_arch5, arm_arch5e): Delete.
3284 (arm_arch5t, arm_arch5te): Declare.
3285 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3286 (arm_arch5t): Declare.
3287 (arm_option_reconfigure_globals): Update for the above.
3288 (arm_options_perform_arch_sanity_checks): Update comment, replace
3289 use of arm_arch5 with arm_arch5t.
3290 (use_return_insn): Likewise.
3291 (arm_emit_call_insn): Likewise.
3292 (output_return_instruction): Likewise.
3293 (arm_final_prescan_insn): Likewise.
3294 (arm_coproc_builtin_available): Likewise.
3295 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3296 arm_arch5e with arm_arch5t and arm_arch5te.
3297 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3298 (arm_arch5t, arm_arch5te): Declare.
3299 * config/arm/arm-tables.opt: Regenerate.
3300 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3301 * config/arm/t-multilib: Likewise.
3302 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3303 instead of arm_arch5.
3304 (*call_reg_thumb1): Likewise.
3305 (*call_value_reg_thumb1_v5): Likewise.
3306 (*call_value_reg_thumb1): Likewise.
3307 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3309 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3311 2018-05-18 Martin Liska <mliska@suse.cz>
3313 PR gcov-profile/84846
3314 * doc/gcov.texi: Document -t option of gcov tool.
3316 2018-05-18 Martin Liska <mliska@suse.cz>
3318 PR gcov-profile/84846
3319 * gcov.c (print_usage): Add new -t option.
3320 (process_args): Handle the option.
3321 (generate_results): Use stdout as output when requested by
3324 2018-05-18 Martin Liska <mliska@suse.cz>
3326 PR gcov-profile/84846
3327 * coverage.c (coverage_init): Write PWD to .gcno file.
3328 * doc/gcov.texi: Document how working directory is printed.
3329 * gcov-dump.c (dump_gcov_file): Print PWD.
3330 * gcov.c (output_intermediate_file): Likewise.
3331 (read_graph_file): Read PWD string.
3332 (output_lines): Print PWD.
3334 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3337 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3338 for retval and return false if all args to phi are zero.
3340 2018-05-18 Richard Biener <rguenther@suse.de>
3342 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3344 (evrp_dom_walker::before_dom_children): Call it.
3346 2018-05-18 Richard Biener <rguenther@suse.de>
3348 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3349 results when processing array refs with variable index.
3351 2018-05-18 Toon Moene <toon@moene.org>
3353 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3354 directly after that of -floop-interchange. Indicate that both
3355 options are enabled by default when specifying -O3.
3357 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3359 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3360 iterator. Delete separate integer-mode vec_set<mode> expander.
3361 (aarch64_simd_vec_setv2di): Delete.
3362 (vec_setv2di): Delete.
3363 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3364 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3365 the "w, r" alternative.
3367 2018-05-18 Martin Liska <mliska@suse.cz>
3369 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3370 * tree-pass.h (make_pass_lower_switch_O0): New function.
3371 * tree-switch-conversion.c (node_has_low_bound): Remove.
3372 (node_has_high_bound): Likewise.
3373 (node_is_bounded): Likewise.
3374 (class pass_lower_switch): Make it a template type and create
3376 (pass_lower_switch::execute): Add template argument.
3377 (make_pass_lower_switch): New function.
3378 (make_pass_lower_switch_O0): New function.
3379 (do_jump_if_equal): Remove.
3380 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3381 all the hard work to tree optimization passes.
3383 2018-05-18 Martin Liska <mliska@suse.cz>
3385 * dbgcnt.c (limit_low): Renamed from limit.
3386 (limit_high): New variable.
3387 (dbg_cnt_is_enabled): Check for upper limit.
3388 (dbg_cnt): Adjust dumping.
3389 (dbg_cnt_set_limit_by_index): Add new argument for high
3391 (dbg_cnt_set_limit_by_name): Likewise.
3392 (dbg_cnt_process_single_pair): Parse new format.
3393 (dbg_cnt_process_opt): Use strtok.
3394 (dbg_cnt_list_all_counters): Remove 'value' and add
3396 * doc/invoke.texi: Document changes.
3398 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3400 * doc/sourcebuild.texi (scalar_all_fma): Document.
3401 * tree.def (FMA_EXPR): Delete.
3402 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3403 * internal-fn.c (ternary_direct): New macro.
3404 (expand_ternary_optab_fn): Likewise.
3405 (direct_ternary_optab_supported_p): Likewise.
3406 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3407 * builtins.c (fold_builtin_fma): Delete.
3408 (fold_builtin_3): Don't call it.
3409 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3410 * expr.c (expand_expr_real_2): Likewise.
3411 * fold-const.c (operand_equal_p): Likewise.
3412 (fold_ternary_loc): Likewise.
3413 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3414 * gimple.c (DEFTREECODE): Likewise.
3415 * gimplify.c (gimplify_expr): Likewise.
3416 * optabs-tree.c (optab_for_tree_code): Likewise.
3417 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3418 * tree-eh.c (operation_could_trap_p): Likewise.
3419 (stmt_could_throw_1_p): Likewise.
3420 * tree-inline.c (estimate_operator_cost): Likewise.
3421 * tree-pretty-print.c (dump_generic_node): Likewise.
3422 (op_code_prio): Likewise.
3423 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3424 * tree-ssa-operands.c (get_expr_operands): Likewise.
3425 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3426 * fold-const-call.h (fold_fma): Delete.
3427 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3428 CFN_FNMA and CFN_FNMS.
3430 * genmatch.c (combined_fn): New enum.
3431 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3432 (commutative_op): New function.
3433 (commutate): Use it. Handle more than 2 operands.
3434 (dt_operand::gen_gimple_expr): Use commutative_op.
3435 (parser::parse_expr): Allow :c to be used with non-binary
3436 operators if the commutative operand is known.
3437 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3438 CFN_FMS, CFN_FNMA and CFN_FNMS.
3439 (backprop::process_assign_use): Remove FMA_EXPR handling.
3440 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3441 (gen_hsa_fma): New function.
3442 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3443 IFN_FNMA and IFN_FNMS.
3444 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3445 * gimple-fold.h (follow_all_ssa_edges): Declare.
3446 * gimple-fold.c (follow_all_ssa_edges): New function.
3447 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3448 gimple_build interface and use follow_all_ssa_edges to fold the result.
3449 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3450 instead of checking for optabs directly.
3451 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3452 rather than FMA_EXPRs.
3453 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3454 call to IFN_FMA instead of an FMA_EXPR.
3456 2018-05-17 Jim Wilson <jimw@sifive.com>
3458 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3459 sign extended promoted subreg, and the range does not have the sign bit
3460 set, then do a sign extend.
3462 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3463 test, check for sign extended subreg and/or constant operands, and
3464 do a sign extend in that case.
3466 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3468 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3469 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3471 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3472 Change logics_shift_reg to logics_shift_imm.
3473 (thunderx2t99_fp_loadpair_basic): Delete.
3474 (thunderx2t99_fp_storepair_basic): Delete.
3475 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3476 (thunderx2t99_asimd_polynomial): Delete.
3477 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3478 and neon_fp_mul_d_scalar_q.
3479 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3480 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3481 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3482 (thunderx2t99_asimd_lut): Add missing tbl types.
3483 (thunderx2t99_asimd_ext): Delete.
3484 (thunderx2t99_asimd_load1_1_mult): Delete.
3485 (thunderx2t99_asimd_load1_2_mult): Delete.
3486 (thunderx2t99_asimd_load1_ldp): New.
3487 (thunderx2t99_asimd_load1): New.
3488 (thunderx2t99_asimd_load2): Add missing *load2* types.
3489 (thunderx2t99_asimd_load3): New.
3490 (thunderx2t99_asimd_load4): New.
3491 (thunderx2t99_asimd_store1_1_mult): Delete.
3492 (thunderx2t99_asimd_store1_2_mult): Delete.
3493 (thunderx2t99_asimd_store2_mult): Delete.
3494 (thunderx2t99_asimd_store2_onelane): Delete.
3495 (thunderx2t99_asimd_store_stp): New.
3496 (thunderx2t99_asimd_store1): New.
3497 (thunderx2t99_asimd_store2): New.
3498 (thunderx2t99_asimd_store3): New.
3499 (thunderx2t99_asimd_store4): New.
3501 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3503 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3504 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3506 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3507 Segher Boessenkool <segher@kernel.crashing.org>
3510 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3513 2018-05-17 Richard Biener <rguenther@suse.de>
3515 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3516 for pruning loop and prune defs feeding only already visited PHIs.
3518 2018-05-17 Richard Biener <rguenther@suse.de>
3520 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3522 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3523 Richard Biener <rguenther@suse.de>
3525 PR tree-optimization/85793
3526 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3527 for VMAT_ELEMENTWISE.
3529 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3531 * internal-fn.h (lookup_internal_fn): Declare
3532 * internal-fn.c (lookup_internal_fn): New function.
3533 * gimple.c (gimple_build_call_from_tree): Handle calls to
3535 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3536 internal function names.
3537 * tree-pretty-print.c (dump_generic_node): Likewise.
3538 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3540 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3542 * gimple-fold.h (gimple_build): Make the function forms take
3543 combined_fn rather than built_in_function.
3544 (gimple_simplify): Likewise.
3545 * gimple-match-head.c (gimple_simplify): Likewise.
3546 * gimple-fold.c (gimple_build): Likewise.
3547 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3548 rather than gimple_build_call_internal.
3549 (get_initial_defs_for_reduction): Likewise.
3550 (vect_create_epilog_for_reduction): Likewise.
3551 (vectorizable_live_operation): Likewise.
3553 2018-05-17 Martin Liska <mliska@suse.cz>
3555 * gimple-ssa-sprintf.c (format_directive): Do not use
3556 space in between 'G_' and '('.
3558 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3561 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3562 even if the mask is not all ones.
3565 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3567 (ix86_gimple_fold_builtin): Likewise.
3570 * config/i386/i386.c: Include tree-vector-builder.h.
3571 (ix86_vector_shift_count): New function.
3572 (ix86_fold_builtin): Fold shift builtins by scalar count.
3573 (ix86_gimple_fold_builtin): Likewise.
3575 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3576 _mm512_setzero): New intrinsics.
3578 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3579 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3581 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3582 code generation for cases where splatting a value is not useful.
3583 * simplify-rtx.c (simplify_ternary_operation): Simplify
3584 vec_merge across a vec_duplicate and a paradoxical subreg forming
3585 a vector mode to a vec_concat.
3587 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3589 * config.gcc: Support "goldmont-plus".
3590 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3592 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3593 PROCESSOR_GOLDMONT_PLUS.
3594 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3595 (processor_target_table): Add "goldmont-plus".
3596 (PTA_GOLDMONT_PLUS): Define.
3597 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3598 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3599 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3600 (fold_builtin_cpu): Add "goldmont-plus".
3601 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3602 (ix86_option_override_internal): Add "goldmont-plus".
3603 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3604 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3605 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3606 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3608 2018-05-17 Richard Biener <rguenther@suse.de>
3610 PR tree-optimization/85757
3611 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3612 remove defs that only feed that PHI from further processing.
3614 2018-05-16 Jim Wilson <jimw@sifive.com>
3616 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3618 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3620 2018-05-16 Mark Wielaard <mark@klomp.org>
3622 * dwarf2out.c (count_index_strings): New function.
3623 (output_indirect_strings): Call count_index_strings and generate
3624 header for dwarf_version >= 5.
3626 2018-05-16 Mark Wielaard <mark@klomp.org>
3628 * dwarf2out.c (dwarf_FORM): New function.
3629 (set_indirect_string): Use dwarf_FORM.
3630 (reset_indirect_string): Likewise.
3631 (size_of_die): Likewise.
3632 (value_format): Likewise.
3633 (output_die): Likewise.
3634 (add_skeleton_AT_string): Likewise.
3635 (output_macinfo_op): Likewise.
3636 (index_string): Likewise.
3637 (output_index_string_offset): Likewise.
3638 (output_index_string): Likewise.
3639 (count_index_strings): Likewise.
3641 2018-05-16 Carl Love <cel@us.ibm.com>
3643 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3644 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3646 2018-05-16 Martin Jambor <mjambor@suse.cz>
3648 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3649 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3651 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3653 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3654 (fnma<mode>4): Likewise.
3655 (fms<mode>4): Likewise.
3656 (fnms<mode>4): Likewise.
3657 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3658 (aarch64_fnma<mode>4): Likewise.
3659 (aarch64_fms<mode>4): Likewise.
3660 (aarch64_fnms<mode>4): Likewise.
3661 (aarch64_fnmadd<mode>4): Likewise.
3663 2018-05-16 Jason Merrill <jason@redhat.com>
3665 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3667 2018-05-16 Richard Biener <rguenther@suse.de>
3669 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3670 (dump_stmt_cost): Declare.
3671 (add_stmt_cost): Dump cost we add.
3672 (add_stmt_costs): New function.
3673 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3675 (vect_analyze_stmt): Adjust prototype.
3676 (vectorizable_condition): Likewise.
3677 (vectorizable_live_operation): Likewise.
3678 (vectorizable_reduction): Likewise.
3679 (vectorizable_induction): Likewise.
3680 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3681 cost vector to pass to vectorizable_ and record afterwards.
3682 (vect_model_reduction_cost): Take cost vector argument and adjust.
3683 (vect_model_induction_cost): Likewise.
3684 (vectorizable_reduction): Likewise.
3685 (vectorizable_induction): Likewise.
3686 (vectorizable_live_operation): Likewise.
3687 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3688 SLP_TREE_NUMBER_OF_VEC_STMTS.
3689 (vect_analyze_slp_cost_1): Remove.
3690 (vect_analyze_slp_cost): Likewise.
3691 (vect_slp_analyze_node_operations): Take visited args and
3692 a target cost vector. Avoid processing already visited stmt sets.
3693 (vect_slp_analyze_operations): Use a local cost vector to gather
3694 costs and register those of non-discarded instances.
3695 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3696 (vect_schedule_slp_instance): Remove copying of
3697 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3699 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3700 adding cost. Record cost entry location.
3701 (vect_prologue_cost_for_slp_op): Function to compute cost of
3702 a constant or invariant generated for SLP vect in the prologue,
3703 split out from vect_analyze_slp_cost_1.
3704 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3705 (vect_model_promotion_demotion_cost): Likewise.
3706 (vect_model_store_cost): Likewise, make static.
3707 (vect_model_load_cost): Likewise.
3708 (vectorizable_bswap): Add cost vector arg and adjust.
3709 (vectorizable_call): Likewise.
3710 (vectorizable_simd_clone_call): Likewise.
3711 (vectorizable_conversion): Likewise.
3712 (vectorizable_assignment): Likewise.
3713 (vectorizable_shift): Likewise.
3714 (vectorizable_operation): Likewise.
3715 (vectorizable_store): Likewise.
3716 (vectorizable_load): Likewise.
3717 (vectorizable_condition): Likewise.
3718 (vectorizable_comparison): Likewise.
3719 (can_vectorize_live_stmts): Likewise.
3720 (vect_analyze_stmt): Likewise.
3721 (vect_transform_stmt): Adjust calls to vectorizable_*.
3722 * tree-vectorizer.c: Include gimple-pretty-print.h.
3723 (dump_stmt_cost): New function.
3725 2018-05-16 Richard Biener <rguenther@suse.de>
3727 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3728 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3729 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3730 (check_name): New callback.
3731 (dse_classify_store): Track cycles via a visited bitmap of PHI
3732 defs and simplify handling of in-loop and across loop dead stores
3733 and properly fail for loop-variant refs. Handle byte-tracking with
3734 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3737 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3739 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3740 (vect_get_mask_type_for_stmt): Likewise.
3741 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3743 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3744 to determine the statement's vector type and the vector type that
3745 should be used for calculating nunits. Deal with cases in which
3746 the type has to be deferred.
3747 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3748 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3749 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3750 (vect_determine_vf_for_stmt): New functions, split out from...
3751 (vect_determine_vectorization_factor): ...here.
3752 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3753 (vect_get_mask_type_for_stmt): New functions, split out from
3754 vect_determine_vectorization_factor.
3756 2018-05-16 Richard Biener <rguenther@suse.de>
3758 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3759 verify the [VEC_]COND_EXPR embedded comparison.
3761 2018-05-15 Martin Sebor <msebor@redhat.com>
3763 PR tree-optimization/85753
3764 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3765 RECORD_TYPE in addition to ARRAY_TYPE.
3767 2018-05-15 Martin Sebor <msebor@redhat.com>
3770 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3772 2018-05-15 Richard Biener <rguenther@suse.de>
3774 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3775 add by_clobber_p one. Change algorithm to collect all defs
3776 representing uses we need to walk and try reducing them to
3777 a single one before failing.
3778 (dse_dom_walker::dse_optimize_stmt): Adjust.
3780 2018-05-13 Mark Wielaard <mark@klomp.org>
3782 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3783 (size_of_loc_descr): Likewise.
3784 (output_loc_operands): Likewise.
3785 (output_loc_operands_raw): Likewise.
3786 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3787 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3788 (hash_loc_operands): Likewise.
3789 (compare_loc_operands): Likewise.
3791 2018-05-14 Mark Wielaard <mark@klomp.org>
3793 * dwarf2out.c (count_index_addrs): New function.
3794 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3796 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3798 PR tree-optimization/83648
3799 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3800 return value as malloc candidate.
3802 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3805 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3806 param as true in call to suggest_attribute.
3808 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3810 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3811 -mreadonly-in-sdata.
3813 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3815 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3817 (aarch64_crypto_aesd_fused): Likewise.
3819 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3821 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3822 (movsi_aarch64): Likewise.
3823 (load_pairsi): Likewise.
3824 (load_pairdi): Likewise.
3825 (store_pairsi): Likewise.
3826 (store_pairdi): Likewise.
3827 (load_pairsf): Likewise.
3828 (load_pairdf): Likewise.
3829 (store_pairsf): Likewise.
3830 (store_pairdf): Likewise.
3831 (zero_extend): Likewise.
3832 (trunc): Swap alternatives.
3833 (fcvt_target): Add '?' to prefer w over r.
3835 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3838 * config/i386/i386.md: Disallow non-commutative arithmetics in
3839 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3840 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3841 in the peephole2 before it.
3843 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3845 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3846 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3847 (ix86_handle_option): Handle -mcldemote.
3848 * config.gcc: New header.
3849 * config/i386/cldemoteintrin.h: New file.
3850 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3851 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3853 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3854 OPTION_MASK_ISA_CLDEMOTE.
3855 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3856 (ix86_valid_target_attribute_inner_p): Ditto.
3857 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3858 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3859 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3860 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3861 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3863 * config/i386/i386.opt: Add -mcldemote.
3864 * config/i386/x86intrin.h: New header.
3865 * doc/invoke.texi: Add -mcldemote.
3867 2018-05-14 Richard Biener <rguenther@suse.de>
3869 * doc/match-and-simplify.texi: Adjust :s documentation.
3871 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3873 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3874 intended memcpy size.
3875 (REORDER_45): Likewise.
3877 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3879 * sort.cc: New file.
3880 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3881 * vec.c (qsort_chk): Use gcc_qsort.
3882 * Makefile.in (OBJS-libcommon): Add sort.o.
3883 (build/sort.o): New target. Use it...
3884 (BUILD_RTL): ... here, and...
3885 (build/gencfn-macros): ... here, and...
3886 (build/genmatch): ... here.
3888 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3889 Chung-Ju Wu <jasonwucj@gmail.com>
3891 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3892 * config/nds32/nds32-graywolf.md: New file.
3893 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3894 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3896 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3897 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3898 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3899 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3900 * config/nds32/pipelines.md: Include n15 settings.
3902 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3903 Chung-Ju Wu <jasonwucj@gmail.com>
3905 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3906 * config/nds32/nds32-n13.md: New file.
3907 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3908 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3910 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3911 * config/nds32/nds32.md (pipeline_model): Add n13.
3912 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3913 * config/nds32/pipelines.md: Include n13 settings.
3915 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3916 Chung-Ju Wu <jasonwucj@gmail.com>
3918 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3919 * config/nds32/nds32-n10.md: New file.
3920 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3921 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3923 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3924 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3925 * config/nds32/nds32.md (pipeline_model): Add n10.
3926 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
3927 * config/nds32/pipelines.md: Include n10 settings.
3929 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
3930 Kito Cheng <kito.cheng@gmail.com>
3931 Chung-Ju Wu <jasonwucj@gmail.com>
3933 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
3934 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3935 Add enum values for DSP extension instructions.
3936 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
3938 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
3939 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
3941 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
3942 * config/nds32/nds32-dspext.md: New file for DSP implementation.
3943 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
3944 * config/nds32/nds32-intrinsic.md: Likewise.
3945 * config/nds32/nds32_intrinsic.h: Likewise.
3946 * config/nds32/nds32-md-auxiliary.c: Likewise.
3947 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
3948 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
3949 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
3950 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
3951 * config/nds32/nds32-protos.h: New declarations for DSP extension.
3952 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
3953 TYPE_DMAC in switch statement.
3954 * config/nds32/nds32.c: New checking and implementation for DSP
3955 extension instructions.
3956 * config/nds32/nds32.h: Likewise.
3957 * config/nds32/nds32.md: Likewise.
3958 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
3959 * config/nds32/predicates.md: Implement new predicates for DSP
3962 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
3964 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
3965 Reformat alternatives and attributes so it is easier to identify
3966 which constraints/attributes go with which instruction.
3967 (mov<mode>_hardfloat32, FMOVE64): Likewise.
3968 (mov<mode>_softfloat32, FMOVE64): Likewise.
3969 (mov<mode>_hardfloat64, FMOVE64): Likewise.
3970 (mov<mode>_softfloat64, FMOVE64): Likewise.
3972 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
3974 * doc/extend.texi (PowerPC Built-in Functions): Rename this
3976 (Basic PowerPC Built-in Functions): The new name of the
3977 subsection previously known as "PowerPC Built-in Functions".
3978 (Basic PowerPC Built-in Functions Available on all Configurations):
3980 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
3981 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
3982 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
3983 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
3985 2018-05-11 Martin Jambor <mjambor@suse.cz>
3988 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
3991 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
3994 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
3996 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
3998 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
3999 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4000 (ix86_handle_option): Handle -mwaitpkg.
4001 * config.gcc: New header.
4002 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4003 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4004 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4006 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4007 OPTION_MASK_ISA_WAITPKG.
4008 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4009 (ix86_option_override_internal): Add PTA_WAITPKG.
4010 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4011 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4012 IX86_BUILTIN_TPAUSE.
4013 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4014 __builtin_ia32_umwait and __builtin_ia32_tpause.
4015 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4016 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4017 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4018 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4019 UNSPECV_TPAUSE): New.
4020 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4021 * config/i386/i386.opt: Add -mwaitpkg.
4022 * config/i386/waitpkgintrin.h: New file.
4023 * config/i386/x86intrin.h: New header.
4024 * doc/invoke.texi: Add -mwaitpkg.
4026 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4029 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4031 (cortex-m0): Use armv6s-m isa.
4032 (cortex-m0plus): Likewise.
4033 (cortex-m1): Likewise.
4034 (cortex-m0.small-multiply): Likewise.
4035 (cortex-m0plus.small-multiply): Likewise.
4036 (cortex-m1.small-multiply): Likewise.
4038 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4039 Jakub Jelinek <jakub@redhat.com>
4041 PR tree-optimization/85692
4042 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4043 source permute as well.
4045 2018-05-11 Martin Liska <mliska@suse.cz>
4048 * doc/extend.texi: Document LLVM style format for no_sanitize
4051 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4053 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4054 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4055 (mode_supports_vsx_dform_quad): Likewise.
4056 (mode_supports_vmx_dform): Move these functions to be next to the
4057 other mode_supports functions.
4058 (mode_supports_dq_form): Likewise.
4059 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4060 mode_supports_dq_form.
4061 (reg_offset_addressing_ok_p): Likewise.
4062 (offsettable_ok_by_alignment): Likewise.
4063 (rs6000_legitimate_offset_address_p): Likewise.
4064 (legitimate_lo_sum_address_p): Likewise.
4065 (rs6000_legitimize_address): Likewise.
4066 (rs6000_legitimize_reload_address): Likewise.
4067 (rs6000_secondary_reload_inner): Likewise.
4068 (rs6000_preferred_reload_class): Likewise.
4069 (rs6000_output_move_128bit): Likewise.
4071 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4073 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4074 Generate SImode target register for null target.
4075 <case IX86_BUILTIN_XGETBV>: Ditto.
4076 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4077 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4079 2018-05-10 Carl Love <cel@us.ibm.com>
4081 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4082 dcbtt and dcbtstt if operands[2] is 0.
4084 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4087 * config/i386/sse.md (usadv64qi): New expander.
4089 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4091 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4092 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4093 -maltivec=be support.
4094 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4095 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4096 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4097 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4098 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4099 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4100 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4101 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4102 altivec_vsumsws): Adjust.
4103 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4104 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4105 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4107 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4108 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4109 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4110 (altivec_lve<VI_char>x): Delete expand.
4111 (*altivec_lve<VI_char>x_internal): Rename to...
4112 (altivec_lve<VI_char>x): ... this.
4113 (altivec_lvxl_<mode>): Delete expand.
4114 (*altivec_lvxl_<mode>_internal): Rename to ...
4115 (altivec_lvxl_<mode>): ... this.
4116 (altivec_stvxl_<mode>): Delete expand.
4117 (*altivec_stvxl_<mode>_internal): Rename to ...
4118 (altivec_stvxl_<mode>): ... this.
4119 (altivec_stve<VI_char>x): Delete expand.
4120 (*altivec_stve<VI_char>x_internal): Rename to ...
4121 (altivec_stve<VI_char>x): ... this.
4122 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4123 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4124 reduc_plus_scal_<mode>): Adjust.
4125 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4127 (rs6000_cpu_cpp_builtins): Adjust.
4128 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4129 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4130 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4131 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4132 -maltivec=be support.
4133 (rs6000_split_vec_extract_var): Adjust.
4134 (rs6000_split_v4si_init): Adjust.
4135 (swap_selector_for_mode): Delete.
4136 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4137 altivec_expand_stvex_be): Delete.
4138 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4139 -maltivec=be support.
4140 (rs6000_gimple_fold_builtin): Ditto.
4141 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4143 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4144 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4145 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4146 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4147 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4148 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4149 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4150 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4151 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4152 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4153 anonymous split): Adjust.
4154 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4155 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4157 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4159 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4160 when --with-gxx-include-dir is also specified.
4161 * configure: Regenerate.
4163 2018-05-09 Jim Wilson <jimw@sifive.com>
4166 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4167 * config/riscv/t-withmultilib: New.
4168 * config/riscv/withmultilib.h: New.
4169 * doc/install.texi: Document RISC-V --with-multilib-list support.
4171 2018-05-09 Richard Biener <rguenther@suse.de>
4173 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4175 (vect_bb_vectorization_profitable_p): Adjust. Compute
4176 actual scalar cost using the cost vector and the add_stmt_cost
4179 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4181 PR rtl-optimization/85645
4182 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4183 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4185 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4187 PR rtl-optimization/85645
4188 * shrink-wrap.c (spread_components): Return a boolean saying if
4189 anything was changed.
4190 (try_shrink_wrapping_separate): Iterate spread_components until
4191 nothing changes anymore.
4193 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4195 PR rtl-optimization/85645
4196 * regrename.c (build_def_use): Also kill the chains that include the
4197 destination of a REG_CFA_REGISTER note.
4199 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4201 PR rtl-optimization/85645
4202 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4203 insn that has a REG_CFA_REGISTER note.
4205 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4207 * cfgexpand.c (expand_clobber): New function.
4208 (expand_gimple_stmt_1): Use it.
4209 * tree-vect-stmts.c (vect_clobber_variable): New function,
4211 (vectorizable_simd_clone_call): ...here.
4212 (vectorizable_store): Emit a clobber either side of an
4213 IFN_STORE_LANES sequence.
4214 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4216 2018-05-09 Tom de Vries <tom@codesourcery.com>
4219 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4220 (define_insn "trap_if_false"): Add exit after trap.
4222 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4224 PR rtl-optimization/85638
4225 * bb-reorder.c: Include common/common-target.h.
4226 (create_forwarder_block): New function extracted from...
4227 (fix_up_crossing_landing_pad): ...here. Rename into...
4228 (dw2_fix_up_crossing_landing_pad): ...this.
4229 (sjlj_fix_up_crossing_landing_pad): New function.
4230 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4231 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4232 from both partitions and exit the loop after one iteration.
4234 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4237 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4239 (Basic PowerPC Built-in Functions): The new name of the
4240 subsection previously known as "PowerPC Built-in Functions".
4241 (Basic PowerPC Built-in Functions Available on all Configurations):
4243 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4245 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4247 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4249 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4252 2018-05-08 Jim Wilson <jimw@sifive.com>
4254 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4255 (LD_EMUL_SUFFIX): New.
4256 (LINK_SPEC): Use it.
4258 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4260 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4262 (Basic PowerPC Built-in Functions): The new name of the
4263 subsection previously known as "PowerPC Built-in Functions".
4264 (Basic PowerPC Built-in Functions Available on all Configurations):
4266 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4268 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4270 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4272 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4275 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4278 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4279 after cmpelim optimization.
4281 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4283 * config.gcc: Support "goldmont".
4284 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4285 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4287 * config/i386/i386.c (m_GOLDMONT): Define.
4288 (processor_target_table): Add "goldmont".
4289 (PTA_GOLDMONT): Define.
4290 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4291 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4292 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4293 (fold_builtin_cpu): Add "goldmont".
4294 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4295 (ix86_option_override_internal): Add "goldmont".
4296 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4297 (processor_type): Add PROCESSOR_GOLDMONT.
4298 * config/i386/i386.md: Add CPU "glm".
4299 * config/i386/glm.md: New file.
4300 * config/i386/x86-tune.def: Add m_GOLDMONT.
4301 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4303 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4306 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4308 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4309 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4310 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4313 * config/i386/i386.c (ix86_fold_builtin): Handle
4314 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4317 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4318 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4320 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4323 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4324 (check_arch): Likewise.
4325 (check_fpu): Return the result rather than printing it.
4326 (end arch): Fix operator precedence.
4327 (end cpu): Likewise.
4328 (END): Print the result from check_fpu.
4330 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4331 Alan Hayward <alan.hayward@arm.com>
4332 David Sherwood <david.sherwood@arm.com>
4334 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4335 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4336 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4337 (*fcmuo<mode>_and): New patterns.
4339 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4341 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4342 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4343 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4344 (cmp_op, sve_imm_con): New code attributes.
4345 (SVE_COND_INT_CMP, imm_con): Delete.
4346 (cmp_op): Remove above unspecs from int attribute.
4347 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4349 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4350 comparison-specific unspecs.
4351 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4352 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4353 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4354 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4355 (*vec_fcm<cmp_op><mode>): Rename to...
4356 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4357 (*vec_fcmuo<mode>): Rename to...
4358 (*fcmuo<mode>): ...this and adjust likewise.
4359 (*pred_fcm<cmp_op><mode>): New pattern.
4360 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4361 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4363 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4365 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4366 (aarch64_emit_sve_predicated_cond): New function.
4367 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4368 (aarch64_emit_unspec_cond_or): Replace with...
4369 (aarch64_emit_sve_or_conds): ...this new function. Use
4370 aarch64_emit_sve_ptrue_op for the individual comparisons and
4371 aarch64_emit_binop to OR them together.
4372 (aarch64_emit_inverted_unspec_cond): Replace with...
4373 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4374 aarch64_emit_sve_ptrue_op for the comparison and
4375 aarch64_emit_unop to invert the result.
4376 (aarch64_expand_sve_vec_cmp_float): Update after the above
4377 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4379 2018-05-07 Nathan Sidwell <nathan@acm.org>
4381 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4382 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4383 (Backwards Compatibility): Likewise.
4385 2018-05-07 Luis Machado <luis.machado@linaro.org>
4389 2018-05-07 Luis Machado <luis.machado@linaro.org>
4391 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4392 <prefetch_dynamic_strides>: New const bool field.
4393 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4394 prefetch_dynamic_strides.
4395 (exynosm1_prefetch_tune): Likewise.
4396 (thunderxt88_prefetch_tune): Likewise.
4397 (thunderx_prefetch_tune): Likewise.
4398 (thunderx2t99_prefetch_tune): Likewise.
4399 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4401 (aarch64_override_options_internal): Update to set
4402 PARAM_PREFETCH_DYNAMIC_STRIDES.
4403 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4404 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4405 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4406 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4407 prefetch-dynamic-strides setting.
4409 2018-05-07 Luis Machado <luis.machado@linaro.org>
4411 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4412 <minimum_stride>: New const int field.
4413 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4414 minimum_stride field.
4415 (exynosm1_prefetch_tune): Likewise.
4416 (thunderxt88_prefetch_tune): Likewise.
4417 (thunderx_prefetch_tune): Likewise.
4418 (thunderx2t99_prefetch_tune): Likewise.
4419 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4420 (aarch64_override_options_internal): Update to set
4421 PARAM_PREFETCH_MINIMUM_STRIDE.
4422 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4423 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4424 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4425 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4426 stride is constant and is below the minimum stride threshold.
4428 2018-05-07 Luis Machado <luis.machado@linaro.org>
4430 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4433 2018-05-07 Luis Machado <luis.machado@linaro.org>
4435 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4436 <prefetch_dynamic_strides>: New const bool field.
4437 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4438 prefetch_dynamic_strides.
4439 (exynosm1_prefetch_tune): Likewise.
4440 (thunderxt88_prefetch_tune): Likewise.
4441 (thunderx_prefetch_tune): Likewise.
4442 (thunderx2t99_prefetch_tune): Likewise.
4443 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4445 (aarch64_override_options_internal): Update to set
4446 PARAM_PREFETCH_DYNAMIC_STRIDES.
4447 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4448 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4449 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4450 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4451 prefetch-dynamic-strides setting.
4453 2018-05-07 Luis Machado <luis.machado@linaro.org>
4455 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4456 <minimum_stride>: New const int field.
4457 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4458 minimum_stride field.
4459 (exynosm1_prefetch_tune): Likewise.
4460 (thunderxt88_prefetch_tune): Likewise.
4461 (thunderx_prefetch_tune): Likewise.
4462 (thunderx2t99_prefetch_tune): Likewise.
4463 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4464 (aarch64_override_options_internal): Update to set
4465 PARAM_PREFETCH_MINIMUM_STRIDE.
4466 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4467 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4468 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4469 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4470 stride is constant and is below the minimum stride threshold.
4472 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4475 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4476 the type is addressable. Don't force op into register if it has
4479 2018-05-05 Roland McGrath <mcgrathr@google.com>
4482 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4483 any section for which we don't know a specific type it should have,
4484 regardless of name. Previously this was done only for the exact
4485 names ".init_array", ".fini_array", and ".preinit_array".
4486 (default_elf_asm_named_section): Add comment about
4487 relationship with default_section_type_flags and SECTION_NOTYPE.
4488 (get_section): Don't consider it a type conflict if one side has
4489 SECTION_NOTYPE and the other doesn't, as long as neither has the
4490 SECTION_BSS et al used in the default_section_type_flags logic.
4492 2018-05-05 Tom de Vries <tom@codesourcery.com>
4495 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4496 (workaround_barsyncs): New function.
4497 (nvptx_reorg): Use workaround_barsyncs.
4498 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4499 (define_expand "nvptx_membar_cta"): New define_expand.
4500 (define_insn "*nvptx_membar_cta"): New insn.
4502 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4504 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4505 To improve optimization opportunities.
4506 * builtin-types.def: The new needed builtin types for the above.
4508 2018-05-04 Richard Biener <rguenther@suse.de>
4510 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4511 * gimple-ssa-store-merging.c
4512 (imm_store_chain_info::output_merged_store): Remove redundant create,
4513 release split_store vector contents on failure.
4514 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4515 scalar stmt vector on cache hit.
4517 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4519 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4521 * config.gcc (powerpc-xilinx-eabi*): Remove.
4522 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4524 (fusion_addis_mem_combo_load): Ditto.
4525 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4527 (rs6000_cpu_cpp_builtins): Ditto.
4528 * config/rs6000/rs6000-linux.c
4529 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4530 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4531 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4533 (rs6000_setup_reg_addr_masks): Ditto.
4534 (rs6000_init_hard_regno_mode_ok): Ditto.
4535 (rs6000_option_override_internal): Ditto.
4536 (legitimate_lo_sum_address_p): Ditto.
4537 (rs6000_legitimize_address): Ditto.
4538 (rs6000_legitimize_reload_address): Ditto.
4539 (rs6000_legitimate_address_p): Ditto.
4540 (abi_v4_pass_in_fpr): Ditto.
4541 (setup_incoming_varargs): Ditto.
4542 (rs6000_gimplify_va_arg): Ditto.
4543 (rs6000_split_multireg_move): Ditto.
4544 (rs6000_savres_strategy): Ditto.
4545 (rs6000_emit_prologue_components): Ditto.
4546 (rs6000_emit_epilogue_components): Ditto.
4547 (rs6000_emit_prologue): Ditto.
4548 (rs6000_emit_epilogue): Ditto.
4549 (rs6000_elf_file_end): Ditto.
4550 (rs6000_function_value): Ditto.
4551 (rs6000_libcall_value): Ditto.
4552 * config/rs6000/rs6000.h: Ditto.
4553 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4554 (TARGET_MINMAX): ... this. New.
4555 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4556 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4557 (*movsi_internal1_single): Delete.
4558 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4559 mfpu=, mxilinx-fpu): Delete.
4560 * config/rs6000/singlefp.h: Delete.
4561 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4562 * config/rs6000/t-rs6000: Ditto.
4563 * config/rs6000/t-xilinx: Delete.
4564 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4565 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4566 (VStype_simple): Delete.
4567 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4568 * config/rs6000/xfpu.h: Delete.
4569 * config/rs6000/xfpu.md: Delete.
4570 * config/rs6000/xilinx.h: Delete.
4571 * config/rs6000/xilinx.opt: Delete.
4572 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4573 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4575 2018-05-04 Tom de Vries <tom@codesourcery.com>
4578 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4581 2018-05-04 Richard Biener <rguenther@suse.de>
4584 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4585 (expand_complex_div_wide): Likewise.
4586 (expand_complex_operations_1): Likewise.
4587 (expand_complex_libcall): Preserve EH info of the original stmt.
4588 (tree_lower_complex): Handle removed blocks.
4589 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4590 on complex multiplication and division libcall builtins.
4592 2018-05-04 Richard Biener <rguenther@suse.de>
4595 * fold-const.c (negate_expr_p): Restrict negation of operand
4596 zero of a division to when we know that can happen without
4598 (fold_negate_expr_1): Likewise.
4600 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4603 * real.h (real_nextafter): Declare.
4604 * real.c (real_nextafter): New function.
4605 * fold-const-call.c (fold_const_nextafter): New function.
4606 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4607 CASE_CFN_NEXTTOWARD.
4608 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4609 even when arg1_mode is different from arg0_mode.
4611 2018-05-03 Nathan Sidwell <nathan@acm.org>
4613 * doc/extend.texi (Deprecated Features): Remove
4615 (Backwards Compatibility): Likewise.
4616 * doc/invoke.texi (C++ Language Options): Likewise.
4617 (C++ Dialect Options): Likewise.
4619 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4622 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4623 _mm512_mask_mullox_epi64): New intrinsics.
4625 2018-05-03 Tom de Vries <tom@codesourcery.com>
4628 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4629 dump files): Add offload-tree.
4631 2018-05-03 Richard Biener <rguenther@suse.de>
4633 PR tree-optimization/85615
4634 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4635 to loops not nested in BBs loop father to avoid creating multi-entry
4638 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4640 PR tree-optimization/70291
4641 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4642 arguments. Change return type to tree. Emit libcall as a new
4643 statement rather than replacing existing one when inplace_p is true.
4644 (expand_complex_multiplication_components): New function.
4645 (expand_complex_multiplication): Expand floating-point complex
4646 multiplication using the above.
4647 (expand_complex_division): Rename inner_type parameter to type.
4648 Update expand_complex_libcall call-site.
4649 (expand_complex_operations_1): Update expand_complex_multiplication
4650 and expand_complex_division call-sites.
4652 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4655 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4656 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4657 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4658 the highest significant bit of the shift count mask is clear. In
4659 check whether and[sq]i3 is needed verify that all significant bits
4660 of the shift count other than the highest are set.
4662 2018-05-02 Tom de Vries <tom@codesourcery.com>
4665 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4666 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4667 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4668 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4669 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4670 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4671 __builtin_goacc_parlevel_size.
4673 2018-05-02 Richard Biener <rguenther@suse.de>
4675 PR tree-optimization/85597
4676 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4677 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4680 2018-05-02 Tom de Vries <tom@codesourcery.com>
4683 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4684 dump files): Add ltrans-tree.
4686 2018-05-02 Tom de Vries <tom@codesourcery.com>
4689 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4690 dump files): Add wpa-ipa.
4692 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4694 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4695 powerpc*-*-linux*paired* target.
4696 * config/rs6000/750cl.h: Delete.
4697 * config/rs6000/paired.h: Delete.
4698 * config/rs6000/paired.md: Delete.
4699 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4701 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4702 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4703 comment. Remove paired float support.
4704 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4705 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4707 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4708 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4710 * config/rs6000/rs6000.c: Remove paired float support.
4711 (paired_expand_vector_init, paired_expand_vector_move,
4712 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4713 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4714 paired_expand_builtin, paired_expand_predicate_builtin,
4715 paired_init_builtins): Delete.
4716 * config/rs6000/rs6000.h: Remove paired float support.
4717 * config/rs6000/rs6000.md: Remove paired float support.
4718 (move_from_CR_ov_bit): Delete.
4719 * config/rs6000/rs6000.opt (mpaired): Delete.
4720 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4721 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4723 2018-05-02 Richard Biener <rguenther@suse.de>
4726 * gimplify.c (gimplify_save_expr): When in SSA form allow
4727 SAVE_EXPRs to compute to SSA vars.
4729 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4732 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4733 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4734 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4735 clobber operands[2], instead use a new pseudo. Formatting fixes.
4737 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4739 PR tree-optimization/85586
4740 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4741 exit early for statements in the same group if the accesses are
4744 2018-05-02 Tom de Vries <tom@codesourcery.com>
4747 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4750 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4752 PR tree-optimization/85143
4753 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4755 2018-05-01 Tom de Vries <tom@codesourcery.com>
4758 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4759 not found" error message.
4761 2018-05-01 Tom de Vries <tom@codesourcery.com>
4764 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4765 * vec.c (test_ordered_remove_if): New function.
4766 (vec_c_tests): Call test_ordered_remove_if.
4767 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4768 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4769 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4770 VEC_ORDERED_REMOVE_IF.
4772 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4774 PR tree-optimization/82665
4775 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4776 pointer subtraction where arguments come from a memchr call.
4778 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4780 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4781 --push-state --as-needed and --pop-state instead of --as-needed and
4782 --no-as-needed if ld supports it.
4783 * configure: Regenerated.
4786 * doc/install.texi2html: Replace _002d with - and _002a with * in
4787 generated html files using sed.
4789 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4792 * gcc-rich-location.c (blank_line_before_p): New function.
4793 (use_new_line): New function.
4794 (gcc_rich_location::add_fixit_insert_formatted): New function.
4795 * gcc-rich-location.h
4796 (gcc_rich_location::add_fixit_insert_formatted): New function.
4798 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4800 * selftest.c (assert_streq): Rename "expected" and "actual" to
4801 "val1" and "val2". Extend NULL-handling to cover both inputs
4802 symmetrically, while still requiring both to be non-NULL for a pass.
4803 * selftest.h (assert_streq): Rename "expected" and "actual" to
4805 (ASSERT_EQ): Likewise.
4806 (ASSERT_EQ_AT): Likewise.
4807 (ASSERT_KNOWN_EQ): Likewise.
4808 (ASSERT_KNOWN_EQ_AT): Likewise.
4809 (ASSERT_NE): Likewise.
4810 (ASSERT_MAYBE_NE): Likewise.
4811 (ASSERT_MAYBE_NE_AT): Likewise.
4812 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4813 the assertion to pass.
4814 (ASSERT_STREQ_AT): Likewise.
4816 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4818 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4819 interaction with -pie.
4821 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4823 * selftest.h: Fix alphabetization of per-source-file selftest
4826 2018-04-30 Jason Merrill <jason@redhat.com>
4828 PR c++/61982 - dead stores to destroyed objects.
4829 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4832 2018-04-30 Jason Merrill <jason@redhat.com>
4834 * tree.c (build_clobber): New.
4835 * tree.h: Declare it.
4836 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4838 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4840 * diagnostic-show-locus.c (layout::layout): Update for
4841 location_get_source_line returning a char_span.
4842 (struct char_span): Move to input.h.
4843 (struct correction): Update for fields in char_span becoming
4845 (struct source_line): Update for location_get_source_line
4846 returning a char_span.
4847 (layout::print_line): Likewise.
4848 * edit-context.c (edited_file::print_content): Likewise.
4849 (edited_file::print_diff_hunk): Likewise.
4850 (edited_file::print_run_of_changed_lines): Likewise.
4851 (edited_file::get_num_lines): Likewise.
4852 (edited_line::edited_line): Likewise.
4853 * final.c (asm_show_source): Likewise.
4854 * input.c (location_get_source_line): Convert return type
4855 from const char * to char_span, losing the final "line_len"
4857 (dump_location_info): Update for the above.
4858 (get_substring_ranges_for_loc): Likewise. Use a char_span
4859 when handling the literal within the line.
4860 (test_reading_source_line): Update for location_get_source_line
4861 returning a char_span.
4862 * input.h (class char_span): Move here from
4863 diagnostic-show-locus.c, converting from a struct to a class.
4864 Make data members private.
4865 (char_span::operator bool): New.
4866 (char_span::length): New.
4867 (char_span::get_buffer): New.
4868 (char_span::operator[]): New.
4869 (char_span::subspan): Make const.
4870 (char_span::xstrdup): New.
4871 (location_get_source_line): Convert return type from const char *
4872 to char_span, losing the final "line_size" param.
4874 2018-04-30 Jan Hubicka <jh@suse.cz>
4876 * lto-wrapper.c (ltrans_priorities): New static var.
4877 (cmp_priority): New.
4878 (run_gcc): Read priorities and if doing parallel build order
4879 the Makefile by them.
4881 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4883 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4885 2018-04-30 Richard Biener <rguenther@suse.de>
4887 * tree-cfg.c (verify_address): Remove base argument, add
4888 flag whether to check TREE_ADDRESSABLE and do that.
4889 (verify_expr): Remove.
4890 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4891 (verify_gimple_assign_single): Likewise.
4892 (verify_gimple_switch): Likewise.
4893 (verify_expr_location_1): Dereference tp once. Add (disabled)
4894 piece from verify_expr.
4895 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4897 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4899 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4901 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4903 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4904 (small_data_pattern): Likewise.
4905 (arc_rewrite_small_data): Likewise.
4906 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4907 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4908 (get_symbol_alignment): New function.
4909 (legitimate_small_data_address_p): Likewise.
4910 (legitimate_scaled_address): Update, call
4911 legitimate_small_data_address_p.
4912 (output_sdata): New static variable.
4913 (arc_print_operand): Update how we handle small data operands.
4914 (arc_print_operand_address): Likewise.
4915 (arc_legitimate_address_p): Update, use
4916 legitimate_small_data_address_p.
4917 (arc_rewrite_small_data_p): Remove.
4918 (arc_rewrite_small_data_1): Likewise.
4919 (arc_rewrite_small_data): Likewise.
4920 (small_data_pattern): Likewise.
4921 (compact_sda_memory_operand): Update to use
4922 legitimate_small_data_address_p and get_symbol_alignment.
4923 (prepare_move_operands): Don't rewite sdata pattern.
4924 (prepare_extend_operands): Remove.
4925 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
4927 (zero_extendqisi2): Likewise.
4928 (zero_extendhisi2): Likewise.
4929 (extendqihi2): Likewise.
4930 (extendqisi2): Likewise.
4931 (extendhisi2): Likewise.
4935 * config/arc/constraints.md (Usd): Change it to memory constraint.
4937 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4939 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
4940 as source of std instructions.
4941 * config/arc/arc.md (movsi_insn): Update pattern predicate to
4942 allow 6-bit constants as source for store instructions.
4943 (movdi_insn): Update instruction pattern to allow 6-bit constants
4944 as source for store instructions.
4946 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4948 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
4950 2018-04-30 Nathan Sidwell <nathan@acm.org>
4951 Sandra Loosemore <sandra@codesourcery.com>
4953 * dumpfile.c (dump_open): Allow '-' for stdout.
4954 * doc/invoke.texi (Developer Options): Document dump filename
4955 determination early. Document stdin/stdout selection.
4957 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
4959 Microblaze Target: PIC data text relative
4961 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
4962 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
4964 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
4965 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
4966 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
4967 New addressing mode for data-text relative position indepenedent code.
4968 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
4969 'ADDRESS_SYMBOLIC_TXT_REL'.
4970 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
4971 (microblaze_legitimate_pic_operand): Exclude function calls from
4972 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
4973 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
4975 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4976 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
4977 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
4978 for 'address + offset'.
4979 (microblaze_expand_prologue): Add new function prologue call for
4981 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
4982 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
4983 table in case of TARGET_PIC_DATA_TEXT_REL.
4984 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
4985 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
4986 Add new macros 'UNSPEC_TEXT',
4987 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
4988 + exclude function calls from 'UNSPEC_PLT' in case of data text
4990 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4991 new target hook for generating address diff vector tables in case of
4993 * doc/tm.texi : Regenerate.
4994 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
4995 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
4996 of addr diff vector generation.
4997 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
4998 target hook definition.
4999 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5000 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5001 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5002 option for data text relative.
5004 2018-04-30 Richard Biener <rguenther@suse.de>
5006 * tree-chrec.h (evolution_function_is_constant_p): Remove
5008 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5010 2018-04-30 Richard Biener <rguenther@suse.de>
5013 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5015 2018-04-30 Richard Biener <rguenther@suse.de>
5017 PR tree-optimization/28364
5018 PR tree-optimization/85275
5019 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5020 copying first exit test.
5022 2018-04-28 Mark Wielaard <mark@klomp.org>
5024 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5026 (dwarf_AT): Handle DW_AT_addr_base.
5027 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5029 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5032 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5033 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5034 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5035 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5037 2018-04-28 Richard Biener <rguenther@suse.de>
5039 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5040 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5041 to reflect use. Only add interesting stmts.
5043 2018-04-27 Martin Jambor <mjambor@suse.cz>
5046 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5047 the jump function allows for passing through aggregate values.
5049 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5051 * input.h (in_system_header_at): Convert from macro to inline
5053 (from_macro_expansion_at): Likewise.
5054 (from_macro_definition_at): Likewise.
5056 2018-04-27 Jeff Law <law@redhat.com>
5058 * config.gcc: Mark tile* targets as deprecated/obsolete.
5060 2018-04-27 Richard Biener <rguenther@suse.de>
5062 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5065 2018-04-27 Richard Biener <rguenther@suse.de>
5067 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5069 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5071 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5072 with Yd constraint. Set "preferred_for_speed" attribute from
5073 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5075 (*movdi_internal): Ditto.
5076 (movti_interunit splitters): Remove
5077 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5078 (movdi_interunit splitters): Ditto.
5079 * config/i386/constraints.md (Ye): Remove.
5080 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5082 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5085 * config/aarch64/constraints.md (Usg): Limit to 31.
5088 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5090 PR tree-optimization/85529
5091 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5092 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5093 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5094 zero extension or masking of the MSB bit.
5095 (optimize_range_tests): Add FIRST_BB argument, pass it through
5096 to optimize_range_tests_var_bound.
5097 (maybe_optimize_range_tests, reassociate_bb): Adjust
5098 optimize_range_tests callers.
5100 2018-04-26 Richard Biener <rguenther@suse.de>
5101 Jakub Jelinek <jakub@redhat.com>
5103 * cgraph.h (symbol_table): Just declare debug method here.
5104 * symtab.c (symbol_table::debug): Define.
5106 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5108 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5110 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5112 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5113 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5114 (*movdi_internal): Substitute Yi and Yj constraint with x
5115 and Ym and Yn constraint with y constraint. Update "isa"
5116 attribute and set "preferred_for_speed" attribute from
5117 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5118 (*movsi_internal): Ditto.
5119 (*movdf_internal): Ditto.
5120 (*movsf_internal): Ditto.
5121 (*zero_extendsidi2): Ditto.
5122 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5123 (sse2_loadld): Ditto.
5124 (*vec_extract<ssevecmodelower>_0): Ditto.
5125 (*vec_extractv4si_0_zext_sse4): Ditto.
5126 (vec_concatv2di): Ditto.
5127 (*vec_dup<mode>): Ditto.
5128 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5129 * config/i386/constraints.md (Yi): Remove.
5134 2018-04-26 Nathan Sidwell <nathan@acm.org>
5136 * dumpfile.c (dump_open): New.
5137 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5138 (dump_finish): Detect stdio/stderr by value not name.
5140 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5142 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5144 2018-04-26 Tom de Vries <tom@codesourcery.com>
5147 * config/nvptx/nvptx.c (verify_neutering_jumps)
5148 (verify_neutering_labels): New function
5149 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5151 2018-04-26 Tom de Vries <tom@codesourcery.com>
5154 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5155 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5158 2018-04-26 Richard Biener <rguenther@suse.de>
5159 Tom de Vries <tom@codesourcery.com>
5162 * lto-streamer-out.c (output_function): Fixup loops if required to match
5163 discovery done in the reader.
5165 2018-04-26 Richard Biener <rguenther@suse.de>
5167 PR tree-optimization/85116
5168 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5169 have a loop exit from the single latch predecessor. Remove
5170 case of header with just condition.
5171 (ch_base::copy_headers): Exclude infinite loops from any
5173 (pass_ch::execute): Record exits.
5175 2018-04-26 Richard Biener <rguenther@suse.de>
5177 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5178 prologue cost vector and pass it to vect_get_load_cost.
5179 (vect_get_peeling_costs_all_drs): Likewise.
5180 (vect_peeling_hash_get_lowest_cost): Likewise.
5181 (vect_enhance_data_refs_alignment): Likewise.
5183 2018-04-26 Richard Biener <rguenther@suse.de>
5186 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5187 checking of integer<->pointer conversions.
5188 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5189 sign-/zero-extending pointer types.
5190 (expand_omp_for_static_chunk): Likewise.
5192 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5193 Jean Lee <xiaoyur347@gmail.com>
5195 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5196 (TARGET_ASAN_SHADOW_OFFSET): Define.
5197 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5198 true for -fsanitize=address.
5200 2018-04-25 Mark Wielaard <mark@klomp.org>
5202 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5205 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5207 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5208 than "alu", remove explicit "memory" and "imm_disp" attributes.
5209 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5212 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5213 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5214 gen_lowpart_no_emit.
5216 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5219 * config/i386/i386.c (ix86_expand_builtin): Change memory
5220 operand to XI, extend p0 to Pmode.
5221 * config/i386/i386.md: Change unspec volatile and operand
5222 1 mode to XI, change operand 0 mode to P.
5224 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5226 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5227 GET_MODE_MASK before any checking.
5228 (nds32_can_use_bset_p): Likewise.
5229 (nds32_can_use_btgl_p): Likewise.
5231 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5233 * config/nds32/nds32-doubleword.md: New define_split pattern for
5234 illegal register number.
5236 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5238 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5240 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5242 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5244 2018-04-25 Richard Biener <rguenther@suse.de>
5246 * lto-streamer.h (LTO_major_version): Bump to 8.
5248 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5250 * BASE-VER: Set to 9.0.0.
5252 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5254 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5255 in __abskf2 and __powikf2.
5257 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5260 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5261 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5262 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5263 Use the above on operand 2. Reindent.
5264 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5266 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5269 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5271 * config/i386/i386.opt (mcet): Removed.
5272 * doc/install.texi: Remove -mcet documentation.
5273 * doc/invoke.texi: Likewise.
5275 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5278 * doc/install.texi: Remove -mcet from bootstrap-cet.
5280 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5283 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5284 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5288 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5289 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5290 containing a CONST_VECTOR.
5292 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5294 * doc/install.texi: Update newlib dependency for nvptx.
5296 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5299 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5300 instead of INTVAL when shifting x left.
5302 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5304 PR tree-optimization/85478
5305 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5306 vect_grouped_store_supported for single element vectors.
5308 2018-04-24 Richard Biener <rguenther@suse.de>
5311 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5312 load cost increase to the case of non-constant step.
5314 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5317 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5318 destination if any_malformed_asm.
5320 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5323 * expr.c (store_field): In the bitfield case, if the value comes from
5324 a function call and is returned in registers by means of a PARALLEL,
5325 do not change the mode of the temporary unless BLKmode and VOIDmode.
5327 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5329 PR rtl-optimization/85423
5330 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5331 dependencies to debug insns when the previous insn is non-debug.
5333 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5335 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5336 enums into a single definition.
5337 (fls): Fix predicates and printing.
5340 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5342 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5343 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5344 and short u6 immediate.
5345 (check_if_valid_sleep_operand): Remove.
5346 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5348 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5350 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5351 flag_always_save_lp condition.
5352 * config/nds32/nds32.opt (malways-save-lp): New option.
5354 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5356 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5357 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5358 * config/nds32/nds32.h
5359 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5360 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5362 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5364 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5365 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5367 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5368 Chung-Ju Wu <jasonwucj@gmail.com>
5370 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5372 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5373 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5375 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5377 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5379 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5381 * config/nds32/nds32-protos.h (nds32_data_alignment,
5382 nds32_local_alignment): Declare.
5383 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5384 nds32_local_alignment): New functions.
5385 (TARGET_CONSTANT_ALIGNMENT): Define.
5386 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5388 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5390 * config/nds32/nds32.c
5391 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5392 (TARGET_MODES_TIEABLE_P): Likewise.
5394 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5396 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5399 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5400 Chung-Ju Wu <jasonwucj@gmail.com>
5402 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5403 for unaligned access.
5404 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5406 * config/nds32/nds32-intrinsic.md: Likewise.
5407 * config/nds32/nds32_intrinsic.h: Likewise.
5408 * config/nds32/nds32.h (nds32_builtins): Likewise.
5409 * config/nds32/nds32.opt (munaligned-access): New option.
5410 * config/nds32/nds32.c (nds32_asm_file_start): Display
5411 flag_unaligned_access status.
5413 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5415 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5416 -mno-relax is present.
5417 * config/riscv/linux.h (LINK_SPEC): Ditto.
5419 2018-04-20 Martin Sebor <msebor@redhat.com>
5422 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5424 (gimple_fold_builtin_stxcpy_chk): Same.
5425 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5427 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5430 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5431 __powikf2 when long double is IEEE 128-bit.
5433 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5435 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5436 step to make sure stack always aligned.
5438 2018-04-20 Carl Love <cel@us.ibm.com>
5441 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5442 size check for arg0.
5444 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5445 Tom de Vries <tom@codesourcery.com>
5448 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5449 Emit insns for calls too.
5450 (nvptx_find_par): Always look for worker-level predecessor insn.
5451 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5453 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5454 (nvptx_process_pars): Propagate frames for calls.
5456 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5459 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5461 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5462 (ix86_handle_option): Don't handle OPT_mibt.
5463 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5465 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5467 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5468 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5469 (ix86_target_macros): Define __CET__ with flag_cf_protection
5470 for -fcf-protection.
5471 * config/i386/i386.c (isa2_opts): Remove -mibt.
5472 * config/i386/i386.h (TARGET_IBT): Removed.
5473 (TARGET_IBT_P): Likewise.
5474 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5475 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5476 * config/i386/i386.opt (mcet): Update help message.
5479 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5480 -mcet as an alias for -mshstk.
5482 2018-04-20 Richard Biener <rguenther@suse.de>
5485 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5486 complexity by forcing a single use of the multiply operand.
5488 2018-04-20 Martin Jambor <mjambor@suse.cz>
5491 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5492 recursion dependency to only apply to non-clones.
5494 2018-04-20 Martin Jambor <mjambor@suse.cz>
5497 * ipa-cp.c (create_specialized_node): Check that clones of
5498 self-recursive edges exist during IPA-CP.
5500 2018-04-19 Toon Moene <toon@moene.org>
5502 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5505 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5507 PR tree-optimization/85467
5508 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5509 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5510 VECTOR_CST element to type.
5512 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5515 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5516 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5517 (builtin_longjmp): Likewise.
5518 (save_stack_nonlocal): New pattern.
5519 (restore_stack_nonlocal): Likewise.
5521 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5524 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5525 Replace ASM_OUTPUT_LABEL with fprintf.
5527 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5530 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5531 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5532 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5533 define __IBT__ and __SHSTK__ for -fcf-protection.
5534 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5536 (ix86_trampoline_init): Likewise.
5537 (x86_output_mi_thunk): Likewise.
5538 (ix86_notrack_prefixed_insn_p): Likewise.
5539 (ix86_option_override_internal): Don't disallow -fcf-protection.
5540 * config/i386/i386.md (rdssp<mode>): Also enable for
5542 (incssp<mode>): Likewise.
5543 (nop_endbr): Likewise.
5544 * config/i386/i386.opt (mcet): Change help message to built-in
5548 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5549 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5550 enable CET built-in functions.
5552 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5554 * common/config/i386/i386-common.c
5555 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5556 OPTION_MASK_ISA_MOVDIRI_UNSET,
5557 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5558 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5559 * config.gcc (movdirintrin.h): New header.
5560 * config/i386/cpuid.h (bit_MOVDIRI,
5561 bit_MOVDIR64B): New bits.
5562 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5564 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5565 (VOID, PVOID, PCVOID)): New function types.
5566 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5567 __builtin_ia32_directstoreu_u64,
5568 __builtin_ia32_movdir64b): New builtins.
5569 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5570 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5572 (ix86_valid_target_attribute_inner_p): Ditto.
5573 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5574 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5575 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5576 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5577 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5578 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5579 (movdiri<mode>, movdir64b_<mode>): New.
5580 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5581 * config/i386/immintrin.h: Include movdirintrin.h.
5582 * config/i386/movdirintrin.h: New file.
5583 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5585 2018-04-19 Richard Biener <rguenther@suse.de>
5588 * cfg.c (clear_bb_flags): When loop state says we have
5589 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5591 2018-04-19 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/84737
5594 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5595 copying restrict info.
5596 (vect_setup_realignment): Use it.
5597 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5598 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5599 the first DR to all generated stores.
5600 (vectorizable_load): Likewise for loads.
5602 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5604 PR tree-optimization/85446
5605 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5606 the integral and pointer types to have the same precision.
5608 * doc/install.texi: Document --disable-cet being the default and
5611 2018-04-18 Martin Liska <mliska@suse.cz>
5613 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5616 2018-04-18 Martin Liska <mliska@suse.cz>
5619 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5622 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5623 arguments if they are comparable.
5625 2018-04-18 Martin Liska <mliska@suse.cz>
5628 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5631 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5634 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5637 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5638 ENDBR after calling __morestack.
5640 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5643 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5644 by using gcc_base_ver to generate a gcc_driver_version, and use
5645 it when generating GCC_DRIVER_NAME.
5646 * configure: Regenerate.
5648 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5651 * config.gcc: Obsolete powerpc*-*-*spe*.
5653 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5656 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5657 (stabstr_D): Change type of unum from unsigned int to
5658 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5661 2018-04-17 Jim Wilson <jimw@sifive.com>
5664 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5665 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5666 Set arg_pointer_offset after using pretend_args_size.
5668 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5670 PR rtl-optimization/85431
5671 * dse.c (record_store): Ignore zero width stores.
5674 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5675 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5676 __builtin_stack_restore rather than after it.
5677 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5678 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5679 argument instead of virtual_dynamic_stack_rtx.
5681 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5683 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5685 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5686 Add note to error message to explain internal mapping of overloaded
5687 built-in function name to non-overloaded built-in function name.
5688 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5691 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5694 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5695 where the inputs overlap with the output.
5697 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5699 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5700 (=v, v) alternative and explicit "memory" attribute.
5701 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5702 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5704 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5705 "sselog1" type instead of "sselog".
5706 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5707 "sselog". Remove explicit "memory" attribute.
5708 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5709 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5711 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5712 "sselog1" type instead of "sselog". Remove explicit "memory"
5714 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5715 use "sselog1" type instead of "sselog". Remove explicit "memory"
5717 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5718 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5720 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5721 "sselog1" type instead of "sselog". Remove explicit "memory"
5723 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5724 use "sselog1" type instead of "sselog". Remove explicit "memory"
5728 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5731 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5734 2018-04-17 Martin Jambor <mjambor@suse.cz>
5737 * ipa-cp.c (create_specialized_node): Call
5738 expand_all_artificial_thunks if necessary.
5740 2018-04-17 Martin Liska <mliska@suse.cz>
5743 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5744 in message, remote space in between '_G' and '('.
5746 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5749 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5750 avx512f_vmcmp<mode>3<round_saeonly_name>,
5751 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5752 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5753 avx512f_rndscale<mode><round_saeonly_name>,
5754 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5755 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5756 Use %<iptr>2 instead of %2 for -masm=intel.
5757 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5758 avx512f_vcvttss2usi<round_saeonly_name>,
5759 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5761 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5762 avx512f_vcvttsd2usi<round_saeonly_name>,
5763 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5764 Use %q1 instead of %1 for -masm=intel.
5765 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5766 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5767 of %3 for -masm=intel.
5768 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5770 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5772 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5774 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5775 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5777 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5779 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5780 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5781 %g1 and one with %0 and %1.
5782 (avx512er_vmrcp28<mode><round_saeonly_name>,
5783 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5785 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5786 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5787 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5788 of %0 and %{%4%} for -masm=intel.
5789 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5790 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5791 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5792 order of %0 and %{%5%}%{z%} for -masm=intel.
5794 2018-04-17 Jan Hubicka <jh@suse.cz>
5797 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5799 2018-04-17 Martin Liska <mliska@suse.cz>
5802 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5803 for target_clone error message. Make default implementation
5804 clone to be a local declaration.
5805 (separate_attrs): Add new argument and check for an empty
5807 (expand_target_clones): Handle it.
5808 (ipa_target_clone): Make redirection just for target_clones
5811 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5812 Tom de Vries <tom@codesourcery.com>
5815 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5816 tiled basic blocks without omp continue statements.
5818 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5821 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5822 vec_extract expression as having side effects to make sure it gets
5825 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5828 * config/i386/i386.c (get_builtin_code_for_version): Check
5831 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5834 * gcc/config.gcc: Support "skylake".
5835 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5837 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5838 (processor_target_table): Add "skylake".
5839 (ix86_option_override_internal): Add "skylake".
5840 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5841 PROCESSOR_CANNONLAKE.
5842 (get_builtin_code_for_version): Fix priority for
5843 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5844 PROCESSOR_SKYLAKE-AVX512.
5845 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5846 (processor_type): Add PROCESSOR_SKYLAKE.
5848 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5849 Jason Merrill <jason@redhat.com>
5852 * convert.c (convert_to_integer_1): Use direct recursion for
5853 enumeral types and types with a precision less than the number
5854 of bits in their mode.
5856 2018-04-16 Julia Koval <julia.koval@intel.com>
5859 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5860 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5862 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5865 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5866 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5867 and -mno-direct-move.
5869 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5872 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5873 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5874 Ensure negative shifts result in {0}.
5876 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5878 PR rtl-optimization/79916
5879 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5880 regs (if any) to define how to gnerate SD moves when LRA is in
5883 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5885 PR rtl-optimization/85393
5886 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5887 * except.c (expand_dw2_landing_pad_for_region): Make static.
5888 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5889 a label and unconditional jump to old_bb, rather than
5890 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5893 PR rtl-optimization/85376
5894 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5895 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5896 instead of a specific value.
5898 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5899 Bin Cheng <bin.cheng@arm.com>
5901 PR tree-optimization/82965
5902 PR tree-optimization/83991
5903 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5904 by_profile_only parameter.
5905 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5906 information if the loop was predicted to iterate too many times.
5907 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5909 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5912 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5915 2018-04-13 Martin Liska <mliska@suse.cz>
5916 Jakub Jelinek <jakub@redhat.com>
5919 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5920 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5921 * builtins.c (expand_builtin_memory_copy_args): Use
5922 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5923 handle dest_addr == pc_rtx.
5925 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
5928 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
5929 asked to not generate direct moves.
5930 (fix_trunc<mode>si2_stfiwx): Similar.
5931 (fix_trunc<mode>si2_internal): Similar.
5933 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5936 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
5937 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
5938 lookup if dest in some wider mode is known to be const0_rtx and
5939 if so, record permanent equivalence for it to be ZERO_EXTEND of
5940 the narrower mode destination.
5942 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5944 * lto-streamer-out.c (output_function): Revert 259346.
5945 * omp-expand.c (expand_oacc_for): Likewise.
5947 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
5949 PR rtl-optimization/85354
5950 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
5951 * sel-sched.c (sel_global_init): ... here.
5953 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
5956 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
5957 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5958 mode for PE-COFF targets.
5959 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
5960 (i386_pe_asm_lto_end): Likewise.
5961 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
5962 (TARGET_ASM_LTO_END): Likewise.
5963 * config/i386/winnt.c (saved_debug_info_level): New static variable.
5964 (i386_pe_asm_lto_start): New function.
5965 (i386_pe_asm_lto_end): Likewise.
5967 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
5968 Richard Biener <rguenther@suse.de>
5971 * lto-streamer-out.c (output_function): Fix CFG loop state before
5973 * omp-expand.c (expand_oacc_for): Handle calls to internal
5974 functions like regular functions.
5976 2018-04-12 Richard Biener <rguenther@suse.de>
5979 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
5980 for the early LTO debug to properly generate references to it
5981 during DIE emission. Do not re-use that for the skeleton for
5983 (dwarf2out_early_finish): Likewise.
5985 2018-04-12 Jakub Jelinek <jakub@redhat.com>
5988 * config/i386/sse.md
5989 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
5990 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
5991 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
5992 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
5993 and output is a reg, avoid creating invalid lowpart subreg, but
5994 instead split into a 512-bit move. Don't split if not AVX512VL,
5995 input is xmm16+ reg and output is a mem.
5996 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
5997 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
5998 xmm16+ reg and output is a mem.
6000 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6002 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6003 also for flag_dwarf2_cfi_asm.
6005 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6007 PR rtl-optimization/85342
6008 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6009 a bool scalar var inside of the loop instead. Don't try to update
6010 recog_data.operand after failed apply_change_group.
6012 2018-04-12 Tom de Vries <tom@codesourcery.com>
6015 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6016 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6017 array with flexible array member as array without given dimension.
6018 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6019 argument for undefined param to true.
6021 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6024 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6025 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6026 from PowerPC section.
6027 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6028 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6029 help text that is too long.
6030 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6031 help text that is too long.
6032 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6033 help text that is too long.
6035 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6037 * config/alpha/alpha.md (stack_probe_internal): Rename
6038 from "probe_stack". Update all callers.
6040 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6042 PR rtl-optimization/84566
6043 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6044 sched_macro_fuse_insns.
6046 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6049 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6050 (compute_block_dependences): ... from here.
6052 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6054 PR tree-optimization/85331
6055 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6056 from int to HOST_WIDE_INT.
6058 2018-04-11 Martin Jambor <mjambor@suse.cz>
6061 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6062 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6063 not the same as the source val.
6064 (cgraph_edge_brings_value_p): New parameter.
6065 (gather_edges_for_value): Pass destination value to
6066 cgraph_edge_brings_value_p.
6067 (perhaps_add_new_callers): Likewise.
6068 (get_info_about_necessary_edges): Likewise and exclude values brought
6069 only by self-recursive edges.
6070 (create_specialized_node): Redirect only clones of self-calling edges.
6071 (+self_recursive_pass_through_p): New function.
6072 (find_more_scalar_values_for_callers_subset): Use it.
6073 (find_aggregate_values_for_callers_subset): Likewise.
6074 (known_aggs_to_agg_replacement_list): Removed.
6075 (decide_whether_version_node): Re-calculate known constants for all
6076 remaining context clones.
6078 2018-04-11 Richard Biener <rguenther@suse.de>
6081 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6082 from early DWARF output.
6083 (dwarf2out_early_finish): Output line info unconditionally into
6084 early DWARF and add reference to it.
6086 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6089 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6090 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6091 other than V2DFmode using iptr mode attribute.
6092 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6094 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6096 PR rtl-optimization/84659
6097 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6099 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6102 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6104 (output_loc_list): Pass address of a dummy size variable even in the
6105 locview handling loop.
6106 (index_location_lists): Add comment on why skip_loc_list_entry can't
6109 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6112 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6115 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6118 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6119 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6120 and -mstring-compare-inline-limit.
6122 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6125 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6126 for stack clash protection in a register whenever we need it to be in
6129 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6131 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6132 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6134 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6137 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6139 (mlong-double-): Ditto.
6140 * config/rs6000/sysv4.opt (msdata=): Ditto.
6141 (mtls-size=): Ditto.
6143 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6145 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6146 erroneous entries for
6147 "vector int vec_ldl (int, long int *)", and
6148 "vector unsigned int vec_ldl (int, unsigned long int *)".
6149 Add comments and entries for
6150 "vector bool char vec_ldl (int, bool char *)",
6151 "vector bool short vec_ldl (int, bool short *)",
6152 "vector bool int vec_ldl (int, bool int *)",
6153 "vector bool long long vec_ldl (int, bool long long *)",
6154 "vector pixel vec_ldl (int, pixel *)",
6155 "vector long long vec_ldl (int, long long *)",
6156 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6157 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6158 type tree bool_long_long_type_node and correct definition of
6159 bool_V2DI_type_node to make reference to this new type tree.
6160 (rs6000_mangle_type): Replace erroneous reference to
6161 bool_long_type_node with bool_long_long_type_node.
6162 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6163 comments to emphasize sign distinctions for char and int types and
6164 replace RS6000_BTI_bool_long constant with
6165 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6166 use of RS6000_BTI_pixel.
6167 (bool_long_type_node): Remove this macro definition.
6168 (bool_long_long_type_node): New macro definition
6170 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6172 PR rtl-optimization/85300
6173 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6174 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6175 simplify_unary_operation fails.
6177 2018-04-10 Martin Liska <mliska@suse.cz>
6179 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6180 cgraph_edge and ipa_ref.
6182 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6186 * config/i386/sse.md
6187 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6188 computation of the VEC_MERGE selector from mask.
6189 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6190 Fix decoding of the VEC_MERGE selector into mask.
6192 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6194 PR tree-optimization/85286
6195 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6197 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6199 * final.c (final_1): Set insn_last_address as well as
6200 insn_current_address.
6202 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6205 * explow.c (emit_stack_probe): Call validize_mem on memory location
6206 before passing it to gen_probe_stack. Create address operand and
6207 legitimize it for the probe_stack_address case.
6209 2018-04-09 Jan Hubicka <jh@suse.cz>
6212 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6213 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6214 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6216 (free_lang_data): Rebuild type inheritance graph.
6218 2018-04-09 Martin Sebor <msebor@redhat.com>
6220 * invoke.texi (-finline-small-functions): Mention other optimization
6222 (-findirect-inlining, -fpartial-inlining): Same.
6223 (-finline-functions-called-once): Same.
6224 (-freorder-blocks-and-partition): Same.
6226 2018-04-09 Jan Hubicka <jh@suse.cz>
6229 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6230 jumps; choose last target that matches the criteria (i.e.
6231 no partition changes for non-crossing jumps).
6232 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6233 support for redirecting crossing jumps to non-crossing.
6235 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6237 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6238 also for naked functions.
6240 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6242 * config/arc/arc.md (add_shift): New pattern.
6243 (add_shift2): Likewise.
6244 (sub_shift): Likewise.
6245 (sub_shift_cmp0_noout): Likewise.
6246 (compare_si_ashiftsi): Likewise.
6247 (xbfu_cmp0_noout): New combine pattern.
6248 (xbfu_cmp0"): Likewise.
6249 (movsi_set_cc_insn): Place the predicable variant first.
6250 (commutative_binary_cmp0_noout): Remove clobber.
6251 (commutative_binary_cmp0): New pattern.
6252 (noncommutative_binary_cmp0): Likewise.
6253 (noncommutative_binary_cmp0_noout): Likewise.
6254 (noncommutative_binary_comparison_result_used): Removed.
6255 (rsub_cmp0): New pattern.
6256 (rsub_cmp0_noout): Likewise.
6257 (extzvsi): Changed, keep only meaningful variants.
6258 (SQH, SEZ): New iterators.
6259 (SQH_postfix): New mode attribute.
6260 (SEZ_prefix): New code attribute.
6261 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6262 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6263 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6265 (noncommutative_operator): Check the availability of barrel
6268 2018-04-09 Richard Biener <rguenther@suse.de>
6270 PR tree-optimization/85284
6271 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6272 Only use the niter constraining form of simple_iv when the exit
6275 2018-04-09 Tom de Vries <tom@codesourcery.com>
6278 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6279 (define_expand "*memory_barrier"): New define_expand.
6280 (define_insn "memory_barrier"): New insn.
6282 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6284 PR rtl-optimization/80463
6285 PR rtl-optimization/83972
6286 PR rtl-optimization/83480
6288 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6289 correct producer for the insn.
6290 (tidy_control_flow): Fixup seqnos in case of debug insns.
6292 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6294 PR rtl-optimization/83913
6296 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6297 different sched-times when merging exprs.
6299 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6301 PR rtl-optimization/83962
6303 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6304 tidy_fallthru_edge and tidy_control_flow.
6306 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6308 PR rtl-optimization/83530
6310 * sel-sched.c (force_next_insn): New global variable.
6311 (remove_insn_for_debug): When force_next_insn is true, also leave only
6312 next insn in the ready list.
6313 (sel_sched_region): When the region wasn't scheduled, make another pass
6314 over it with force_next_insn set to 1.
6316 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6318 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6320 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6321 for interrupt control.
6322 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6323 functions for interrupt control.
6324 * config/nds32/nds32-intrinsic.md: Likewise.
6325 * config/nds32/nds32_intrinsic.h: Likewise.
6326 * config/nds32/nds32.h (nds32_builtins): Likewise.
6328 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6330 * config/nds32/nds32.c (nds32_init_machine_status,
6331 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6332 strict_aligned_p field.
6333 (nds32_expand_to_rtl_hook): New function.
6334 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6335 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6337 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6338 Chung-Ju Wu <jasonwucj@gmail.com>
6340 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6341 * config/nds32/nds32-n7.md: New file.
6342 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6343 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6345 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6346 * config/nds32/nds32.md (pipeline_model): Add n7.
6347 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6348 * config/nds32/pipelines.md: Include n7 settings.
6350 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6351 Chung-Ju Wu <jasonwucj@gmail.com>
6353 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6354 * config/nds32/nds32-e8.md: New file.
6355 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6356 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6358 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6359 * config/nds32/nds32.md (pipeline_model): Add e8.
6360 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6361 * config/nds32/pipelines.md: Include e8 settings.
6363 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6364 Chung-Ju Wu <jasonwucj@gmail.com>
6366 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6367 * config/nds32/nds32-n8.md: New file.
6368 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6369 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6371 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6372 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6373 * config/nds32/nds32.md (pipeline_model): Add n8.
6374 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6375 * config/nds32/pipelines.md: Include n8 settings.
6377 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6378 Chung-Ju Wu <jasonwucj@gmail.com>
6380 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6381 * config/nds32/nds32-n9-2r1w.md: New file.
6382 * config/nds32/nds32-n9-3r2w.md: New file.
6383 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6384 nds32_register_ports): New or modify for cpu n9.
6385 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6387 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6388 * config/nds32/nds32-utils.c: New file.
6389 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6390 TARGET_MUL_SLOW): Define.
6391 * config/nds32/nds32.md (pipeline_model): New attribute.
6392 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6393 New options that support cpu n9.
6394 * config/nds32/pipelines.md: Include n9 settings.
6395 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6397 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6399 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6400 information if necessary.
6401 (output_cond_branch_compare_zero): Likewise.
6402 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6403 (nds32_target_alignment): Refine for alignment.
6404 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6405 (FUNCTION_BOUNDARY): Modify.
6406 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6408 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6410 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6412 * config/nds32/constants.md (unspec_volatile_element): Add values for
6413 TLB operation and data prefetch.
6414 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6415 functions for TLB operation and data prefetch.
6416 * config/nds32/nds32-intrinsic.md: Likewise.
6417 * config/nds32/nds32_intrinsic.h: Likewise.
6418 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6419 (nds32_print_operand): Likewise.
6420 * config/nds32/nds32.h (nds32_builtins): Likewise.
6422 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6423 Andrew Pinski <pinsika@gcc.gnu.org>
6426 * match.pd: Use constant_boolean_node of correct type instead of
6427 boolean_true_node or boolean_false_node for simplifying
6428 pointer comparisons to zero.
6430 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6432 PR tree-optimization/80021
6433 * tree.c (verify_type_variant): Make error call in verify_variant_match
6434 translatable and remove final full stop.
6436 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6438 * config/nds32/constants.md (unspec_volatile_element): Add
6439 UNSPEC_VOLATILE_EH_RETURN.
6440 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6441 nds32_output_stack_pop): Support dwarf exception handling process.
6442 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6443 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6444 exception handling process.
6445 (nds32_compute_stack_frame): Likewise.
6446 (nds32_return_addr_rtx): Likewise.
6447 (nds32_initial_elimination_offset): Likewise.
6448 (nds32_expand_prologue): Likewise.
6449 (nds32_expand_epilogue): Likewise.
6450 (nds32_dynamic_chain_address): New function.
6451 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6453 (DYNAMIC_CHAIN_ADDRESS): Define.
6454 (EH_RETURN_DATA_REGNO): Define.
6455 (EH_RETURN_STACKADJ_RTX): Define.
6456 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6457 patterns for dwarf exception handling.
6459 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6461 * config/nds32/nds32.h: Clean up obsolete macros.
6463 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6465 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6466 Add enum values for particular instructions.
6467 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6468 particular intrinsic functions.
6469 * config/nds32/nds32-intrinsic.md: Likewise.
6470 * config/nds32/nds32_intrinsic.h: Likewise.
6471 * config/nds32/nds32.h (nds32_builtins): Likewise.
6472 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6473 (btst, ave): New patterns for particular instructions.
6475 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6477 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6478 Add enum values for atomic load/store and memory sync.
6479 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6481 * config/nds32/nds32-intrinsic.md: Likewise.
6482 * config/nds32/nds32_intrinsic.h: Likewise.
6483 * config/nds32/nds32.h (nds32_builtins): Likewise.
6485 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6487 PR tree-optimization/85257
6488 * fold-const.c (native_encode_vector): If not all elts could fit
6489 and off is -1, return 0 rather than offset.
6490 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6491 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6492 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6493 adjust buffer in native_interpret_expr call.
6495 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6497 * config/nds32/constants.md (unspec_volatile_element): Add cache
6498 control enum values.
6499 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6500 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6501 * config/nds32/nds32.c (nds32_cctl_names): New.
6502 (nds32_print_operand): Handle cache control register names.
6503 * config/nds32/nds32.h (nds32_builtins): New enum values.
6504 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6506 * config/nds32/nds32.md (type): Add mmu.
6507 * config/nds32/pipelines.md (simple_insn): Add mmu.
6509 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6511 * config/nds32/nds32.md (type): Remove call.
6512 * config/nds32/pipelines.md (simple_insn): Likewise.
6514 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6516 * config/nds32/constants.md (unspec_volatile_element): Add
6517 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6518 UNSPEC_VOLATILE_FMFCFG.
6519 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6520 description for fmfcfg and fmfcsr.
6521 (bdesc_1arg): Add fmtcsr.
6522 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6523 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6524 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6525 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6526 unspec_fmfcfg): New patterns.
6527 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6528 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6529 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6530 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6531 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6532 __nds32__fmfcfg): Define.
6534 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6536 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6537 intrinsic register names.
6538 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6539 intrinsic register enum values and macros.
6541 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6543 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6544 for load/store addressing form.
6545 (nds32_print_operand_address): Likewise.
6547 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6550 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6551 based on LABEL_REF. Remove useless assertion.
6552 (pic_address_needs_scratch): Fix formatting.
6553 (sparc_legitimize_pic_address): Minor tweaks.
6554 (sparc_delegitimize_address): Adjust assertion accordingly.
6555 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6556 into symbolic_operand.
6557 (movsi_high_pic_label_ref): Likewise.
6558 (movsi_lo_sum_pic_label_ref): Likewise.
6559 (movdi_pic_label_ref): Likewise.
6560 (movdi_high_pic_label_ref): Likewise.
6561 (movdi_lo_sum_pic_label_ref): Likewise.
6563 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6565 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6566 custom LIB_SPEC setup.
6568 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6569 Kito Cheng <kito.cheng@gmail.com>
6571 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6572 * config/riscv/freebsd.h: New.
6574 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6576 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6577 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6580 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6581 Kito Cheng <kito.cheng@gmail.com>
6583 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6584 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6585 * config/nds32/nds32-protos.h (nds32_output_call,
6586 nds32_output_return): Declare.
6587 * config/nds32/nds32.md: Refine all the call and return patterns.
6589 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6592 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6593 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6595 PR rtl-optimization/84872
6596 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6597 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6600 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6602 * expr.c (copy_blkmode_to_reg): Revert 254862.
6603 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6605 2018-04-06 Richard Biener <rguenther@suse.de>
6608 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6609 after seeing a component reference with an adjacent field. Treat
6610 refs to arrays at struct end of external decls similar to
6611 refs to unconstrained commons.
6613 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6616 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6617 look through SAVE_EXPRs with non-side-effects argument. Adjust
6619 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6622 2018-04-06 Richard Biener <rguenther@suse.de>
6625 * alias.c (find_base_term): New wrapper around find_base_term
6626 unwinding CSELIB_VAL_PTR changes.
6627 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6630 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6632 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6634 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6635 constant definitions.
6636 ("nop"): lr 0,0 -> nopr r0
6637 ("nop_lr0", "nop_lr1"): New insn definitions.
6639 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6641 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6642 NDS32_V3PUSH_AVAILABLE_P macro.
6644 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6645 Chung-Ju Wu <jasonwucj@gmail.com>
6647 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6648 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6649 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6650 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6651 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6652 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6653 * config/nds32/constraints.md: New constraints and checking for hard
6654 float configuration.
6655 * config/nds32/iterators.md: New mode iterator and attribute for hard
6656 float configuration.
6657 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6659 * config/nds32/nds32-fpu.md: New file.
6660 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6661 deal with hard float code generation.
6662 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6664 (abi_type, float_reg_number): New enum type.
6665 * config/nds32/nds32-predicates.c: New predicates for hard float.
6666 * config/nds32/nds32-protos.h: Declare functions for hard float.
6667 * config/nds32/nds32.c: Implementation for hard float configuration.
6668 * config/nds32/nds32.h: Definitions for hard float configuration.
6669 * config/nds32/nds32.md: Include hard float machine description and
6670 modify patterns for hard float configuration.
6671 * config/nds32/nds32.opt: New options for hard float configuration.
6672 * config/nds32/predicates.md: New predicates for hard float
6675 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6677 * common/config/nds32/nds32-common.c
6678 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6680 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6683 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6684 CONSTRUCTOR_ELT (ctor, ...)->value.
6686 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6689 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6691 2018-04-05 Tom de Vries <tom@codesourcery.com>
6694 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6697 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6698 Kito Cheng <kito.cheng@gmail.com>
6700 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6701 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6702 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6703 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6705 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6706 Kito Cheng <kito.cheng@gmail.com>
6708 * config/nds32/constraints.md (Ufe): New memory constraint.
6709 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6710 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6711 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6713 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6714 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6716 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6718 * config/nds32/nds32.md: Use optimize_size in the condition for
6719 alu-shift instructions.
6721 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6723 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6725 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6727 * config/nds32/nds32.md (negsi2): Refine pattern.
6729 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6730 Chung-Ju Wu <jasonwucj@gmail.com>
6732 * config/nds32/iterators.md (shift_rotate): New code iterator.
6733 (shift): New code attribute.
6734 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6735 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6736 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6737 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6738 bit-wise operations.
6739 (andsi3, *andsi3): Ditto.
6740 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6741 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6742 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6743 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6744 nds32_ior_operand, nds32_xor_operand): New predicates.
6746 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6748 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6749 (addsi3, subsi3): ... this.
6751 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6753 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6755 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6757 * config/nds32/nds32.md: Adjust indention.
6759 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6761 * config/nds32/nds32.md (feature): New attribute.
6763 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6765 * config/nds32/nds32.md (subtype): New attribute.
6767 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6770 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6771 expansion to perform a bitwise AND of the argument followed by a
6772 boolean negation of the result.
6774 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6776 PR rtl-optimization/84878
6777 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6778 the basic block. Assert the use reference is not artificial and that
6779 it has an associated insn.
6781 2018-04-04 Michael Matz <matz@suse.de>
6783 * builtins.c (compute_objsize): Pass correct operand
6784 to array_at_struct_end_p.
6786 2018-04-04 Richard Biener <rguenther@suse.de>
6789 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6790 from contexts for DINFO_LEVEL_TERSE and below.
6792 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6794 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6795 resiter_operand condition.
6796 * config/nds32/nds32.md (*move<mode>): Ditto.
6798 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6799 Monk Chiang <sh.chiang04@gmail.com>
6801 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6803 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6805 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6807 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6808 Kito Cheng <kito.cheng@gmail.com>
6810 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6811 nds32_cond_code_str, output_cond_branch,
6812 output_cond_branch_compare_zero, nds32_expand_cbranch,
6813 nds32_expand_cstore, nds32_expand_movcc,
6814 nds32_output_cbranchsi4_equality_zero,
6815 nds32_output_cbranchsi4_equality_reg,
6816 nds32_output_cbranchsi4_equality_reg_or_const_int,
6817 nds32_output_cbranchsi4_greater_less_zero: New functions.
6818 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6819 nds32_expand_cstore, nds32_expand_movcc,
6820 nds32_output_cbranchsi4_equality_zero,
6821 nds32_output_cbranchsi4_equality_reg,
6822 nds32_output_cbranchsi4_equality_reg_or_const_int,
6823 nds32_output_cbranchsi4_greater_less_zero): Declare.
6824 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6825 nds32_rimm11s_operand): New predicates.
6826 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6827 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6830 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6832 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6833 * config/nds32/nds32.md: Ditto.
6834 * config/nds32/pipelines.md: Ditto.
6836 2018-04-04 Richard Biener <rguenther@suse.de>
6838 PR tree-optimization/85168
6839 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6840 propagating abnormals.
6842 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6844 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6846 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6847 Kito Cheng <kito.cheng@gmail.com>
6849 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6850 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6851 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6852 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6853 * config/nds32/nds32.md (sibcall_internal): New.
6854 (sibcall_register): Remove.
6855 (sibcall_immediate): Remove.
6856 (sibcall_value_internal): New.
6857 (sibcall_value_register): Remove.
6858 (sibcall_value_immediate): Remove.
6859 * config/nds32/predicates.md (nds32_general_register_operand): New.
6860 (nds32_call_address_operand): New.
6862 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6864 PR rtl-optimization/85167
6865 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6866 bb_defs if *split_p, instead preinitialize it to NULL.
6868 PR tree-optimization/85156
6869 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6870 evaluating the argument multiple times.
6872 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6874 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6876 (_mm_cvtpd_ps): Likewise.
6877 (_mm_cvttpd_epi32): Likewise.
6878 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6879 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6880 vector, pixel, and bool following altivec.h include.
6882 2018-04-03 Martin Sebor <msebor@redhat.com>
6884 * doc/extend.texi (Common Function Attributes): Clarify.
6885 (const attribute): Likewise.
6886 (pure attribute): Likewise.
6888 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6891 * config/i386/i386.c (ix86_expand_vector_set): Use
6892 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6894 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6896 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6897 instructions when changing rounding bits to preserve precision bits
6898 in the x87 control word.
6900 2018-04-03 Martin Liska <mliska@suse.cz>
6902 PR tree-optimization/82491
6903 * rtl.h (strip_offset_and_add): Replace += suboffset with
6904 poly_uint64 () + suboffset.
6906 2018-03-29 Martin Liska <mliska@suse.cz>
6907 Martin Jambor <mjambor@suse.cz>
6910 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6911 param_type is not an integral or pointer type.
6913 2018-04-03 Richard Biener <rguenther@suse.de>
6915 * sese.h (recompute_all_dominators): Remove.
6917 2018-04-02 Martin Sebor <msebor@redhat.com>
6919 * doc/invoke.texi (-Wrestrict): Fix typos.
6921 2018-04-02 Jim Wilson <jimw@sifive.com>
6923 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6924 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6925 (<optab>di3, <optab>si3_extend): Likewise.
6926 (<optab>si3_mask, <optab>si3_mask_1): New.
6927 (<optab>di3_mask, <optab>di3_mask_1): New.
6928 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
6929 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
6930 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
6932 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
6934 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
6937 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
6939 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
6940 (nds32_canonicalize_comparison): New function.
6942 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6943 Kito Cheng <kito.cheng@gmail.com>
6944 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6946 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
6947 * config/nds32/constants.md (unspec_volatile_element): Add
6948 UNSPEC_VOLATILE_RELAX_GROUP.
6949 * config/nds32/nds32-relax-opt.c: New file.
6950 * config/nds32/nds32-predicates.c
6951 (nds32_symbol_load_store_p): New function.
6952 * config/nds32/nds32-protos.h
6953 (nds32_symbol_load_store_p): Declare function.
6954 (make_pass_nds32_relax_opt): Declare new rtl pass function.
6955 * config/nds32/nds32.c
6956 (nds32_register_pass): New function to register pass.
6957 (nds32_register_passes): New function to register passes.
6958 * config/nds32/nds32.md (relax_group): New pattern.
6959 * config/nds32/nds32.opt (mrelax-hint): New option.
6960 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
6962 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
6964 * config/nds32/t-nds32: Modify files dependency.
6966 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6968 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
6969 (PROFILE_HOOK): Define its implementation.
6971 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
6973 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
6974 type and 32-bit size.
6976 2018-04-01 Jakub Jelinek <jakub@redhat.com>
6979 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
6980 (V_128_256): New mode iterator.
6981 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
6982 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
6983 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
6985 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
6986 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
6988 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
6991 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
6992 NaN inputs correctly.
6994 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
6997 * config/rs6000/vsx.md (??r): New mode attribute.
6998 (*vsx_mov<mode>_64bit): Use it.
6999 (*vsx_mov<mode>_32bit): Likewise.
7001 2018-03-30 Martin Sebor <msebor@redhat.com>
7003 PR tree-optimization/84818
7004 * builtins.c (check_access): Use warning_n.
7006 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7009 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7011 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7014 2018-03-30 Julia Koval <julia.koval@intel.com>
7017 * x86-tune.def (movx, partial_reg_dependency): Enable for
7020 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7023 * lra-constraints.c (process_alt_operands): Move setting
7024 this_alternative_matches below.
7026 2018-03-29 Martin Liska <mliska@suse.cz>
7029 * doc/invoke.texi: Document how LTO works with debug info.
7030 Describe auto-load support of binutils. Mention 'x86-64'
7031 as valid option value of -march option.
7033 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7035 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7038 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7039 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7040 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7043 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7046 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7047 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7048 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7049 for RS6000_BTM_POWERPC64.
7050 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7051 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7052 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7057 2018-03-28 Carl Love <cel@us.ibm.com>
7060 2017-09-27 Carl Love <cel@us.ibm.com>
7062 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7063 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7064 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7067 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7069 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7070 instead of __vector bool.
7071 (_mm_max_pu8): Likewise.
7072 (_mm_min_pi16): Likewise.
7074 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7077 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7078 (DIVWEUO): Likewise.
7080 (DIVDEUO): Likewise.
7081 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7082 DIVWEUO and DIVDEUO.
7083 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7084 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7085 (div_extend): Likewise.
7086 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7088 (__builtin_divweuo): Likewise.
7089 (__builtin_divdeo): Likewise.
7090 (__builtin_divdeuo): Likewise.
7092 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7095 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7096 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7098 PR tree-optimization/82004
7099 * gimple-match-head.c (optimize_pow_to_exp): New function.
7100 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7101 Don't fold to exp if optimize_pow_to_exp is false.
7103 2018-03-28 Martin Liska <mliska@suse.cz>
7106 * calls.c (initialize_argument_information): Fix trailing space.
7107 * common.opt: Fix typo and provide better explanation for
7108 -fsanitize-coverage option.
7109 * config/i386/i386.opt: Fix typo.
7111 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7112 Martin Liska <mliska@suse.cz>
7115 * gimplify.c (asan_poison_variable): Don't do the check for
7116 gimplify_omp_ctxp here.
7117 (gimplify_decl_expr): Do it here.
7118 (gimplify_target_expr): Likewise.
7120 2018-03-28 Martin Liska <mliska@suse.cz>
7123 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7124 chkp_type_bounds_count if MPX is not enabled.
7126 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7128 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7130 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7133 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7134 function to create the function decl for complex long double
7135 multiply and divide for -mabi=ieeelongdouble.
7136 (init_float128_ieee): Call it.
7138 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7141 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7142 -fcf-protection=branch -mibt.
7143 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7145 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7148 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7149 arm_word_relocations.
7151 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7154 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7155 extern array declarations.
7157 2018-03-27 Richard Biener <rguenther@suse.de>
7160 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7161 explicit single_use checks.
7163 2018-03-27 Richard Biener <rguenther@suse.de>
7165 PR tree-optimization/85082
7166 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7169 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7171 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7172 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7173 Turn on fasynchronous-unwind-tables and funwind-tables.
7175 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7178 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7179 (*bmi_blsr_<mode>_ccz): Ditto.
7181 2018-03-26 Tom de Vries <tom@codesourcery.com>
7183 PR tree-optimization/85063
7184 * omp-general.c (offloading_function_p): New function. Factor out
7186 * omp-offload.c (pass_omp_target_link::gate): ... here.
7187 * omp-general.h (offloading_function_p): Declare.
7188 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7189 with attribute omp declare target for offloading functions.
7191 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7193 PR tree-optimization/84005
7194 * tree-data-ref.h (get_base_for_alignment): Declare.
7195 * tree-data-ref.c (get_base_for_alignment_1): New function.
7196 (get_base_for_alignment): Likewise.
7197 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7198 get_base_for_alignment to find a suitable base object, instead
7199 of always using drb->base_address.
7201 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7204 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7205 known size by default.
7207 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7210 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7211 and non BLKmode operands.
7213 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7216 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7217 Clean up attributes.
7219 2018-03-23 Richard Biener <rguenther@suse.de>
7222 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7223 we are going to emit early debug for LTO.
7225 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7228 * function.c (match_asm_constraints_1): Don't optimize if input
7229 doesn't satisfy general_operand predicate for output's mode.
7232 * alias.c (write_dependence_p): Don't require for x_canonicalized
7233 non-VOIDmode if x has VOIDmode.
7236 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7237 just don't try to optimize it rather than assert it never happens.
7239 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7241 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7242 macro expansions for definition of ST_INTERNAL_<mode> and
7243 LD_INTERNAL_<mode> builtins.
7244 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7246 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7248 (altivec_expand_st_builtin): Likewise.
7249 (altivec_expand_builtin): Remove calls to deleted functions.
7250 (rs6000_address_for_altivec): Delete this function.
7251 * config/rs6000/vector.md: Remove expands for
7252 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7254 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7257 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7258 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7259 re-computing once computed.
7260 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7261 (arm_init_machine_status): Initialize
7262 machine->static_chain_stack_bytes.
7264 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7267 * doc/extend.texi: Add four new prototypes for vec_ld.
7268 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7269 definitions for more logical presentation.
7270 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7271 entries for V1TI variants of __builtin_altivec_ld builtin.
7272 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7273 handling of V1TI variant of LVX icode pattern.
7274 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7275 (rs6000_gimple_fold_builtin): Likewise.
7276 (altivec_init_builtins): Add code to define
7277 __builtin_altivec_lvx_v1ti function.
7279 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7282 * function.c (match_asm_constraints_1): Don't do the optimization
7283 if input isn't a REG, SUBREG, MEM or constant.
7285 2018-03-22 Tom de Vries <tom@codesourcery.com>
7287 PR tree-optimization/84956
7288 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7289 bb_has_abnormal_pred.
7291 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7294 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7295 DECL_INITIAL (decl) to decl at the end.
7296 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7299 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7301 * doc/extend.texi (__builtin_tgmath): Document when complex
7302 integer types are treated as _Complex _Float64.
7304 2018-03-21 Tom de Vries <tom@codesourcery.com>
7306 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7308 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7310 PR tree-optimization/84960
7311 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7312 if it is ENTRY block, move them into single succ of ENTRY in that case.
7314 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7316 PR tree-optimization/84811
7317 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7318 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7319 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7320 (known_subrange_p): Update accordingly. Cast each value involved
7321 in the size comparison, rather than casting the result of the
7324 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7326 PR tree-optimization/84982
7327 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7328 by flipping the least significant bit rather than all bits from
7329 bitpos to bitpos + bitsize - 1.
7331 2018-03-21 Nathan Sidwell <nathan@acm.org>
7333 * doc/extend.texi (Deprecated Features): Remove mention of
7334 long-deleted deprecations.
7336 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7339 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7340 * configure: Regenerate.
7342 2018-03-21 Tom de Vries <tom@codesourcery.com>
7344 PR tree-optimization/83126
7345 * tree-parloops.c (num_phis): New function.
7346 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7348 2018-03-21 Nathan Sidwell <nathan@acm.org>
7350 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7351 mention anon-struct/union members and trailing attributes.
7353 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7355 PR tree-optimization/84969
7356 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7357 builtin memset partitions if they set different rhs values.
7359 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7361 PR rtl-optimization/84989
7362 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7363 VEC_DUPLICATE with scalar result mode.
7365 2018-03-21 Martin Liska <mliska@suse.cz>
7368 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7369 not intended return statement.
7371 2018-03-21 Martin Liska <mliska@suse.cz>
7374 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7375 (chkp_find_bound_slots_1): Limit number of iterations.
7377 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7380 * Minor grammar fixes for x86 options.
7382 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7385 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7386 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7388 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7391 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7392 (altivec_lvx_<mode>_1op): Likewise.
7393 (altivec_stvx_<mode>_2op): Likewise.
7394 (altivec_stvx_<mode>_1op): Likewise.
7395 (altivec_lvx_<VM2:mode>): New define_expand.
7396 (altivec_stvx_<VM2:mode>): Likewise.
7397 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7398 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7399 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7400 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7401 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7402 (rs6000_gen_lvx): Likewise.
7403 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7404 (altivec_expand_stv_builtin): Likewise.
7405 (altivec_expand_builtin): Likewise.
7406 * config/rs6000/vector.md: Likewise.
7408 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7411 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7414 2018-03-20 Richard Biener <rguenther@suse.de>
7417 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7418 sign-conversions as zero, fall back to standard scalar_stmt
7421 2018-03-20 Martin Liska <mliska@suse.cz>
7424 * predict.c (rebuild_frequencies): Handle case when we have
7425 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7427 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7430 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7431 flag_section_anchors.
7432 * varasm.c (use_blocks_for_decl_p): Remove hack for
7433 dw2_force_const_mem.
7436 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7438 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7439 be created, use lowpart_subreg of operands[0] rather than operands[0]
7441 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7442 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7443 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7444 and n constraint instead of aarch64_shift_imm_di and Usd.
7445 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7446 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7448 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7451 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7452 to favor GPR over NEON registers.
7453 (<shift>di3_neon): Likewise.
7455 2018-03-20 Tom de Vries <tom@codesourcery.com>
7458 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7459 (nvptx_process_pars): Emit bar.sync asap and alap.
7461 2018-03-20 Tom de Vries <tom@codesourcery.com>
7464 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7465 seen_label if seen_label is already set.
7467 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7470 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7471 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7472 Use 1U instead of 1. Formatting fixes.
7475 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7476 instead of TREE_TYPE (s1) for the return value.
7478 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7480 PR tree-optimization/84946
7481 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7482 bitsize + bitsize in poly_uint64 rather than poly_int64.
7485 * dwarf2asm.c: Include fold-const.c.
7486 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7487 of decl rather than decl itself.
7489 PR rtl-optimization/84643
7490 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7492 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7495 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7496 calling assemble_variable.
7498 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7501 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7502 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7504 2018-03-19 Jim Wilson <jimw@sifive.com>
7507 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7508 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7509 (riscv_first_stack_step): Likewise.
7510 (riscv_option_override): Use STACK_BOUNDARY instead of
7512 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7514 (BIGGEST_ALIGNMENT): Set to 128.
7515 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7516 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7519 2018-03-19 Richard Biener <rguenther@suse.de>
7521 PR tree-optimization/84933
7522 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7523 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7525 2018-03-19 Richard Biener <rguenther@suse.de>
7527 PR tree-optimization/84859
7528 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7529 (cond_if_else_store_replacement): Perform sinking operation on
7530 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7531 Generalize what a BB with a single eligible store is.
7533 2018-03-19 Richard Biener <rguenther@suse.de>
7535 PR tree-optimization/84929
7536 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7537 chrec_is_positive against non-chrec arg.
7539 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7542 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7544 2018-03-18 Martin Liska <mliska@suse.cz>
7546 PR rtl-optimization/84635
7547 * regrename.c (build_def_use): Use matches_mode only when
7550 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7552 PR tree-optimization/84913
7553 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7554 vectorize chains of COND_EXPRs.
7556 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7558 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7560 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7562 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7564 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7566 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7568 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7569 Kito Cheng <kito.cheng@gmail.com>
7571 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7572 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7573 (nds32_adjust_reg_alloc_order): New function.
7574 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7576 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7578 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7579 nds32_print_operand, nds32_print_operand_address): Use
7580 HOST_WIDE_INT_PRINT_DEC instead.
7582 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7584 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7586 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7589 * config/i386/i386.c (initial_ix86_tune_features,
7590 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7592 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7593 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7594 rather than 1u << ix86_tune. Formatting fix.
7595 (ix86_option_override_internal): Change ix86_arch_mask from
7596 unsigned int to unsigned HOST_WIDE_INT, initialize to
7597 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7598 (ix86_function_specific_restore): Likewise.
7600 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7603 * postreload.c (reload_combine_recognize_pattern): Perform
7604 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7605 truncate_int_for_mode the result for the destination's mode.
7608 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7609 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7611 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7614 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7617 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7618 regno_allocno_class_array and sorted_pseudos.
7619 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7620 insns where regno is used.
7622 2018-03-16 Martin Liska <mliska@suse.cz>
7625 * multiple_target.c (create_dispatcher_calls): Redirect
7626 reference in the symbol table.
7628 2018-03-16 Martin Liska <mliska@suse.cz>
7631 * multiple_target.c (create_dispatcher_calls): Redirect also
7634 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7638 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7639 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7640 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7642 2018-03-16 Julia Koval <julia.koval@intel.com>
7644 * doc/invoke.texi (Skylake Server): Add CLWB.
7645 Cannonlake): Remove CLWB.
7647 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7649 PR tree-optimization/84841
7650 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7652 (FLOAT_ONE_CONST_TYPE): Define.
7653 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7654 (sort_by_operand_rank): Put entries with higher constant_type last
7655 rather than first to match comments.
7657 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7659 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7662 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7665 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7666 check and use address of target always.
7668 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7671 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7672 (indirect_thunk_bnd_needed): Likewise.
7673 (indirect_thunks_used): Likewise.
7674 (indirect_thunks_bnd_used): Likewise.
7675 (indirect_return_needed): New.
7676 (indirect_return_bnd_needed): Likewise.
7677 (output_indirect_thunk_function): Add a bool argument for
7679 (output_indirect_thunk_function): Don't generate alias for
7680 function return thunk.
7681 (ix86_code_end): Call output_indirect_thunk_function to generate
7682 function return thunks.
7683 (ix86_output_function_return): Set indirect_return_bnd_needed
7684 and indirect_return_needed instead of indirect_thunk_bnd_needed
7685 and indirect_thunk_needed.
7687 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7689 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7690 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7691 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7693 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7694 Paul Hua <paul.hua.gm@gmail.com>
7697 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7699 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7701 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7702 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7705 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7708 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7709 instead of GET_MODE_SIZE when comparing Units.
7711 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7714 * varasm.c (hash_section): Return an unchangeble hash value
7715 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7716 Return !aarch64_can_use_per_function_literal_pools_p ().
7718 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7721 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7722 rather than address of cmode as last argument to prepare_cmp_insn.
7724 2018-03-15 Julia Koval <julia.koval@intel.com>
7726 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7727 F_AVX512VNNI, F_AVX512BITALG): New.
7729 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7732 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7733 insn for floating-point loads and stores.
7735 2018-03-14 Carl Love <cel@us.ibm.com>
7737 * config/rs6000/rs6000-c.c: Add macro definitions for
7738 ALTIVEC_BUILTIN_VEC_PERMXOR.
7739 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7740 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7741 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7742 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7744 * config/doc/extend.texi: Add prototypes for vec_permxor.
7746 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7749 * diagnostic-show-locus.c (class layout_point): Convert m_line
7750 from int to linenum_type.
7751 (line_span::comparator): Use linenum "compare" function when
7752 comparing line numbers.
7753 (test_line_span): New function.
7754 (layout_range::contains_point): Convert param "row" from int to
7756 (layout_range::intersects_line_p): Likewise.
7757 (layout::will_show_line_p): Likewise.
7758 (layout::print_source_line): Likewise.
7759 (layout::should_print_annotation_line_p): Likewise.
7760 (layout::print_annotation_line): Likewise.
7761 (layout::print_leading_fixits): Likewise.
7762 (layout::annotation_line_showed_range_p): Likewise.
7763 (struct line_corrections): Likewise for field m_row.
7764 (line_corrections::line_corrections): Likewise for param "row".
7765 (layout::print_trailing_fixits): Likewise.
7766 (layout::get_state_at_point): Likewise.
7767 (layout::get_x_bound_for_row): Likewise.
7768 (layout::print_line): Likewise.
7769 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7770 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7771 * input.c (selftest::test_linenum_comparisons): New function.
7772 (selftest::input_c_tests): Call it.
7773 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7774 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7775 * selftest.h (ASSERT_GT): New macro.
7776 (ASSERT_GT_AT): New macro.
7777 (ASSERT_LT): New macro.
7778 (ASSERT_LT_AT): New macro.
7780 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7782 PR rtl-optimization/84780
7783 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7785 2018-03-14 Martin Liska <mliska@suse.cz>
7787 * tree.c (record_node_allocation_statistics): Use
7788 get_stats_node_kind.
7789 (get_stats_node_kind): New function extracted from
7790 record_node_allocation_statistics.
7791 (free_node): Use get_stats_node_kind.
7793 2018-03-14 Richard Biener <rguenther@suse.de>
7795 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7796 that the value-set of ANTIC_IN doesn't grow.
7799 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7801 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7802 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7804 2018-03-14 Julia Koval <julia.koval@intel.com>
7806 * config.gcc (icelake-client, icelake-server): New.
7808 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7809 (initial_ix86_arch_features): Ditto.
7810 (PTA_SKYLAKE): Add SGX.
7811 (PTA_ICELAKE): Remove.
7812 (PTA_ICELAKE_CLIENT): New.
7813 (PTA_ICELAKE_SERVER): New.
7814 (ix86_option_override_internal): Split up icelake on icelake client and
7816 (get_builtin_code_for_version): Ditto.
7817 (fold_builtin_cpu): Ditto.
7818 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7819 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7820 * config/i386/i386.h (processor_type): Ditto.
7821 * doc/invoke.texi: Ditto.
7823 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7826 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7827 INTEGER_CST offset, add it together with bitpos / 8 and
7828 sign extend based on POINTER_SIZE.
7832 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7835 * config/i386/constraints.md (Yc): New register constraint.
7836 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7837 Use Yc constraint for alternative 2 of operand 0. Remove
7838 preferred_for_speed attribute.
7840 2018-03-14 Richard Biener <rguenther@suse.de>
7842 PR tree-optimization/84830
7843 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7844 with the old one to avoid oscillations.
7846 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7849 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7851 (assign_by_spills): Return a flag of reload assignment failure.
7852 Do not process the reload assignment failures. Do not spill other
7853 reload pseudos if they has the same reg class. Update n if
7855 (lra_assign): Add a return arg. Set up from the result of
7856 assign_by_spills call.
7857 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7858 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7859 usage_insns if it is not NULL.
7860 (spill_hard_reg_in_range): New function.
7861 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7862 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7863 function prototypes.
7864 (lra_assign): Change prototype.
7865 * lra.c (lra): Add code to deal with fails by splitting hard reg
7868 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7870 * config/riscv/riscv.opt (mrelax): New option.
7871 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7872 "norelax" when riscv_mrelax is disabled.
7873 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7875 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7878 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7879 reassociation for int modes.
7881 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7883 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7884 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7886 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7887 * config/aarch64/aarch64-sve.md
7888 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7889 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7890 (*extend<mode><Vwide>2): Rename to...
7891 (aarch64_sve_extend<mode><Vwide>2): ...this.
7892 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7893 renaming the old pattern to...
7894 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7896 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7897 define_expand, renaming the old pattern to...
7898 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7899 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7900 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7901 account when deciding which SVE instruction the optab should use.
7902 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7904 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7906 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7907 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7908 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7909 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7910 (tlsdesc_small_<mode>): Turn a define_expand and use
7911 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7912 (tlsdesc_small_advsimd_<mode>): ...this.
7913 (tlsdesc_small_sve_<mode>): New pattern.
7915 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7917 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7918 (UNSPEC_UMUL_HIGHPART): New constants.
7919 (MUL_HIGHPART): New int iteraor.
7920 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7921 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7923 (*<su>mul<mode>3_highpart): New define_insn.
7925 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
7928 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
7931 2018-03-13 Martin Liska <mliska@suse.cz>
7934 * (sem_item_optimizer::sem_item_optimizer): Initialize new
7936 (sem_item_optimizer::~sem_item_optimizer): Release it.
7937 (sem_item_optimizer::merge_classes): Register variable aliases.
7938 (sem_item_optimizer::fixup_pt_set): New function.
7939 (sem_item_optimizer::fixup_points_to_sets): Likewise.
7940 * ipa-icf.h: Declare new variables and functions.
7942 2018-03-13 Jakub Jelinek <jakub@redhat.com>
7945 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
7946 integer_pow2p@2 and test integer_pow2p in condition.
7947 (A < 0 ? C : 0): Similarly for @1.
7950 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
7951 characters starting at p contain '\0' character, don't look beyond
7955 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
7956 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
7959 * reg-stack.c (change_stack): Change update_end var from int to
7960 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
7961 also call set_block_for_insn on the newly added insns and rescan.
7964 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
7965 on the last operand.
7968 * tree.c (stabilize_reference_1): Return save_expr (e) for
7969 STATEMENT_LIST even if it doesn't have side-effects.
7971 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
7973 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
7975 2018-03-12 Renlin Li <renlin.li@arm.com>
7977 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
7978 aarch64_output_scalar_simd_mov_immediate.
7980 2018-03-12 Martin Sebor <msebor@redhat.com>
7982 PR tree-optimization/83456
7983 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
7984 for perfectly overlapping calls to memcpy.
7985 (gimple_fold_builtin_memory_chk): Same.
7986 (gimple_fold_builtin_strcpy): Handle no-warning.
7987 (gimple_fold_builtin_stxcpy_chk): Same.
7988 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
7990 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
7992 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
7993 parameter. Use it for SFmode.
7994 (rs6000_function_arg_advance_1): Adjust.
7995 (rs6000_function_arg): Adjust.
7996 (rs6000_gimplify_va_arg): Pass false for that new parameter.
7998 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8000 PR rtl-optimization/84169
8001 PR rtl-optimization/84780
8002 * combine.c (can_combine_p): Check for a 2-insn combination whether
8003 the destination register is used between the two insns, too.
8005 2018-03-12 Richard Biener <rguenther@suse.de>
8007 PR tree-optimization/84803
8008 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8009 for refs DR analysis didn't process.
8011 2018-03-12 Richard Biener <rguenther@suse.de>
8013 PR tree-optimization/84777
8014 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8015 force-vectorize loops ignore whether we are optimizing for size.
8017 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8019 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8020 (TARGET_MD_ASM_ADJUST): Define.
8022 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8023 Kito Cheng <kito.cheng@gmail.com>
8024 Chung-Ju Wu <jasonwucj@gmail.com>
8026 * config/nds32/nds32.c (nds32_compute_stack_frame,
8027 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8028 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8029 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8030 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8031 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8032 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8033 * config/nds32/nds32.md (prologue, epilogue): Use macro
8034 NDS32_V3PUSH_AVAILABLE_P to do checking.
8036 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8039 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8040 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8041 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8042 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8043 addition of most attributes on !orig_type_die or the attribute not
8044 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8046 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8047 Chung-Ju Wu <jasonwucj@gmail.com>
8049 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8051 * config/nds32/nds32.opt (mvh): New option.
8053 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8054 Chung-Ju Wu <jasonwucj@gmail.com>
8056 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8058 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8059 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8062 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8063 Chung-Ju Wu <jasonwucj@gmail.com>
8065 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8067 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8068 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8070 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8071 Kito Cheng <kito.cheng@gmail.com>
8072 Chung-Ju Wu <jasonwucj@gmail.com>
8074 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8075 UNSPEC_FFMISM and UNSPEC_FLMISM.
8076 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8077 for ffb, ffmism and flmism.
8078 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8079 (unspec_ffmism): Ditto.
8080 (unspec_flmism): Ditto.
8081 (nds32_expand_builtin_impl): Check if string extension is available.
8082 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8083 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8085 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8088 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8091 * lra-assigns.c (assign_by_spills): Return a flag of reload
8092 assignment failure. Do not process the reload assignment
8093 failures. Do not spill other reload pseudos if they has the same
8095 (lra_assign): Add a return arg. Set up from the result of
8096 assign_by_spills call.
8097 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8098 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8099 usage_insns if it is not NULL.
8100 (spill_hard_reg_in_range): New function.
8101 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8102 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8103 function prototypes.
8104 (lra_assign): Change prototype.
8105 * lra.c (lra): Add code to deal with fails by splitting hard reg
8108 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8111 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8113 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8116 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8117 (dw_val_node): Add val_symbolic_view.
8118 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8119 (symview_upper_bound): New.
8120 (new_line_info_table): Initialize symviews_since_reset.
8121 (dwarf2out_source_line): Count symviews_since_reset and set
8122 symview_upper_bound.
8123 (dw_val_equal_p): Handle symview.
8124 (add_AT_symview): New.
8125 (print_dw_val): Handle symview.
8126 (attr_checksum, attr_checksum_ordered): Likewise.
8127 (same_dw_val_p, size_of_die): Likewise.
8128 (value_format, output_die): Likewise.
8129 (add_high_low_attributes): Use add_AT_symview for entry_view.
8130 (dwarf2out_finish): Reset symview_upper_bound, clear
8133 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8136 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8137 Add strict argument and use it.
8138 (rs6000_split_multireg_move): Update for new strict argument.
8139 (mem_operand_gpr): Disallow all non-offsettable addresses.
8140 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8142 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8145 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8146 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8147 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8150 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8151 decl, use remap_type if we want to use the type.
8153 2018-03-09 Martin Sebor <msebor@redhat.com>
8155 PR tree-optimization/84526
8156 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8158 (builtin_access::generic_overlap): Be prepared to handle non-array
8161 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8163 PR rtl-optimization/84682
8164 * lra-constraints.c (process_address_1): Check is_address flag
8165 for address constraints.
8166 (process_alt_operands): Likewise.
8167 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8168 preprocess_constraints.
8169 * recog.h (preprocess_constraints): Add oploc parameter.
8171 * recog.c (preprocess_constraints): Test address_operand for
8172 CT_ADDRESS constraints.
8174 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8177 * lra-assigns.c (assign_by_spills): Return a flag of reload
8178 assignment failure. Do not process the reload assignment
8179 failures. Do not spill other reload pseudos if they has the same
8181 (lra_assign): Add a return arg. Set up from the result of
8182 assign_by_spills call.
8183 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8184 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8185 usage_insns if it is not NULL.
8186 (spill_hard_reg_in_range): New function.
8187 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8188 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8189 function prototypes.
8190 (lra_assign): Change prototype.
8191 * lra.c (lra): Add code to deal with fails by splitting hard reg
8194 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8197 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8198 Accept complain bool parameter. Only emit errors if it is true.
8199 (arm_parse_cpu_option_name): Likewise.
8200 (arm_target_thumb_only): Adjust callers of the above.
8201 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8202 prototype to take a default true bool parameter.
8203 (arm_parse_arch_option_name): Likewise.
8205 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8206 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8210 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8211 * configure: Regenerate.
8212 * configure.ac ("linker --version-script option"): New.
8213 ("linker soname option"): New.
8215 2018-03-09 Richard Biener <rguenther@suse.de>
8217 PR tree-optimization/84775
8218 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8219 immediate uses of predicate stmts and mark them modified.
8222 PR tree-optimization/84178
8223 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8225 (version_loop_for_if_conversion): Delay update_ssa call.
8226 (tree_if_conversion): Delay update_ssa until after predicate
8229 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8232 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8233 when the function accesses prior frames.
8235 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8238 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8239 gen_llsym, otherwise call maybe_gen_llsym.
8242 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8243 has ',' character inside of it.
8245 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8248 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8249 as clobbering CC_REGNUM.
8251 2018-03-08 Richard Biener <rguenther@suse.de>
8254 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8255 (follow_copies_to_constant): Do not follow SSA names registered
8258 2018-03-08 Richard Biener <rguenther@suse.de>
8260 PR tree-optimization/84178
8261 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8263 (version_loop_for_if_conversion): Delay update_ssa call.
8264 (tree_if_conversion): Delay update_ssa until after predicate
8267 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8269 PR tree-optimization/84178
8270 * tree-if-conv.c (release_bb_predicate): Remove the
8271 the assertion that the stmts have NULL use_ops.
8272 Discard the statements, asserting that they haven't
8273 yet been added to a BB.
8275 2018-03-08 Richard Biener <rguenther@suse.de>
8277 PR tree-optimization/84746
8278 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8279 (phi_translate): Pass in destination ANTIC_OUT set.
8280 (phi_translate_1): Likewise. For a simplified result lookup
8281 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8282 (phi_translate_set): Adjust.
8283 (do_pre_regular_insertion): Likewise.
8284 (do_pre_partial_partial_insertion): Likewise.
8286 2018-03-08 Martin Liska <mliska@suse.cz>
8288 PR gcov-profile/84735
8289 * doc/gcov.texi: Document usage of profile files.
8290 * gcov-io.h: Document changes in the format.
8292 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8296 * dwarf2out.c (struct dw_line_info_table): Update comments for
8298 (FORCE_RESET_NEXT_VIEW): New.
8299 (FORCE_RESETTING_VIEW_P): New.
8300 (RESETTING_VIEW_P): Check for -1 too.
8301 (ZERO_VIEW_P): Likewise.
8302 (new_line_info_table): Force-reset next view.
8303 (dwarf2out_begin_function): Likewise.
8304 (dwarf2out_source_line): Simplify zero_view_p initialization.
8305 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8306 view directly. Omit view when omitting .loc at line 0.
8308 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8310 PR tree-optimization/84740
8311 * tree-switch-conversion.c (process_switch): Call build_constructors
8312 only if info.phi_count is non-zero.
8314 PR tree-optimization/84739
8315 * tree-tailcall.c (find_tail_calls): Check call arguments against
8316 DECL_ARGUMENTS (current_function_decl) rather than
8317 DECL_ARGUMENTS (func) when checking for tail recursion.
8319 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8321 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8322 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8323 Volker Reichelt's entry and add entries for people that perform
8324 GCC fuzzy testing and report numerous bugs.
8326 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8329 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8330 readonly data in sdata, if that is disabled.
8331 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8332 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8333 -mreadonly-in-sdata option.
8335 2018-03-07 Martin Sebor <msebor@redhat.com>
8337 PR tree-optimization/84468
8338 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8339 basic block when looking for nul assignment.
8341 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8344 * except.h (output_function_exception_table): Adjust prototype.
8345 * except.c (output_function_exception_table): Remove FNNAME parameter
8346 and add SECTION parameter. Ouput one part of the table at a time.
8347 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8348 the first part of the exception table and emit unwind directives.
8349 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8350 (i386_pe_seh_cold_init): Likewise.
8351 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8352 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8353 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8354 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8355 * config/i386/winnt.c: Include except.h.
8356 (struct seh_frame_state): Add reg_offset, after_prologue and
8357 in_cold_section fields.
8358 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8359 (i386_pe_seh_cold_init): New function.
8360 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8361 to seh->in_cold_section.
8362 (seh_emit_push): Record the offset of the push.
8363 (seh_emit_save): Record the offet of the save.
8364 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8365 Test seh->after_prologue to disregard the epilogue.
8366 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8367 (i386_pe_end_cold_function): New function.
8369 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8372 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8373 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8376 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8378 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8379 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8382 * multiple_target.c: Include tree-inline.h and intl.h.
8383 (expand_target_clones): Diagnose and fail if node->definition and
8384 !tree_versionable_function_p (node->decl).
8386 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8388 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8390 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8391 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8392 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8394 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8397 * combine.c (try_combine): Use reg_or_subregno instead of handling
8398 just paradoxical SUBREGs and REGs.
8400 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8402 * config/arc/arc.c (arc_finalize_pic): Remove function.
8403 (arc_must_save_register): We use single base PIC register, remove
8404 checks to save/restore the PIC register.
8405 (arc_expand_prologue): Likewise.
8406 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8408 (arc_verify_short): Likewise.
8409 (arc_attr_type): Likewise.
8410 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8411 (walk_stores): Likewise.
8412 (arc_address_cost): Make it static.
8413 (arc_verify_short): Likewise.
8414 (branch_dest): Likewise.
8415 (arc_attr_type): Likewise.
8416 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8417 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8418 (arc_final_prescan_insn): Remove inserting the nops due to
8419 hardware hazards. It is done in reorg step.
8420 (insn_length_variant_t): Remove.
8421 (insn_length_parameters_t): Likewise.
8422 (arc_insn_length_parameters): Likewise.
8423 (arc_get_insn_variants): Likewise.
8424 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8426 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8429 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8432 PR tree-optimization/84687
8433 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8435 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8437 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8439 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8440 Rename to ppc_speculation_barrier.
8441 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8442 __builtin_ppc_speculation_barrier.
8444 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8447 * combine.c (combine_simplify_rtx): Don't try to simplify if
8448 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8451 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8453 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8454 to 32 bytes when compiling for POWER9.
8456 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8459 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8460 regparm >= 3 with no arg reg available also for calls with
8461 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8464 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8466 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8468 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8471 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8473 2018-03-05 Richard Biener <rguenther@suse.de>
8475 PR tree-optimization/84486
8476 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8477 When inserting a __builtin_assume_aligned call set the LHS
8478 SSA name alignment info accordingly.
8480 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8482 PR tree-optimization/84114
8483 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8484 Avoid reassociation of FLOAT_MODE addition.
8486 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8488 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8489 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8490 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8491 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8492 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8493 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8494 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8496 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8497 __builtin_ia32_wbinvd): New builtins.
8498 (SPECIAL_ARGS2): New.
8499 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8500 (SPECIAL_ARGS2): New.
8501 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8502 (ix86_valid_target_attribute_inner_p): Ditto.
8503 (ix86_init_mmx_sse_builtins): Add special_args2.
8504 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8505 TARGET_WBNOINVD_P): New.
8506 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8507 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8508 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8509 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8510 * config/i386/pconfigintrin.h: New file.
8511 * config/i386/wbnoinvdintrin.h: Ditto.
8512 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8514 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8516 2018-03-05 Richard Biener <rguenther@suse.de>
8518 PR tree-optimization/84670
8519 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8521 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8522 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8523 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8524 doesn't grow if all successors (recursively) were visited at least
8527 2018-03-05 Richard Biener <rguenther@suse.de>
8529 PR tree-optimization/84650
8530 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8531 if executed in the loop pipeline.
8533 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8535 * doc/configfiles.texi (Configuration Files): Move info about
8536 conditionalizing $target-protos.h to...
8537 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8538 differs from $target-protos.h.
8540 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8541 Chung-Ju Wu <jasonwucj@gmail.com>
8543 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8544 * config/nds32/nds32-multiple.md (setmemsi): Define.
8545 * config/nds32/nds32-memory-manipulation.c
8546 (nds32_gen_dup_4_byte_to_word_value): New.
8547 (emit_setmem_word_loop): New.
8548 (emit_setmem_byte_loop): New.
8549 (nds32_expand_setmem_loop): New.
8550 (nds32_expand_setmem_loop_v3m): New.
8551 (nds32_expand_setmem_unroll): New.
8552 (nds32_expand_setmem): New.
8554 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8555 Chung-Ju Wu <jasonwucj@gmail.com>
8557 * config/nds32/nds32-memory-manipulation.c
8558 (nds32_emit_load_store): New.
8559 (nds32_emit_post_inc_load_store): New.
8560 (nds32_emit_mem_move): New.
8561 (nds32_emit_mem_move_block): New.
8562 (nds32_expand_movmemsi_loop_unknown_size): New.
8563 (nds32_expand_movmemsi_loop_known_size): New.
8564 (nds32_expand_movmemsi_loop): New.
8565 (nds32_expand_movmemsi_unroll): New.
8566 (nds32_expand_movmemqi): Rename ...
8567 (nds32_expand_movmemsi): ... to this.
8568 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8569 (movmemsi): ... to this.
8570 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8571 (nds32_expand_movmemsi): ... to this.
8573 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8574 Monk Chiang <sh.chiang04@gmail.com>
8575 Chung-Ju Wu <jasonwucj@gmail.com>
8577 * config/nds32/nds32-protos.h
8578 (nds32_expand_load_multiple): New arguments.
8579 (nds32_expand_store_multiple): Ditto.
8580 (nds32_valid_multiple_load_store): Rename ...
8581 (nds32_valid_multiple_load_store_p): ... to this.
8582 * config/nds32/nds32-memory-manipulation.c
8583 (nds32_expand_load_multiple): Refine implementation.
8584 (nds32_expand_store_multiple): Ditto.
8585 * config/nds32/nds32-multiple.md
8586 (load_multiple): Update nds32_expand_load_multiple interface.
8587 (store_multiple): Update nds32_expand_store_multiple interface.
8588 * config/nds32/nds32-predicates.c
8589 (nds32_valid_multiple_load_store): Rename ...
8590 (nds32_valid_multiple_load_store_p): ... to this and refine
8592 * config/nds32/predicates.md
8593 (nds32_load_multiple_and_update_address_operation): New predicate.
8594 (nds32_store_multiple_and_update_address_operation): New predicate.
8596 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8597 Chung-Ju Wu <jasonwucj@gmail.com>
8599 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8600 (combo): New attribute.
8601 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8603 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8605 * config/nds32/nds32.opt: Change -mcmodel= default value.
8607 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8608 Monk Chiang <sh.chiang04@gmail.com>
8609 Chung-Ju Wu <jasonwucj@gmail.com>
8611 * config/nds32/constants.md (unspec_element): New enum.
8612 * config/nds32/constraints.md (Umw): New constraint.
8613 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8614 * config/nds32/nds32-intrinsic.md: Likewise.
8615 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8616 (nds32_valid_smw_lwm_base_p): New.
8617 (nds32_output_smw_single_word): New.
8618 (nds32_output_lmw_single_word): New.
8619 (nds32_expand_unaligned_load): New.
8620 (nds32_expand_unaligned_store): New.
8621 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8622 (nds32_output_smw_single_word): Declare.
8623 (nds32_output_lmw_single_word): Declare.
8624 (nds32_expand_unaligned_load): Declare.
8625 (nds32_expand_unaligned_store): Declare.
8626 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8627 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8628 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8629 NDS32_BUILTIN_UASTORE_DW.
8630 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8633 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8634 Kito Cheng <kito.cheng@gmail.com>
8635 Chung-Ju Wu <jasonwucj@gmail.com>
8637 * config/nds32/nds32-intrinsic.c
8638 (nds32_expand_builtin_null_ftype_reg): Delete.
8639 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8640 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8641 (nds32_read_argument): New.
8642 (nds32_legitimize_target): Ditto.
8643 (nds32_legitimize_argument): Ditto.
8644 (nds32_check_constant_argument): Ditto.
8645 (nds32_expand_unop_builtin): Ditto.
8646 (nds32_expand_unopimm_builtin): Ditto.
8647 (nds32_expand_binop_builtin): Ditto.
8648 (nds32_builtin_decl_impl): Ditto.
8649 (builtin_description): Ditto.
8650 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8651 (nds32_init_builtins_impl): Ditto.
8652 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8653 (nds32_builtin_decl): New.
8654 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8655 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8657 2018-03-02 Jeff Law <law@redhat.com>
8659 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8660 (redundant_insn, fill_simple_delay_slots): Likewise.
8661 (fill_slots_from_thread): Likewise.
8662 * resource.c (mark_referenced_resources): Likewise.
8663 (mark_set_resources, find_dead_or_set_registers): Likewise.
8665 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8667 * substring-locations.h (format_warning_va): Formatting fix for
8669 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8671 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8672 * substring-locations.c: Include intl.h.
8673 (format_warning_va): Turned into small wrapper around
8674 format_warning_n_va, renamed to ...
8675 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8676 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8678 (format_warning_at_substring_n): New function.
8679 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8680 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8681 format_warning_at_substring with just a shorter name instead of
8682 const function pointer.
8683 (fmtwarn_n): New function.
8684 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8685 appropriate, get rid of all the fmtstr temporaries, move conditionals
8686 with G_() wrapped string literals directly into fmtwarn arguments,
8687 cast dir.len to (int), formatting fixes.
8689 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8691 * doc/invoke.texi: Remove "Cilk Plus" references.
8693 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8694 Richard Biener <rguenther@suse.de>
8697 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8698 for error or warning attributes if CALL_FROM_THUNK_P is set.
8701 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8704 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8705 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8708 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8709 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8711 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8712 instead of -1U in last predictors element's probability member.
8714 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8717 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8718 arguments if they are comparable.
8720 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8722 PR tree-optimization/84634
8723 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8724 masks and masked_loop_p with a single loop_masks, making sure it's
8725 null for bb vectorization.
8727 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8729 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8730 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8731 loop->force_vectorize to check whether there is no alias.
8733 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8736 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8738 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8740 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8741 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8742 instead of a loop around prev_real_insn.
8743 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8747 * config/i386/i386.c (ix86_print_operand): Use conditional
8748 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8751 2018-03-02 Richard Biener <rguenther@suse.de>
8753 PR tree-optimization/84427
8754 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8755 (bitmap_set_subtract_values): Rewrite to handle multiple
8758 (prune_clobbered_mems): Likewise.
8759 (phi_translate): Take edge instead of pred/phiblock.
8760 (phi_translate_1): Likewise.
8761 (phi_translate_set): Likewise. Insert all translated
8762 exprs for a value into the set, keeping possibly multiple
8763 expressions per value.
8764 (compute_antic_aux): Adjust for phi_translate changes.
8765 When intersecting union the expressions and prune those
8766 not in the final value set, keeping possibly multiple
8767 expressions per value. Do not use value-insertion
8768 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8769 all expressions. Add verification that the value-sets
8770 only shrink during iteration.
8771 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8772 (do_pre_regular_insertion): Likewise.
8773 (do_pre_partial_partial_insertion): Likewise.
8775 2018-03-02 Richard Biener <rguenther@suse.de>
8778 * config/darwin.c (saved_debug_info_level): New static global.
8779 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8780 (darwin_asm_lto_end): Restore debug info generation settings.
8782 2018-03-01 Martin Liska <mliska@suse.cz>
8785 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8788 2018-03-01 Richard Biener <rguenther@suse.de>
8791 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8792 type annotation with LTO when debug was disabled at compile-time.
8794 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8796 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8798 (mips_final_postscan_insn): Likewise.
8800 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8802 PR rtl-optimization/84528
8803 * alias.c (init_alias_target): Add commentary.
8804 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8805 a unique base value if the frame pointer is not eliminated
8806 to the stack pointer.
8808 2018-03-01 Tom de Vries <tom@codesourcery.com>
8810 PR rtl-optimization/83327
8811 * lra-int.h (hard_regs_spilled_into): Declare.
8812 * lra.c (hard_regs_spilled_into): Define.
8813 (init_reg_info): Init hard_regs_spilled_into.
8814 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8815 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8816 (process_bb_lives): Handle hard_regs_spilled_into.
8817 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8818 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8820 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8822 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8823 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8824 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8825 * config/rs6000/aix72.h: New file.
8827 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8829 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8830 instead of warning_at with conditional singular and plural messages
8834 * stor-layout.c (update_alignment_for_field): For
8835 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8836 && !DECL_PACKED (field), do the alignment update, just use
8837 only desired_align instead of MAX (type_align, desired_align)
8839 (place_field): Don't do known_align < desired_align handling
8840 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8841 is non-NULL, instead do it after rli->prev_field handling and
8842 only if not within a bitfield word. For DECL_PACKED (field)
8843 use type_align of BITS_PER_UNIT.
8845 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8847 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8848 superfluous parentheses and trailing spaces.
8850 2018-02-28 Richard Biener <rguenther@suse.de>
8852 PR tree-optimization/84584
8853 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8854 SCoPs with fake exit edge.
8856 2018-02-28 Martin Liska <mliska@suse.cz>
8859 * timevar.c (timer::print): Fix format to properly print 100%
8862 2018-02-28 Richard Biener <rguenther@suse.de>
8865 * genmatch.c (capture_info::walk_match): Do not mark
8866 captured expressions without operands as expr_p given
8867 they act more like predicates and should be subject to
8868 "lost tail" side-effect preserving.
8870 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8872 PR rtl-optimization/81611
8873 * auto-inc-dec.c (attempt_change): Move dead note from
8874 mem_insn if it's the next use of regno
8875 (find_address): Take address use of reg holding
8876 non-incremented value. Add parm to limit search to the named
8878 (merge_in_block): Attempt to use a mem insn that is the next
8879 use of the original regno.
8881 2018-02-27 Martin Sebor <msebor@redhat.com>
8884 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8885 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8887 2018-02-27 Martin Sebor <msebor@redhat.com>
8889 PR translation/84207
8890 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8891 n argument to unsigned HOST_WIDE_INT.
8892 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8893 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8894 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8895 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8897 2018-02-27 Richard Biener <rguenther@suse.de>
8899 PR tree-optimization/84512
8900 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8901 Do not use the estimate returned from record_stmt_cost for
8902 the scalar iteration cost but sum properly using add_stmt_cost.
8904 2018-02-27 Richard Biener <rguenther@suse.de>
8906 PR tree-optimization/84466
8907 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8908 Adjust last change to less strictly validate use operands.
8910 2018-02-27 Martin Liska <mliska@suse.cz>
8912 PR gcov-profile/84548
8913 * gcov.c (process_file): Allow partial overlap and consider it
8914 also as group functions.
8915 (output_lines): Properly calculate range of lines for a group.
8917 2018-02-27 Martin Liska <mliska@suse.cz>
8919 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8920 'ggc' suffixes. Change first column width.
8921 (timer::print): Fix formatting of the column.
8923 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8925 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8926 preserve inline entry blocks for the sake of debug inline
8927 entry point markers alone.
8928 (remove_unused_locals): Suggest in comments a better place to
8929 force the preservation of inline entry blocks that are
8930 otherwise unused, but do not preserve them.
8932 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8934 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
8936 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8939 * config/i386/constraints.md (Bs): Replace
8940 ix86_indirect_branch_register with
8941 TARGET_INDIRECT_BRANCH_REGISTER.
8943 * config/i386/i386.md (indirect_jump): Likewise.
8944 (tablejump): Likewise.
8945 (*sibcall_memory): Likewise.
8946 (*sibcall_value_memory): Likewise.
8947 Peepholes of indirect call and jump via memory: Likewise.
8948 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
8949 (*sibcall_value_GOT_32): Likewise.
8950 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8951 (GOT_memory_operand): Likewise.
8952 (call_insn_operand): Likewise.
8953 (sibcall_insn_operand): Likewise.
8954 (GOT32_symbol_operand): Likewise.
8955 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
8957 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8959 PR rtl-optimization/83496
8960 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
8961 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
8962 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
8963 redundant insn, if any.
8964 (relax_delay_slots): Likewise.
8965 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
8967 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
8969 PR tree-optimization/83965
8970 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
8971 that grouped statements are part of a reduction chain. Return
8972 true if the statement is not marked as a reduction itself but
8974 (vect_recog_dot_prod_pattern): Don't check whether the statement
8975 is part of a group here.
8976 (vect_recog_sad_pattern): Likewise.
8977 (vect_recog_widen_sum_pattern): Likewise.
8979 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
8982 * final.c (rest_of_clean_state): Also look for calls inside sequences.
8984 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
8987 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
8989 (ix86_output_indirect_function_return): New prototype.
8990 (ix86_split_simple_return_pop_internal): Likewise.
8991 * config/i386/i386.c (indirect_return_via_cx): New.
8992 (indirect_return_via_cx_bnd): Likewise.
8993 (indirect_thunk_name): Handle return va CX_REG.
8994 (output_indirect_thunk_function): Create alias for
8995 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
8996 (ix86_output_indirect_jmp): Remove the bool argument.
8997 (ix86_output_indirect_function_return): New function.
8998 (ix86_split_simple_return_pop_internal): Likewise.
8999 * config/i386/i386.md (*indirect_jump): Don't pass false
9000 to ix86_output_indirect_jmp.
9001 (*tablejump_1): Likewise.
9002 (simple_return_pop_internal): Change it to define_insn_and_split.
9003 Call ix86_split_simple_return_pop_internal to split it for
9005 (simple_return_indirect_internal): Call
9006 ix86_output_indirect_function_return instead of
9007 ix86_output_indirect_jmp.
9009 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9012 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9013 memset and value initialization afterwards.
9015 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9017 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9019 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9022 * common/config/aarch64/aarch64-common.c
9023 (aarch_option_optimization_table[]): Switch
9024 off fomit-frame-pointer
9026 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9027 Chung-Ju Wu <jasonwucj@gmail.com>
9029 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9031 (store_multiple): Ditto.
9033 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9035 * config.gcc: Add --with-cpu support for nds32 target.
9036 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9037 * config/nds32/nds32.opt: Add -mcpu= option.
9039 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9041 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9042 isel=yes): Warn for these deprecated options.
9044 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9046 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9047 ISA_2_5_MASKS_EMBEDDED.
9049 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9051 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9052 p->max as pointers rather than using iterative_hash_expr.
9054 2018-02-23 Carl Love <cel@us.ibm.com>
9056 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9057 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9059 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9060 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9061 P8V_BUILTIN_VEC_VUNSIGNED2.
9063 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9066 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9067 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9069 * lra-constraints.c (curr_insn_transform): Set up
9070 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9072 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9073 * lra-eliminations.c (spill_pseudos): Ditto.
9074 (process_insn_for_elimination): Ditto.
9075 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9076 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9077 LRA_NON_CLOBBERED_ALT.
9079 2018-02-22 Martin Sebor <msebor@redhat.com>
9081 PR tree-optimization/84480
9082 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9083 to maybe_diag_stxncpy_trunc. Call it.
9084 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9085 from gimple_fold_builtin_strcpy. Print inlining stack.
9086 (handle_builtin_stxncpy): Print inlining stack.
9087 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9089 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9092 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9093 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9094 and -fcheck-pointer-bounds are used together.
9095 (indirect_thunk_prefix): New enum.
9096 (indirect_thunk_need_prefix): New function.
9097 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9098 "_nt" instead of "_bnd" for NOTRACK prefix.
9099 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9100 (output_indirect_thunk_function): Likewise.
9102 (ix86_code_end): Update output_indirect_thunk_function calls.
9103 (ix86_output_indirect_branch_via_reg): Replace
9104 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9105 (ix86_output_indirect_branch_via_push): Likewise.
9106 (ix86_output_function_return): Likewise.
9107 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9108 incompatible with -fcf-protection=branch and
9109 -fcheck-pointer-bounds.
9111 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9114 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9115 Change gcc_assert call to output_operand_lossage.
9117 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9119 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9121 2018-02-22 DJ Delorie <dj@redhat.com>
9122 Sebastian Perta <sebastian.perta@renesas.com>
9123 Oleg Endo <olegendo@gcc.gnu.org>
9125 * config/rx/rx.c (rx_rtx_costs): New function.
9126 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9128 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9130 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9132 2018-02-22 Martin Liska <mliska@suse.cz>
9135 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9136 Add "native" as a possible value.
9138 2018-02-22 Martin Liska <mliska@suse.cz>
9141 * config/i386/i386.c (ix86_option_override_internal):
9142 Add "native" as a possible value for -march and -mtune.
9144 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9147 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9148 to all type variants.
9150 PR tree-optimization/84503
9151 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9152 width as info->bitpos + info->bitsize - start.
9153 (merged_store_group::merge_overlapping): Simplify width computation.
9154 (check_no_overlap): New function.
9155 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9156 start + width and last_order of the group, fail if check_no_overlap
9158 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9159 to group if check_no_overlap fails.
9161 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9163 * config/rs6000/altivec.md: Delete contraint arguments to
9164 define_expand, define_split, and define_peephole2, and in
9165 define_insn_and_split if always unused.
9166 * config/rs6000/darwin.md: Ditto.
9167 * config/rs6000/dfp.md: Ditto.
9168 * config/rs6000/rs6000.md: Ditto.
9169 * config/rs6000/sync.md: Ditto.
9170 * config/rs6000/vector.md: Ditto.
9171 * config/rs6000/vsx.md: Ditto.
9173 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9175 * config/rs6000/altivec.md: Write output control strings as braced
9176 blocks instead of double-quoted strings.
9177 * config/rs6000/darwin.md: Ditto.
9178 * config/rs6000/rs6000.md: Ditto.
9179 * config/rs6000/vector.md: Ditto.
9180 * config/rs6000/vsx.md: Ditto.
9182 2018-02-21 Jason Merrill <jason@redhat.com>
9184 PR c++/84314 - ICE with templates and fastcall attribute.
9185 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9187 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9189 * ipa-cp.c (determine_versionability): Fix comment typos.
9191 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9194 * ipa-cp.c (determine_versionability): Do not version functions caling
9197 2018-02-21 Martin Liska <mliska@suse.cz>
9200 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9201 Add "native" as a possible value.
9202 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9203 the macro when native cpu detection is available.
9205 2018-02-21 Martin Liska <mliska@suse.cz>
9208 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9209 Add "native" as a possible value.
9210 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9211 when native cpu detection is available.
9213 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9214 Martin Sebor <msebor@redhat.com>
9216 PR tree-optimization/84478
9217 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9219 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9220 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9221 support which is conservatively correct, for 2 only stay conservative
9222 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9223 argument to the 2 argument get_range_strlen, adjust 6 arg
9224 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9226 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9227 (gimple_fold_builtin_strlen): Pass true as last argument to
9230 2018-02-20 Martin Sebor <msebor@redhat.com>
9233 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9234 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9235 (builtin_memref::builtin_memref): Factor out parts into
9236 set_base_and_offset and call it.
9238 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9241 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9242 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9243 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9244 search at the associated MODE_INT.
9246 2018-02-20 Jeff Law <law@redhat.com>
9249 PR tree-optimization/81592
9251 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9252 for range data rather than using global data.
9253 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9254 range data rather than using global data.
9255 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9256 pass it to children as needed.
9257 (struct directive::fmtresult): Similarly.
9258 (struct directive::set_width): Similarly.
9259 (struct directive::set_precision): Similarly.
9260 (format_integer, format_directive, parse_directive): Similarly.
9261 (format_none): Accept unnamed vr_values parameter.
9262 (format_percent, format_floating, format_character): Similarly.
9263 (format_string, format_plain): Similarly.
9264 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9265 the EVRP range analyzer for range data rather than using global data.
9266 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9267 gimple-ssa-evrp-analyze.h
9268 (class sprintf_dom_walker): Add after_dom_children member function.
9269 Add evrp_range_analyzer member.
9270 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9271 range analyzer as needed.
9272 (sprintf_dom_walker::after_dom_children): New member function.
9273 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9275 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9276 (evrp_range_analyzer::pop_to_marker): Likewise.
9278 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9280 PR tree-optimization/84419
9281 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9282 with the required type if its current type is compatible but
9285 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9288 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9289 after vectorization.
9291 2018-02-20 Martin Liska <mliska@suse.cz>
9294 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9295 possible values if we don't have a hint.
9297 2018-02-20 Martin Liska <mliska@suse.cz>
9301 * final.c (shorten_branches): Build align_tab array with one
9303 * opts.c (finish_options): Add alignment option limit check.
9304 (MAX_CODE_ALIGN): Likewise.
9305 (MAX_CODE_ALIGN_VALUE): Likewise.
9306 * doc/invoke.texi: Document maximum allowed option value for
9307 all -falign-* options.
9309 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9312 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9313 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9314 * var-tracking.c (emit_note_insn_var_location): Remove all references
9315 to NOTE_INSN_CALL_ARG_LOCATION.
9316 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9317 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9318 Use copy_rtx_if_shared.
9319 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9320 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9321 (dwarf2out_var_location): Remove handling of
9322 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9324 * final.c (final_scan_insn): Remove all references to
9325 NOTE_INSN_CALL_ARG_LOCATION.
9326 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9327 before dumping final insns.
9328 * except.c (emit_note_eh_region_end): Remove all references to
9329 NOTE_INSN_CALL_ARG_LOCATION.
9330 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9331 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9332 * config/arc/arc.c (hwloop_optimize): Likewise.
9333 * config/arm/arm.c (create_fix_barrier): Likewise.
9334 * config/s390/s390.c (s390_chunkify_start): Likewise.
9335 * config/sh/sh.c (find_barrier): Likewise.
9336 * config/i386/i386.c (rest_of_insert_endbranch,
9337 ix86_seh_fixup_eh_fallthru): Likewise.
9338 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9339 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9340 * config/frv/frv.c (frv_function_prologue): Likewise.
9341 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9343 (note_outside_basic_block_p): Remove all references to
9344 NOTE_INSN_CALL_ARG_LOCATION.
9345 * gengtype.c (adjust_field_rtx_def): Likewise.
9346 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9348 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9349 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9352 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9355 PR tree-optimization/84452
9356 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9357 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9360 2018-02-19 Martin Liska <mliska@suse.cz>
9363 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9365 2018-02-19 Martin Liska <mliska@suse.cz>
9366 Richard Sandiford <richard.sandiford@linaro.org>
9368 PR tree-optimization/82491
9369 * gimple-fold.c (get_base_constructor): Make earlier bail out
9372 2018-02-19 Carl Love <cel@us.ibm.com>
9374 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9375 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9377 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9378 P8V_BUILTIN_VEC_NEG.
9380 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9382 * config/rl78/rl78.md (movdf): New define expand.
9384 2018-02-19 Martin Liska <mliska@suse.cz>
9387 * doc/invoke.texi: Fix typo.
9388 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9390 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9392 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9393 handle rs6000_single_float and rs6000_double_float specially for
9396 2018-02-16 Jeff Law <law@redhat.com>
9398 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9401 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9404 * ipa-inline.c (inline_small_functions): Fix a typo.
9406 2018-02-16 Nathan Sidwell <nathan@acm.org>
9408 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9410 2018-02-16 Carl Love <cel@us.ibm.com>
9412 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9413 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9414 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9415 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9416 expansion to P8V_BUILTIN_VEC_FLOAT2.
9418 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9420 PR rtl-optimization/70023
9421 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9422 src_regno into account.
9424 2018-02-16 Carl Love <cel@us.ibm.com>
9426 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9427 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9428 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9429 * config/rs6000/rs6000.c: Remove case statements for
9430 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9431 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9432 and P9V_BUILTIN_VEC_VINSERT4B.
9433 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9434 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9435 * config/rs6000/vsx.md:
9436 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9439 2018-02-16 Carl Love <cel@us.ibm.com>
9441 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9443 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9445 * config/rs6000/rs6000-c.c: Add the definitions for
9446 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9447 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9448 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9449 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9450 definition for insert4b and define insn *insert3b_internal.
9451 * doc/extend.texi: Add documentation for vec_extract4b.
9453 2018-02-16 Nathan Sidwell <nathan@acm.org>
9455 * doc/extend.texi (Backwards Compatibility): Mention friend
9456 injection. Note for-scope is deprecated.
9457 * doc/invoke.texi (-ffriend-injection): Deprecate.
9459 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9461 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9462 that moved to I2, also allow destinations that are a paradoxical
9463 subreg (instead of a normal reg).
9465 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9468 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9471 2018-02-16 Richard Biener <rguenther@suse.de>
9473 PR tree-optimization/84037
9474 PR tree-optimization/84016
9476 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9477 Adjust vec_construct for the fact we need additional higher latency
9478 128bit inserts for AVX256 and AVX512 vector builds.
9479 (ix86_add_stmt_cost): Scale vector construction cost for
9482 2018-02-16 Richard Biener <rguenther@suse.de>
9484 PR tree-optimization/84417
9485 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9486 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9487 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9489 2018-02-16 Martin Liska <mliska@suse.cz>
9492 * internal-fn.def (ASAN_CHECK): Set proper flags.
9493 (ASAN_MARK): Likewise.
9495 2018-02-16 Julia Koval <julia.koval@intel.com>
9497 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9498 from PTA_CANNONLAKE.
9500 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9503 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9504 Use ++iter rather than iter++ for std::list iterators.
9505 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9506 defer deleting them until all nodes in the forest are processed. Do
9507 free even leaf nodes. Change to_process into auto_vec.
9510 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9511 * vec.h (vec_default_construct): Use memset instead of placement new
9512 if BROKEN_VALUE_INITIALIZATION is defined.
9513 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9514 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9517 PR rtl-optimization/83723
9518 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9519 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9520 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9522 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9524 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9526 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9528 PR rtl-optimization/81443
9529 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9530 from inner REGs to paradoxical SUBREGs.
9532 2018-02-16 Richard Biener <rguenther@suse.de>
9534 PR tree-optimization/84399
9535 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9536 For operands we can analyze at their definition make sure we can
9537 analyze them at each use as well.
9539 2018-02-16 Richard Biener <rguenther@suse.de>
9541 PR tree-optimization/84190
9542 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9543 volatile accesses if the decl isn't volatile.
9545 2018-02-15 Jason Merrill <jason@redhat.com>
9547 PR c++/84314 - ICE with templates and fastcall attribute.
9548 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9549 TYPE_CANONICAL on an existing type.
9551 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9553 PR tree-optimization/84383
9554 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9555 dstoff nor call operand_equal_p if dstbase is NULL.
9557 PR tree-optimization/84334
9558 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9559 also a CONSTANT_CLASS_P, punt.
9561 2018-02-14 Jim Wilson <jimw@sifive.com>
9563 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9564 first SMALL_OPERAND check. New local min_second_step. Move assert
9565 to where locals are set. Add TARGET_RVC support.
9566 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9568 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9570 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9572 2018-02-14 Martin Sebor <msebor@redhat.com>
9574 PR tree-optimization/83698
9575 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9576 arrays constrain the offset range to their bounds.
9577 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9578 (builtin_access::overlap): Avoid setting the size of overlap if it's
9580 (maybe_diag_overlap): Also consider arrays when deciding what values
9581 of offsets to include in diagnostics.
9583 2018-02-14 Martin Sebor <msebor@redhat.com>
9586 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9587 that correspond to the kind of a declaration.
9589 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9592 * config/pa/pa.md: Load address of PIC label using the linkage table
9593 if the label is nonlocal.
9595 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9597 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9598 warning message if user requests -maltivec=be.
9599 * doc/invoke.texi: Document deprecation of -maltivec=be.
9601 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9604 * config/rs6000/rs6000-c.c: Update definitions for
9605 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9606 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9608 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9611 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9612 add _get_ssp intrinsics. Remove argument from
9613 __builtin_ia32_rdssp[d|q].
9614 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9615 * config/i386/i386-builtin.def: Remove argument from
9616 __builtin_ia32_rdssp[d|q].
9617 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9618 ix86_expand_special_args_builtin for _rdssp[d|q].
9619 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9620 Clear register before usage.
9621 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9622 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9624 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9626 PR tree-optimization/84357
9627 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9628 operand 1 of an ARRAY_REF too.
9630 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9633 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9634 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9636 (set_of_reg): New struct.
9637 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9638 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9639 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9641 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9642 Split into bitclr, bitset, bitinvert patterns if appropriate.
9643 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9644 use rx_fuse_in_memory_bitop.
9645 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9646 to named insn, correct maximum insn length.
9648 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9651 * machmode.def: Define a complex mode for PARTIAL_INT.
9652 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9654 * doc/rtl.texi: Document CSPImode.
9655 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9657 (msp430_hard_regno_nregs_with_padding): Likewise.
9659 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9662 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9664 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9666 PR rtl-optimization/84169
9667 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9668 we generated a parallel as new i3 and we split that to new i2 and i3
9669 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9670 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9671 those to i2, not i1. Partially rewrite this scan code.
9673 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9676 * stor-layout.c (place_field): For variable length fields, adjust
9677 offset_align afterwards not just based on the field's alignment,
9678 but also on the size.
9681 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9682 of exps and logs in the use_exp2 case.
9684 2018-02-13 Jeff Law <law@redhat.com>
9686 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9689 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9692 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9696 * common.opt (gas-loc-support, gas-locview-support): New.
9697 (ginline-points, ginternal-reset-location-views): New.
9698 * doc/invoke.texi: Document them. Use @itemx where intended.
9699 (gvariable-location-views): Adjust.
9700 * target.def (reset_location_view): New.
9701 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9702 (TARGET_RESET_LOCATION_VIEW): New.
9703 * doc/tm.texi: Rebuilt.
9704 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9705 (dwarf2out_default_as_locview_support): New.
9706 (output_asm_line_debug_info): Use option variables.
9707 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9708 (output_loc_list): Likewise.
9709 (add_high_low_attributes): Check option variables.
9710 Don't output entry view attribute in strict mode.
9711 (gen_inlined_subroutine_die): Check option variables.
9712 (dwarf2out_inline_entry): Likewise.
9713 (init_sections_and_labels): Likewise.
9714 (dwarf2out_early_finish): Likewise.
9715 (maybe_reset_location_view): New, from...
9716 (dwarf2out_var_location): ... here. Call it.
9717 * debug.h (dwarf2out_default_as_loc_support): Declare.
9718 (dwarf2out_default_as_locview_support): Declare.
9719 * hooks.c (hook_int_rtx_insn_0): New.
9720 * hooks.h (hook_int_rtx_insn_0): Declare.
9721 * toplev.c (process_options): Take -gas-loc-support and
9722 -gas-locview-support from dwarf2out. Enable
9723 -gvariable-location-views by default only with locview
9724 assembler support. Enable -ginternal-reset-location-views by
9725 default only if the target defines the corresponding hook.
9726 Enable -ginline-points by default if location views are
9727 enabled; force it disabled if statement frontiers are
9729 * tree-inline.c (expand_call_inline): Check option variables.
9730 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9732 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9734 PR tree-optimization/84321
9735 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9736 handling. Also check whether the anti-range contains any values
9737 that satisfy the mask; switch to a VR_RANGE if not.
9739 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9742 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9744 2018-02-13 Martin Jambor <mjambor@suse.cz>
9747 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9748 of call statements, also set location of a load to a temporary.
9750 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9752 * config/rl78/rl78.c (add_vector_labels): New function.
9753 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9754 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9755 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9756 which checks that no arguments are passed.
9757 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9758 * doc/extend.texi: Documentation for the new attribute.
9760 2018-02-13 Andreas Schwab <schwab@suse.de>
9762 * config/riscv/linux.h (CPP_SPEC): Define.
9764 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9767 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9768 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9769 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9770 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9771 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9772 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9773 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9774 temporarily for AES and PCLMUL builtins.
9776 PR tree-optimization/84339
9777 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9778 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9782 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9783 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9784 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9786 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9788 * omp-simd-clone.h: New file.
9789 * omp-simd-clone.c: Include omp-simd-clone.h.
9790 (expand_simd_clones): No longer static.
9791 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9792 cgraph.h and omp-simd-clone.h.
9793 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9794 (vect_recog_widen_shift_pattern): Formatting fix.
9795 (vect_pattern_recog_1): Don't check optab for calls.
9798 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9799 operands[2] into a REG before using gen_lowpart on it.
9801 2018-02-12 Jeff Law <law@redhat.com>
9804 * config/sh/sh.c (find_barrier): Consider a sibling call
9807 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9808 successfully back substituting a reg.
9810 2018-02-12 Richard Biener <rguenther@suse.de>
9812 PR tree-optimization/84037
9813 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9814 parameter, move visited init to caller.
9815 (vect_slp_analyze_operations): Separate cost from validity
9816 check, initialize visited once for all instances.
9817 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9819 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9821 (vect_model_promotion_demotion_cost): Likewise.
9822 (vectorizable_bswap): Guard cost modeling with !slp_node
9823 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9825 (vectorizable_call): Likewise.
9826 (vectorizable_conversion): Likewise.
9827 (vectorizable_assignment): Likewise.
9828 (vectorizable_shift): Likewise.
9829 (vectorizable_operation): Likewise.
9830 (vectorizable_store): Likewise.
9831 (vectorizable_load): Likewise.
9832 (vectorizable_condition): Likewise.
9833 (vectorizable_comparison): Likewise.
9835 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9838 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9839 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9840 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9842 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9845 * params.def (inline-min-speedup): Increase from 8 to 15.
9846 (max-inline-insns-auto): Decrease from 40 to 30.
9847 * ipa-split.c (consider_split): Add some buffer for function to
9848 be considered inlining candidate.
9849 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9852 2018-02-12 Richard Biener <rguenther@suse.de>
9854 PR tree-optimization/84037
9855 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9856 matched stmts if we cannot swap the non-matched ones.
9858 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9860 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9861 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9862 _mm_maskz_scalef_round_ss): New intrinsics.
9863 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9864 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9865 __builtin_ia32_scalefss_round): Remove.
9866 (__builtin_ia32_scalefsd_mask_round,
9867 __builtin_ia32_scalefss_mask_round): New intrinsics.
9868 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9869 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9870 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9871 "<round_constraint>")): Changed to ...
9872 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9873 "<round_scalar_constraint>")): ... this.
9874 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9875 %0, %1, %2<round_op3>}"): Changed to ...
9876 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9877 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9878 %2<round_scalar_mask_op3>}"): ... this.
9879 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9881 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9883 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9884 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9885 (_mm_maskz_sqrt_round_ss): New intrinsics.
9886 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9887 (__builtin_ia32_sqrtsd_mask_round)
9888 (__builtin_ia32_sqrtss_mask_round): New builtins.
9889 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9890 (__builtin_ia32_sqrtss_round): Remove.
9891 (__builtin_ia32_sqrtsd_mask_round)
9892 (__builtin_ia32_sqrtss_mask_round): New builtins.
9893 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9894 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9895 ((match_operand:VF_128 1 "vector_operand"
9896 "xBm,<round_constraint>")): Changed to ...
9897 ((match_operand:VF_128 1 "vector_operand"
9898 "xBm,<round_scalar_constraint>")): ... this.
9899 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9900 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9901 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9902 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9903 %<iptr>1<round_scalar_mask_op3>}): ... this.
9904 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9905 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9907 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9910 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9911 Cast vec_cmpeq result to correct type.
9912 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9913 Cast vec_cmpgt result to correct type.
9915 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9917 * final.c (final_scan_insn_1): Renamed from...
9918 (final_scan_insn): ... this. New wrapper, to recover
9919 seen from the outermost call in recursive ones.
9920 * config/sparc/sparc.c (output_return): Drop seen from call.
9921 (output_sibcall): Likewise.
9922 * config/visium/visium.c (output_branch): Likewise.
9924 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9926 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
9929 2018-02-10 Alan Modra <amodra@gmail.com>
9932 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
9933 Specify LR as an input.
9935 2018-02-10 Jakub Jelinek <jakub@redhat.com>
9938 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
9939 remove_member_access_dummy_vars): New functions.
9940 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
9941 lower_omp_1, execute_lower_omp): Use them.
9943 PR rtl-optimization/84308
9944 * shrink-wrap.c (spread_components): Release todo vector.
9946 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
9948 PR rtl-optimization/57193
9949 * ira-color.c (struct allocno_color_data): Add member
9950 conflict_allocno_hard_prefs.
9951 (update_conflict_allocno_hard_prefs): New.
9952 (bucket_allocno_compare_func): Add a preference based on
9953 conflict_allocno_hard_prefs.
9954 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
9955 (color_allocnos): Remove a dead code. Initiate
9956 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
9958 2018-02-09 Jakub Jelinek <jakub@redhat.com>
9961 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
9962 constraint from =wa to wa. Avoid a subreg on the output operand,
9963 instead use a pseudo and subreg it in a move.
9964 (p9_xxbrd_<mode>): Changed to ...
9965 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
9966 (p9_xxbrd_v2df): New expander.
9967 (p9_xxbrw_<mode>): Changed to ...
9968 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
9969 (p9_xxbrw_v4sf): New expander.
9971 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9973 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
9975 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
9978 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
9979 multiply in 32-bit mode.
9980 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
9981 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
9984 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
9986 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
9987 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
9988 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
9989 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
9991 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
9994 * dwarf2out.c (is_trivial_indirect_ref): New function.
9995 (dwarf2out_late_global_decl): Do not generate a location
9996 attribute for variables that have a non-trivial DECL_VALUE_EXPR
9997 and that are not defined in the current unit.
9999 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10001 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10002 instead of a libcall for UNORDERED.
10004 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10007 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10008 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10010 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10013 * config/s390/s390.c (s390_set_current_function): Invoke
10014 s390_indirect_branch_settings also if fndecl didn't change.
10016 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10018 * config/rs6000/rs6000.md (blockage): Set length to zero.
10020 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10022 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10024 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10027 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10028 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10032 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10033 PARALLEL incoming that failed vt_get_decl_and_offset check.
10035 PR middle-end/84237
10036 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10037 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10039 (get_variable_section): For decls in named .bss* sections pass true as
10040 second argument to bss_initializer_p.
10042 2018-02-09 Marek Polacek <polacek@redhat.com>
10043 Jakub Jelinek <jakub@redhat.com>
10046 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10047 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10048 Sync some changes from cxx_fold_indirect_ref.
10050 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10052 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10054 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10055 (BLOCK_INLINE_ENTRY_LABEL): New.
10056 (dwarf2out_var_location): Disregard inline entry markers.
10057 (inline_entry_data): New struct.
10058 (inline_entry_data_hasher): New hashtable type.
10059 (inline_entry_data_hasher::hash): New.
10060 (inline_entry_data_hasher::equal): New.
10061 (inline_entry_data_table): New variable.
10062 (add_high_low_attributes): Add DW_AT_entry_pc and
10063 DW_AT_GNU_entry_view attributes if a pending entry is found
10064 in inline_entry_data_table. Add old entry_pc attribute only
10065 if debug nonbinding markers are disabled.
10066 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10067 markers are enabled.
10068 (block_within_block_p, dwarf2out_inline_entry): New.
10069 (dwarf2out_finish): Check that no entries remained in
10070 inline_entry_data_table.
10071 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10072 (final_scan_insn, notice_source_line): Likewise.
10073 (rest_of_clean_state): Skip inline entry markers.
10074 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10076 * gimple.c (gimple_build_debug_inline_entry): New.
10077 * gimple.h (enum gimple_debug_subcode): Add
10078 GIMPLE_DEBUG_INLINE_ENTRY.
10079 (gimple_build_debug_inline_entry): Declare.
10080 (gimple_debug_inline_entry_p): New.
10081 (gimple_debug_nonbind_marker_p): Adjust.
10082 * insn-notes.def (INLINE_ENTRY): New.
10083 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10084 inline entry marker notes.
10085 (print_insn): Likewise.
10086 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10087 (INSN_DEBUG_MARKER_KIND): Likewise.
10088 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10089 * tree-inline.c (expand_call_inline): Build and insert
10090 debug_inline_entry stmt.
10091 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10092 inline entry blocks early, if nonbind markers are enabled.
10093 (dump_scope_block): Dump fragment info.
10094 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10095 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10096 (gimple_build_debug_inline_entry): New.
10097 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10098 Enable/disable inline entry points too.
10099 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10100 (DEBUG_INSN): Describe inline entry markers.
10102 * common.opt (gvariable-location-views): New.
10103 (gvariable-location-views=incompat5): New.
10104 * config.in: Rebuilt.
10105 * configure: Rebuilt.
10106 * configure.ac: Test assembler for view support.
10107 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10108 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10109 * dwarf2out.c (var_loc_view): New typedef.
10110 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10111 (dwarf2out_locviews_in_attribute): New.
10112 (dwarf2out_locviews_in_loclist): New.
10113 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10114 (enum dw_line_info_opcode): Add LI_adv_address.
10115 (struct dw_line_info_table): Add view.
10116 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10117 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10118 (zero_view_p): New variable.
10119 (ZERO_VIEW_P): New macro.
10120 (output_asm_line_debug_info): New.
10121 (struct var_loc_node): Add view.
10122 (add_AT_view_list, AT_loc_list): New.
10123 (add_var_loc_to_decl): Add view param. Test it against last.
10124 (new_loc_list): Add view params. Record them.
10125 (AT_loc_list_ptr): Handle loc and view lists.
10126 (view_list_to_loc_list_val_node): New.
10127 (print_dw_val): Handle dw_val_class_view_list.
10128 (size_of_die): Likewise.
10129 (value_format): Likewise.
10130 (loc_list_has_views): New.
10131 (gen_llsym): Set vl_symbol too.
10132 (maybe_gen_llsym, skip_loc_list_entry): New.
10133 (dwarf2out_maybe_output_loclist_view_pair): New.
10134 (output_loc_list): Output view list or entries too.
10135 (output_view_list_offset): New.
10136 (output_die): Handle dw_val_class_view_list.
10137 (output_dwarf_version): New.
10138 (output_compilation_unit_header): Use it.
10139 (output_skeleton_debug_sections): Likewise.
10140 (output_rnglists, output_line_info): Likewise.
10141 (output_pubnames, output_aranges): Update version comments.
10142 (output_one_line_info_table): Output view numbers in asm comments.
10143 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10144 Call maybe_gen_llsym.
10145 (loc_list_from_tree_1): Adjust.
10146 (add_AT_location_description): Create view list attribute if
10147 needed, check it's absent otherwise.
10148 (convert_cfa_to_fb_loc_list): Adjust.
10149 (maybe_emit_file): Call output_asm_line_debug_info for test.
10150 (dwarf2out_var_location): Reset views as needed. Precompute
10151 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10152 attribute. Set view.
10153 (new_line_info_table): Reset next view.
10154 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10155 (dwarf2out_source_line): Likewise. Output view resets and labels to
10156 the assembler, or select appropriate line info opcodes.
10157 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10158 (optimize_string_length): Catch it. Adjust.
10159 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10160 dw_val_class_view_list, and remove it if no longer needed.
10161 (hash_loc_list): Hash view numbers.
10162 (loc_list_hasher::equal): Compare them.
10163 (optimize_location_lists): Check whether a view list symbol is
10164 needed, and whether the locview attribute is present, and
10165 whether they match. Remove the locview attribute if no longer
10167 (index_location_lists): Call skip_loc_list_entry for test.
10168 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10169 Use output_dwarf_version.
10170 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10171 (struct dw_val_node): Add val_view_list.
10172 * final.c (SEEN_NEXT_VIEW): New.
10173 (set_next_view_needed): New.
10174 (clear_next_view_needed): New.
10175 (maybe_output_next_view): New.
10176 (final_start_function): Rename to...
10177 (final_start_function_1): ... this. Take pointer to FIRST,
10178 add SEEN parameter. Emit param bindings in the initial view.
10179 (final_start_function): Reintroduce SEEN-less interface.
10180 (final): Rename to...
10181 (final_1): ... this. Take SEEN parameter. Output final pending
10182 next view at the end.
10183 (final): Reintroduce seen-less interface.
10184 (final_scan_insn): Output pending next view before switching
10185 sections or ending a block. Mark the next view as needed when
10186 outputting variable locations. Notify debug backend of section
10187 changes, and of location view changes.
10188 (rest_of_handle_final): Adjust.
10189 * toplev.c (process_options): Autodetect value for debug variable
10190 location views option. Warn on incompat5 without -gdwarf-5.
10191 * doc/invoke.texi (gvariable-location-views): New.
10192 (gvariable-location-views=incompat5): New.
10193 (gno-variable-location-views): New.
10195 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10197 PR tree-optimization/84136
10198 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10199 that the result of find_edge is non-NULL.
10201 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10204 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10205 storing integer register in SImode. Fix cost of 256 and 512
10206 byte aligned SSE register store.
10208 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10210 * config/i386/i386.c (ix86_multiplication_cost): Fix
10211 multiplication cost for TARGET_AVX512DQ.
10213 2018-02-08 Marek Polacek <polacek@redhat.com>
10215 PR tree-optimization/84238
10216 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10219 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10221 PR tree-optimization/84265
10222 * tree-vect-stmts.c (vectorizable_store): Don't treat
10223 VMAT_CONTIGUOUS accesses as grouped.
10224 (vectorizable_load): Likewise.
10226 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10228 PR tree-optimization/81635
10229 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10230 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10231 (test_round_for_mask): New functions.
10232 (wide_int_cc_tests): Call test_round_for_mask.
10233 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10234 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10235 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10236 range returned by get_range_info.
10238 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10241 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10242 * symtab.c: Include builtins.h
10243 (symtab_node::output_to_lto_symbol_table_p): Move here
10244 from lto-streamer-out.c:output_symbol_p.
10245 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10246 (output_symbol_p): Move all logic to symtab.c
10247 (produce_symtab): Update.
10249 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10251 * config/s390/s390-opts.h (enum indirect_branch): Define.
10252 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10253 (s390_indirect_branch_via_thunk)
10254 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10255 (enum s390_indirect_branch_type): Define.
10256 * config/s390/s390.c (struct s390_frame_layout, struct
10257 machine_function): Remove.
10258 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10259 (indirect_branch_table_label_no, indirect_branch_table_name):
10261 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10262 (enum s390_indirect_branch_option): Define.
10263 (s390_return_addr_from_memory): New function.
10264 (s390_handle_string_attribute): New function.
10265 (s390_attribute_table): Add new attribute handler.
10266 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10267 (s390_indirect_branch_via_thunk): New function.
10268 (s390_indirect_branch_via_inline_thunk): New function.
10269 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10270 sibling call optimization for non z10 compiles.
10271 (s390_emit_call): Force indirect branch target to be a single
10272 register. Add r1 clobber for non-z10 compiles.
10273 (s390_emit_epilogue): Emit return jump via return_use expander.
10274 (s390_reorg): Handle JUMP_INSNs as execute targets.
10275 (s390_option_override_internal): Perform validity checks for the
10276 new command line options.
10277 (s390_indirect_branch_attrvalue): New function.
10278 (s390_indirect_branch_settings): New function.
10279 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10280 (s390_output_indirect_thunk_function): New function.
10281 (s390_code_end): Implement target hook.
10282 (s390_case_values_threshold): Implement target hook.
10283 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10285 * config/s390/s390.h (struct s390_frame_layout)
10286 (struct machine_function): Move here from s390.c.
10287 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10288 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10289 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10290 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10291 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10292 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10293 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10294 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10295 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10296 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10297 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10298 (mnemonic attribute): Add values which aren't recognized
10300 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10301 pattern for branch conversion. Fix mnemonic attribute.
10302 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10303 indirect branch via thunk if requested.
10304 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10305 ("*indirect_jump"): Disable for branch conversion using out of
10307 ("indirect_jump_via_thunk<mode>_z10")
10308 ("indirect_jump_via_thunk<mode>")
10309 ("indirect_jump_via_inlinethunk<mode>_z10")
10310 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10311 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10312 ("casesi_jump_via_inlinethunk<mode>_z10")
10313 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10314 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10315 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10316 ("*indirect2_jump"): Disable for branch conversion.
10317 ("casesi_jump"): Turn into expander and expand patterns for branch
10319 ("return_use"): New expander.
10320 ("*return"): Emit return via thunk and rename it to ...
10321 ("*return<mode>"): ... this one.
10322 * config/s390/s390.opt: Add new options and and enum for the
10325 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10327 * lra-constraints.c (match_reload): Unconditionally use
10328 gen_lowpart_SUBREG, rather than selecting between that
10329 and equivalent gen_rtx_SUBREG code.
10331 2018-02-08 Richard Biener <rguenther@suse.de>
10333 PR tree-optimization/84233
10334 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10335 changed flag instead of boguously re-using phi_inserted.
10337 2018-02-08 Martin Jambor <mjambor@suse.cz>
10339 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10340 static local variables.
10342 2018-02-08 Richard Biener <rguenther@suse.de>
10344 PR tree-optimization/84278
10345 * tree-vect-stmts.c (vectorizable_store): When looking for
10346 smaller vector types to perform grouped strided loads/stores
10347 make sure the mode is supported by the target.
10348 (vectorizable_load): Likewise.
10350 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10352 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10353 Increase LDP/STP opportunities by adding adjacent callee-saves.
10355 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10357 PR rtl-optimization/84068
10358 PR rtl-optimization/83459
10359 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10361 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10363 PR tree-optimization/84224
10364 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10365 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10366 non-zero arguments.
10368 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10371 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10372 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10373 count, remove one USE.
10375 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10377 * doc/install.texi (Configuration): Document the
10378 --with-long-double-format={ibm,ieee} PowerPC configuration
10382 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10383 Convert from define_expand to be define_insn_and_split. Rework
10384 float/double/_Float128 conversions to QI/HI/SImode to work with
10385 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10386 conversions to QI/HImode types did a store and then a load to
10387 truncate the value. For conversions to VSX registers, don't split
10388 the insn, instead emit the code directly. Use the code iterator
10389 any_fix to combine signed and unsigned conversions.
10390 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10391 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10392 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10393 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10394 (fix_<mode>di2_hw): Likewise.
10395 (fixuns_<mode>di2_hw): Likewise.
10396 (fix_<mode>si2_hw): Likewise.
10397 (fixuns_<mode>si2_hw): Likewise.
10398 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10399 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10400 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10401 fix<uns>_trunc<SFDF:mode>si2_p8.
10402 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10403 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10404 (fix<uns>_<mode>_mem): Likewise.
10405 (fctiw<u>z_<mode>_mem): Likewise.
10406 (fix<uns>_<mode>_mem): Likewise.
10407 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10408 the register allocator from doing a direct move to the GPRs to do
10409 a store, and instead use the ISA 3.0 store byte/half-word from
10410 vector register instruction. For IEEE 128-bit floating point,
10411 also optimize stores of 32-bit ints.
10412 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10414 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10416 * genextract.c (push_pathstr_operand): New function to support
10418 (walk_rtx): Call push_pathstr_operand.
10419 (print_path): Support [a-zA-Z].
10421 2018-02-07 Richard Biener <rguenther@suse.de>
10423 PR tree-optimization/84037
10424 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10425 (cse_and_gimplify_to_preheader): Declare.
10426 (vect_get_place_in_interleaving_chain): Likewise.
10427 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10429 (_loop_vec_info::~_loop_vec_info): Delete it.
10430 (cse_and_gimplify_to_preheader): New function.
10431 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10432 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10433 (vectorizable_load): Likewise. For grouped stores always base
10434 the IV on the first element.
10435 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10436 condition before gimplifying.
10438 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10440 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10441 *DIV_EXPR and *MOD_EXPR.
10443 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10446 * config/i386/i386.c (ix86_option_override_internal): Mask out
10447 the CF_SET bit when checking -fcf-protection.
10449 2018-02-07 Tom de Vries <tom@codesourcery.com>
10452 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10455 2018-02-07 Richard Biener <rguenther@suse.de>
10457 PR tree-optimization/84204
10458 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10461 PR tree-optimization/84205
10462 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10463 special-case isl_ast_op_zdiv_r.
10465 PR tree-optimization/84223
10466 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10467 Only add conditions from within the region.
10468 (gather_bbs::after_dom_children): Adjust.
10470 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10473 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10474 * config/avr/avr.md: Only post-reload split REG-REG moves if
10475 either register is GENERAL_REG_P.
10477 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10479 PR tree-optimization/84235
10480 * tree-ssa-scopedtables.c
10481 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10482 if the subtraction is performed in floating point type where NaNs are
10483 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10484 build 1. Formatting fix.
10486 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10489 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10490 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10491 and skip it regardless of bb boundaries. Use CALL_P macro,
10492 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10493 unnecessarily, formatting fix.
10495 2018-02-06 Michael Collison <michael.collison@arm.com>
10497 * config/arm/thumb2.md:
10498 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10499 (*thumb_mov_notscc): Ditto.
10501 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10504 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10505 unsigned_fix, not "s".
10507 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10509 * configure.ac (gcc_fn_eh_frame_ro): New function.
10510 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10511 correct .eh_frame permissions.
10512 * configure: Regenerate.
10514 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10516 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10517 irrelevant options.
10519 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10521 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10522 Display warning message for -mno-speculate-indirect-jumps.
10524 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10526 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10528 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10530 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10532 PR tree-optimization/84225
10533 * tree-eh.c (find_trapping_overflow): Only call
10534 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10536 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10539 * config/i386/i386.c: Reimplement the check of possible options
10540 -mibt/-mshstk conbination. Change error messages.
10541 * doc/invoke.texi: Fix a typo: remove extra '='.
10543 2018-02-06 Marek Polacek <polacek@redhat.com>
10545 PR tree-optimization/84228
10546 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10548 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10551 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10552 emitted arch directives.
10553 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10554 __ARM_FEATURE_COPROC before changing architectures.
10556 2018-02-06 Richard Biener <rguenther@suse.de>
10558 * config/i386/i386.c (print_reg): Fix typo.
10559 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10561 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10563 * configure: Regenerate.
10565 2018-02-05 Martin Sebor <msebor@redhat.com>
10567 PR tree-optimization/83369
10568 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10571 2018-02-05 Martin Liska <mliska@suse.cz>
10573 * doc/invoke.texi: Cherry-pick upstream r323995.
10575 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10577 * ira.c (ira_init_register_move_cost): Adjust comment.
10579 2018-02-05 Martin Liska <mliska@suse.cz>
10581 PR gcov-profile/84137
10582 * doc/gcov.texi: Fix typo in documentation.
10584 2018-02-05 Martin Liska <mliska@suse.cz>
10586 PR gcov-profile/83879
10587 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10588 using dlopen functionality.
10590 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10592 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10593 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10594 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10595 _mm_maskz_range_round_ss): New intrinsics.
10596 (__builtin_ia32_rangesd128_round)
10597 (__builtin_ia32_rangess128_round): Remove.
10598 (__builtin_ia32_rangesd128_mask_round,
10599 __builtin_ia32_rangess128_mask_round): New builtins.
10600 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10601 __builtin_ia32_rangess128_round): Remove.
10602 (__builtin_ia32_rangesd128_mask_round,
10603 __builtin_ia32_rangess128_mask_round): New builtins.
10604 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10605 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10606 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10607 "<round_saeonly_constraint>")): Changed to ...
10608 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10609 "<round_saeonly_scalar_constraint>")): ... this.
10610 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10611 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10612 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10613 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10614 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10616 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10618 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10620 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10621 Remove all values except native, 8540 and 8548.
10623 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10625 * config/i386/i386.c (ix86_output_function_return): Pass
10626 INVALID_REGNUM, instead of -1, as invalid register number to
10627 indirect_thunk_name and output_indirect_thunk.
10629 2018-02-02 Julia Koval <julia.koval@intel.com>
10631 * config.gcc: Add -march=icelake.
10632 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10633 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10634 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10635 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10636 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10637 (processor_target_table): Add icelake.
10638 (ix86_option_override_internal): Handle new PTAs.
10639 (get_builtin_code_for_version): Handle icelake.
10640 (M_INTEL_COREI7_ICELAKE): New.
10641 (fold_builtin_cpu): Handle icelake.
10642 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10643 * doc/invoke.texi: Add -march=icelake.
10645 2018-02-02 Julia Koval <julia.koval@intel.com>
10647 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10648 to wide_int_bitmask.
10649 * wide-int-bitmask.h: New.
10651 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10654 * config/i386/i386.md: Replace Pmode with word_mode in
10655 builtin_setjmp_setup and builtin_longjmp to support x32.
10657 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10661 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10663 (rs6000_supported_cpu_names): New static variable.
10664 (linux_cpu_translation_table): Likewise.
10665 (elf_platform) <cpu>: Define new static variable and use it.
10666 Translate kernel AT_PLATFORM name to canonical name if needed.
10667 Error if platform name is unknown.
10669 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10672 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10674 2018-02-01 Jeff Law <law@redhat.com>
10677 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10678 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10679 the scratch if RELEASE_VIA_POP is false.
10680 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10681 If we have to save a temporary register, decrement SIZE appropriately.
10682 Pass new arguments to release_scratch_register_on_entry.
10683 (ix86_adjust_stack_and_probe): Likewise.
10684 (ix86_emit_probe_stack_range): Pass new arguments to
10685 release_scratch_register_on_entry.
10687 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10689 PR rtl-optimization/84157
10690 * combine.c (change_zero_ext): Use REG_P predicate in
10691 front of HARD_REGISTER_P predicate.
10693 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10695 * config/avr/avr.c (avr_option_override): Move disabling of
10696 -fdelete-null-pointer-checks to...
10697 * common/config/avr/avr-common.c (avr_option_optimization_table):
10700 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10702 PR tree-optimization/81635
10703 * tree-data-ref.c (split_constant_offset_1): For types that
10704 wrap on overflow, try to use range info to prove that wrapping
10707 2018-02-01 Renlin Li <renlin.li@arm.com>
10710 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10711 TAILCALL_ADDR_REGS.
10712 (aarch64_register_move_cost): Likewise.
10713 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10714 TAILCALL_ADDR_REGS.
10715 (REG_CLASS_NAMES): Likewise.
10716 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10717 TAILCALL_ADDR_REGS. Remove IP registers.
10718 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10720 2018-02-01 Richard Biener <rguenther@suse.de>
10722 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10723 for specifying RPO order and allow NULL for that.
10724 * domwalk.c (dom_walker::dom_walker): Likewise.
10725 (dom_walker::walk): Handle NULL RPO order.
10726 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10728 (rewrite_update_dom_walker): Likewise.
10729 (mark_def_dom_walker): Likewise.
10731 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10733 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10734 (aarch64_maybe_expand_sve_subreg_move): Declare.
10735 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10736 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10738 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10739 that are semantically a reverse operation.
10740 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10741 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10742 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10744 (aarch64_can_change_mode_class): For big-endian, forbid changes
10745 between two SVE modes if they have different element sizes.
10747 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10749 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10750 the TImode handling for big-endian targets.
10752 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10754 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10755 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10757 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10758 Remove BSWAP handing for big-endian targets and use the form of
10759 LD1RQ appropariate for the mode.
10761 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10763 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10764 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10765 duplicated element.
10767 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10770 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10771 check for operands that need to go through aarch64_sve_reload_be.
10773 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10775 PR tree-optimization/81661
10776 PR tree-optimization/84117
10777 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10778 * tree-eh.c: Include gimplify.h.
10779 (find_trapping_overflow, replace_trapping_overflow,
10780 rewrite_to_non_trapping_overflow): New functions.
10781 * tree-vect-loop.c: Include tree-eh.h.
10782 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10783 * tree-data-ref.c: Include tree-eh.h.
10784 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10786 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10788 PR rtl-optimization/84123
10789 * combine.c (change_zero_ext): Check if hard register satisfies
10790 can_change_dest_mode before calling gen_lowpart_SUBREG.
10792 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10795 * ira.c (ira_init_register_move_cost): Remove assert.
10797 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10799 PR rtl-optimization/84071
10800 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10801 * doc/tm.texi: Regenerate.
10803 2018-01-31 Richard Biener <rguenther@suse.de>
10805 PR tree-optimization/84132
10806 * tree-data-ref.c (analyze_miv_subscript): Properly
10807 check whether evolution_function_is_affine_multivariate_p
10808 before calling gcd_of_steps_may_divide_p.
10810 2018-01-31 Julia Koval <julia.koval@intel.com>
10813 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10814 * config/i386/i386.md (rdpid_rex64) New.
10815 (rdpid): Make 32bit only.
10817 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10820 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10821 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10823 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10826 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10828 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10830 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10832 PR rtl-optimization/84071
10833 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10834 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10836 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10838 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10839 (arc_attribute_table): Add 'aux' attribute.
10840 (arc_in_small_data_p): Consider aux like variables.
10841 (arc_is_aux_reg_p): New function.
10842 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10843 (arc_get_aux_arg): New function.
10844 (prepare_move_operands): Handle aux-register access.
10845 (arc_handle_aux_attribute): New function.
10846 * doc/extend.texi (ARC Variable attributes): Add subsection.
10848 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10850 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10851 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10852 (arc_attribute_table): Add 'uncached' attribute.
10853 (arc_print_operand): Print '.di' flag for uncached memory
10855 (arc_in_small_data_p): Do not consider for small data the uncached
10857 (arc_is_uncached_mem_p): New function.
10858 * config/arc/predicates.md (compact_store_memory_operand): Check
10859 for uncached memory accesses.
10860 (nonvol_nonimm_operand): Likewise.
10861 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10863 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10866 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10867 falign-loops=): Add Optimization flag.
10869 2018-01-30 Jeff Law <law@redhat.com>
10872 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10873 INT_REGISTERS_SAVED. Check it prior to calling
10874 get_scratch_register_on_entry.
10875 (ix86_adjust_stack_and_probe): Similarly.
10876 (ix86_emit_probe_stack_range): Similarly.
10877 (ix86_expand_prologue): Corresponding changes.
10879 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10882 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10883 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10885 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10888 * lra-constraints.c (curr_insn_transform): Process AND in the
10891 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10893 PR rtl-optimization/83986
10894 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10895 dependence against last_pending_memory_flush in addition to
10896 pending_jump_insns.
10898 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10900 PR tree-optimization/81611
10901 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10904 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10907 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10910 2018-01-30 Richard Biener <rguenther@suse.de>
10911 Jakub Jelinek <jakub@redhat.com>
10913 PR tree-optimization/84111
10914 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10915 inner loops added during recursion, as they don't have up-to-date
10918 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10921 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10922 (can_inline_edge_by_limits_p): ... here.
10923 (can_early_inline_edge_p, check_callers,
10924 update_caller_keys, update_callee_keys, recursive_inlining,
10925 add_new_edges_to_heap, speculation_useful_p,
10926 inline_small_functions,
10927 inline_small_functions, flatten_function,
10928 inline_to_all_callers_1): Update.
10930 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10932 * profile-count.c (profile_count::combine_with_ipa_count): Handle
10935 2018-01-30 Richard Biener <rguenther@suse.de>
10937 PR tree-optimization/83008
10938 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
10939 invariant and constant vector uses in stmts when they need
10940 more than one stmt.
10942 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10945 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
10946 * configure: Regenerate.
10948 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10950 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
10952 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
10953 Use gen_rtx_REG rather than gen_lowpart.
10955 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10957 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
10958 rather than 0 when creating partial subregs.
10960 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
10962 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
10965 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
10968 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
10969 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
10970 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
10971 flags. This restores the settings used before the 2017-07-24.
10972 Turning off pre increment/decrement/modify allows IVOPTS to
10973 optimize DF/SF loops where the index is an int.
10975 2018-01-29 Richard Biener <rguenther@suse.de>
10976 Kelvin Nilsen <kelvin@gcc.gnu.org>
10979 * tree-vect-stmts.c (vectorizable_call): Don't call
10980 targetm.vectorize_builtin_md_vectorized_function if callee is
10983 2018-01-22 Carl Love <cel@us.ibm.com>
10985 * doc/extend.tex: Fix typo in second arg in
10986 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
10988 2018-01-29 Richard Biener <rguenther@suse.de>
10990 PR tree-optimization/84086
10991 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
10992 (flush_ssaname_freelist): When SSA names were released reset
10993 the SCEV hash table.
10995 2018-01-29 Richard Biener <rguenther@suse.de>
10997 PR tree-optimization/84057
10998 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
10999 removed paths when removing edges.
11001 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11003 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11004 -mfunction-return=@var{choice}.
11006 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11008 PR diagnostic/84034
11009 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11010 Handle CR like TAB.
11011 (layout::print_source_line): Likewise.
11012 (test_get_line_width_without_trailing_whitespace): Add test cases.
11014 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11016 PR middle-end/84040
11017 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11020 2018-01-26 Jim Wilson <jimw@sifive.com>
11022 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11024 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11027 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11029 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11030 and CMP + SUB-immediate -> SUBS.
11032 2018-01-26 Martin Sebor <msebor@redhat.com>
11034 PR tree-optimization/83896
11035 * tree-ssa-strlen.c (get_string_len): Rename...
11036 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11037 Avoid assuming length is constant.
11038 (handle_char_store): Use HOST_WIDE_INT for string length.
11040 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11043 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11044 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11046 2018-01-26 Richard Biener <rguenther@suse.de>
11048 PR rtl-optimization/84003
11049 * dse.c (record_store): Only record redundant stores when
11050 the earlier store aliases at least all accesses the later one does.
11052 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11054 PR rtl-optimization/83985
11055 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11056 REG_CFA_RESTORE insns.
11057 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11058 REG_CFA_RESTORE insns here.
11061 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11062 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11064 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11066 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11067 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11068 (arc_init): Likewise.
11069 (arc_override_options): Likewise.
11070 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11072 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11074 * config/arc/arc.h (TARGET_DBNZ): Define.
11075 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11076 properly set the tune attribute.
11077 (dbnz): Use TARGET_DBNZ guard.
11078 * config/arc/arc.opt (mtune): Add core3 option.
11080 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11082 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11083 recognize new pic like addresses.
11084 (arc_delegitimize_address): Clean up.
11086 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11088 * config/arc/arc-arches.def: Option mrf16 valid for all
11090 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11091 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11092 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11093 * config/arc/arc-tables.opt: Regenerate.
11094 * config/arc/arc.c (arc_conditional_register_usage): Handle
11095 reduced register file case.
11096 (arc_file_start): Set must have build attributes.
11097 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11098 mrf16 option value.
11099 * config/arc/arc.opt (mrf16): Add new option.
11100 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11101 * config/arc/genmultilib.awk: Handle new mrf16 option.
11102 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11103 * config/arc/t-multilib: Regenerate.
11104 * doc/invoke.texi (ARC Options): Document mrf16 option.
11106 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11108 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11109 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11110 (arc_attribute_table): Add 'secure_call' attribute.
11111 (arc_print_operand): Print secure call operand.
11112 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11114 (arc_is_secure_call_p): New function. * config/arc/arc.md
11115 (call_i): Add support for sjli instruction.
11116 (call_value_i): Likewise.
11117 * config/arc/constraints.md (Csc): New constraint.
11119 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11120 John Eric Martin <John.Martin@emmicro-us.com>
11122 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11123 * config/arc/arc.c (_arc_jli_section): New struct.
11124 (arc_jli_section): New type.
11125 (rc_jli_sections): New static variable.
11126 (arc_handle_jli_attribute): New function.
11127 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11128 (arc_file_end): New function.
11129 (TARGET_ASM_FILE_END): Define.
11130 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11131 (arc_add_jli_section): New function.
11132 (jli_call_scan): Likewise.
11133 (arc_reorg): Call jli_call_scan.
11134 (arc_output_addsi): Remove 'S' from printing asm operand.
11135 (arc_is_jli_call_p): New function.
11136 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11138 (movhi_insn): Likewise.
11139 (movsi_insn): Likewise.
11140 (movsi_set_cc_insn): Likewise.
11141 (loadqi_update): Likewise.
11142 (load_zeroextendqisi_update): Likewise.
11143 (load_signextendqisi_update): Likewise.
11144 (loadhi_update): Likewise.
11145 (load_zeroextendhisi_update): Likewise.
11146 (load_signextendhisi_update): Likewise.
11147 (loadsi_update): Likewise.
11148 (loadsf_update): Likewise.
11149 (movsicc_insn): Likewise.
11150 (bset_insn): Likewise.
11151 (bxor_insn): Likewise.
11152 (bclr_insn): Likewise.
11153 (bmsk_insn): Likewise.
11154 (bicsi3_insn): Likewise.
11155 (cmpsi_cc_c_insn): Likewise.
11156 (movsi_ne): Likewise.
11157 (movsi_cond_exec): Likewise.
11158 (clrsbsi2): Likewise.
11159 (norm_f): Likewise.
11169 (call_i): Remove 'S' asm letter, add jli instruction.
11170 (call_value_i): Likewise.
11171 * config/arc/arc.op (mjli-always): New option.
11172 * config/arc/constraints.md (Cji): New constraint.
11173 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11175 (subsf3_fpx): Likewise.
11176 (mulsf3_fpx): Likewise.
11177 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11179 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11180 function attrbutes.
11181 * doc/invoke.texi (ARC): Document mjli-always option.
11183 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11185 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11186 avoid addition with 0 and use incw and decw where possible.
11188 2018-01-26 Richard Biener <rguenther@suse.de>
11190 PR tree-optimization/81082
11191 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11192 association if it requires casting to unsigned.
11193 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11194 from fold_plusminus_mult_expr to catch important cases late when
11195 range info is available.
11197 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11199 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11200 * configure.ac (hidden_linkonce): New test.
11201 * configure: Regenerate.
11202 * config.in: Regenerate.
11204 2018-01-26 Julia Koval <julia.koval@intel.com>
11206 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11207 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11208 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11209 _mm_mask_bitshuffle_epi64_mask): Fix type.
11210 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11211 USI_FTYPE_V4DI_V4DI_USI): Remove.
11212 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11213 __builtin_ia32_vpshufbitqmb256_mask,
11214 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11215 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11216 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11218 2018-01-26 Alan Modra <amodra@gmail.com>
11221 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11222 UNSPEC_VBPERMQ. Sort other unspecs.
11224 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11226 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11228 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11230 PR middle-end/83055
11231 * predict.c (drop_profile): Do not push/pop cfun; update also
11233 (handle_missing_profiles): Fix logic looking for zero profiles.
11235 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11237 PR middle-end/83977
11238 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11239 on functions with #pragma omp declare simd or functions with simd
11241 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11242 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11243 Remove trailing \n from warning_at calls.
11245 2018-01-25 Tom de Vries <tom@codesourcery.com>
11248 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11249 for neutered workers.
11251 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11254 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11255 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11257 2018-01-24 Jeff Law <law@redhat.com>
11260 * i386.c (get_probe_interval): Move to earlier point.
11261 (ix86_compute_frame_layout): If -fstack-clash-protection and
11262 the frame is larger than the probe interval, then use pushes
11263 to save registers rather than reg->mem moves.
11264 (ix86_expand_prologue): Remove conditional for int_registers_saved
11267 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11270 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11271 min/max for never referenced object.
11273 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11275 PR middle-end/83977
11276 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11278 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11279 attributes from DECL_ATTRIBUTES (decl) without affecting
11280 DECL_ATTRIBUTES (current_function_decl).
11281 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11282 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11284 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11286 PR tree-optimization/83979
11287 * fold-const.c (fold_comparison): Use constant_boolean_node
11288 instead of boolean_{true,false}_node.
11290 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11292 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11295 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11297 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11298 Simplify the clause that sets the length attribute.
11299 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11300 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11301 clause that sets the length attribute.
11302 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11304 2018-01-24 Tom de Vries <tom@codesourcery.com>
11307 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11308 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11309 Add strict parameter.
11310 (prevent_branch_around_nothing): Insert dummy insn between branch to
11311 label and label with no ptx insn inbetween.
11312 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11314 2018-01-24 Tom de Vries <tom@codesourcery.com>
11317 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11318 for neutered threads in warp.
11319 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11321 2018-01-24 Richard Biener <rguenther@suse.de>
11323 PR tree-optimization/83176
11324 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11327 2018-01-24 Richard Biener <rguenther@suse.de>
11329 PR tree-optimization/82819
11330 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11331 code generating pluses that are no-ops in the target precision.
11333 2018-01-24 Richard Biener <rguenther@suse.de>
11335 PR middle-end/84000
11336 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11338 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11340 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11341 to merge probabilities.
11342 * predict.c (probably_never_executed): Also mark as cold functions
11343 with global 0 profile and guessed local profile.
11344 * profile-count.c (profile_probability::combine_with_count): New
11346 * profile-count.h (profile_probability::operator*,
11347 profile_probability::operator*=, profile_probability::operator/,
11348 profile_probability::operator/=): Reduce precision to adjusted
11349 and set value to guessed on contradictory divisions.
11350 (profile_probability::combine_with_freq): Remove.
11351 (profile_probability::combine_wiht_count): Declare.
11352 (profile_count::force_nonzero):: Set to adjusted.
11353 (profile_count::probability_in):: Set quality to adjusted.
11354 * tree-ssa-tail-merge.c (replace_block_by): Use
11355 combine_with_count.
11357 2018-01-23 Andrew Waterman <andrew@sifive.com>
11358 Jim Wilson <jimw@sifive.com>
11360 * config/riscv/riscv.c (riscv_stack_boundary): New.
11361 (riscv_option_override): Set riscv_stack_boundary. Handle
11362 riscv_preferred_stack_boundary_arg.
11363 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11364 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11365 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11366 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11367 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11368 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11370 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11373 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11374 of struct ix86_frame.
11375 (ix86_expand_epilogue): Likewise. Add a local variable for
11376 the reg_save_offset field in struct ix86_frame.
11378 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11380 PR tree-optimization/82604
11381 * tree-loop-distribution.c (enum partition_kind): New enum item
11382 PKIND_PARTIAL_MEMSET.
11383 (partition_builtin_p): Support above new enum item.
11384 (generate_code_for_partition): Ditto.
11385 (compute_access_range): Differentiate cases that equality can be
11386 proven at all loops, the innermost loops or no loops.
11387 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11388 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11389 (finalize_partitions, distribute_loop): Don't fuse partition of
11390 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11391 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11392 parloop is enabled.
11394 2018-01-23 Martin Liska <mliska@suse.cz>
11396 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11397 order to ignore the predictor.
11398 (PRED_POLYMORPHIC_CALL): Likewise.
11399 (PRED_RECURSIVE_CALL): Likewise.
11401 2018-01-23 Martin Liska <mliska@suse.cz>
11403 * tree-profile.c (tree_profiling): Print function header to
11404 aware reader which function we are working on.
11405 * value-prof.c (gimple_find_values_to_profile): Do not print
11406 not interesting value histograms.
11408 2018-01-23 Martin Liska <mliska@suse.cz>
11410 * profile-count.h (enum profile_quality): Add
11411 profile_uninitialized as the first value. Do not number values
11412 as they are zero based.
11413 (profile_count::verify): Update sanity check.
11414 (profile_probability::verify): Likewise.
11416 2018-01-23 Nathan Sidwell <nathan@acm.org>
11418 * doc/invoke.texi (ffor-scope): Deprecate.
11420 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11422 PR tree-optimization/83510
11423 * domwalk.c (set_all_edges_as_executable): New function.
11424 (dom_walker::dom_walker): Convert bool param
11425 "skip_unreachable_blocks" to enum reachability. Move setup of
11426 edge flags to set_all_edges_as_executable and only do it when
11427 reachability is REACHABLE_BLOCKS.
11428 * domwalk.h (enum dom_walker::reachability): New enum.
11429 (dom_walker::dom_walker): Convert bool param
11430 "skip_unreachable_blocks" to enum reachability.
11431 (set_all_edges_as_executable): New decl.
11432 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11433 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11435 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11436 but converting true to REACHABLE_BLOCKS.
11437 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11439 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11440 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11441 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11443 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11444 if check_all_array_refs will be called.
11446 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11448 * tree.c (selftest::test_location_wrappers): Add more test
11451 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11453 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11454 (selftest::test_bit_in_range): Likewise.
11456 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11459 * doc/sourcebuild.texi (vect_float): Say that the selector
11460 only describes the situation when -funsafe-math-optimizations is on.
11461 (vect_float_strict): Document.
11463 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11465 PR tree-optimization/83965
11466 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11467 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11468 instead of checking only for a reduction.
11469 (vect_recog_widen_sum_pattern): Likewise.
11471 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11473 * predict.c (probably_never_executed): Only use precise profile info.
11474 (compute_function_frequency): Skip after inlining hack since we now
11475 have quality checking.
11477 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11479 * profile-count.h (profile_probability::very_unlikely,
11480 profile_probability::unlikely, profile_probability::even): Set
11481 precision to guessed.
11483 2018-01-23 Richard Biener <rguenther@suse.de>
11485 PR tree-optimization/83963
11486 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11487 Properly terminate dominator walk when crossing the exit edge not
11488 when visiting its source block.
11490 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11493 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11494 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11496 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11498 PR tree-optimization/83957
11499 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11500 semicolon after for body surrounded by braces.
11502 PR tree-optimization/83081
11503 * profile-count.h (profile_probability::split): New method.
11504 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11505 Use profile_probability::split.
11506 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11507 when splitting a single conditional jump into 2.
11509 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11511 PR tree-optimization/69452
11512 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11515 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11517 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11518 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11519 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11521 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11523 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11525 * config/rl78/rl78.md (movdi): New define_expand.
11526 * config/rl78/rl78.c (rl78_split_movdi): New function.
11528 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11531 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11533 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11534 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11535 128-bit to produce an UNSPEC move to get the double word with the
11536 signbit and then a shift directly to do signbit.
11537 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11538 implementation with a new version that just does either a direct
11539 move or a regular move. Move memory interface to separate insns.
11540 Move insns so they are next to the expander.
11541 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11542 with signbit move. Split big and little endian case.
11543 (signbit<mode>2_dm_mem_le): Likewise.
11544 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11545 (signbit<mode>2_dm2): Likewise.
11547 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11549 * config/rl78/rl78.md (anddi3): New define_expand.
11551 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11553 * config/rl78/rl78.md (umindi3): New define_expand.
11555 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11557 * config/rl78/rl78.md (smindi3): New define_expand.
11559 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11561 * config/rl78/rl78.md (smaxdi3): New define_expand.
11563 2018-01-22 Carl Love <cel@us.ibm.com>
11565 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11566 LVX_V1TI): Add macro expansion.
11567 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11568 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11569 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11570 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11571 Change check to determine if the instruction is a byte reversing
11572 entry. Fix typo in comment.
11573 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11574 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11575 Add def_builtin calls for new builtins.
11576 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11577 Add define_insn expansion.
11579 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11581 * config/rl78/rl78.md (umaxdi3): New define_expand.
11583 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11585 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11586 for non-QImode registers.
11588 2018-01-22 Richard Biener <rguenther@suse.de>
11590 PR tree-optimization/83963
11591 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11592 including the loop exit block.
11593 (scop_detection::merge_sese): Likewise.
11594 (scop_detection::add_scop): Do it here instead.
11596 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11598 * doc/sourcebuild.texi (arm_softfloat): Document.
11600 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11603 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11604 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11605 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11607 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11608 David Edelsohn <dje.gcc@gmail.com>
11611 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11612 Change "crset eq" to "crset 2".
11613 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11614 (*call_indirect_aix<mode>_nospec): Likewise.
11615 (*call_value_indirect_aix<mode>_nospec): Likewise.
11616 (*call_indirect_elfv2<mode>_nospec): Likewise.
11617 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11618 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11619 change assembly output from . to $.
11620 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11621 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11622 (*tablejump<mode>_internal1_nospec): Likewise.
11624 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11627 * config/sh/sh_optimize_sett_clrt.cc:
11628 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11630 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11632 PR tree-optimization/83940
11633 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11634 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11635 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11636 use it to pass back the definition type.
11637 (vect_check_store_rhs): Likewise rhs_dt_out.
11638 (vect_build_gather_load_calls): Add a mask_dt argument and use
11639 it instead of a call to vect_is_simple_use.
11640 (vectorizable_store): Update calls to vect_check_load_store_mask
11641 and vect_check_store_rhs. Use the dt returned by the latter instead
11642 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11643 instead of calls to vect_is_simple_use. Pass the scalar rather
11644 than the vector operand to vect_is_simple_use when handling
11645 second and subsequent copies of an rhs value.
11646 (vectorizable_load): Update calls to vect_check_load_store_mask
11647 and vect_build_gather_load_calls. Use the cached mask_dt and
11648 gs_info.offset_dt instead of calls to vect_is_simple_use.
11650 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11652 PR middle-end/83945
11653 * tree-emutls.c: Include gimplify.h.
11654 (lower_emutls_2): New function.
11655 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11656 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11657 it before further processing.
11660 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11661 UINTVAL (trueop1) instead of INTVAL (op1).
11663 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11667 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11668 INCOMING_FRAME_SP_OFFSET if not defined.
11669 (scan_trace): Add ENTRY argument. If true and
11670 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11671 emit a note to adjust the CFA offset.
11672 (create_cfi_notes): Adjust scan_trace callers.
11673 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11674 INCOMING_FRAME_SP_OFFSET in the CIE.
11675 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11676 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11678 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11679 * doc/tm.texi: Regenerated.
11681 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11683 PR rtl-optimization/83147
11684 * lra-constraints.c (remove_inheritance_pseudos): Use
11685 lra_substitute_pseudo_within_insn.
11687 2018-01-19 Tom de Vries <tom@codesourcery.com>
11688 Cesar Philippidis <cesar@codesourcery.com>
11691 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11693 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11696 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11697 spaces for function labels.
11699 2018-01-19 Martin Liska <mliska@suse.cz>
11701 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11702 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11703 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11704 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11705 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11706 (PRED_CONST_RETURN): Change from 69 to 65.
11707 (PRED_NULL_RETURN): Change from 91 to 71.
11708 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11709 (PRED_LOOP_GUARD): Change from 66 to 73.
11711 2018-01-19 Martin Liska <mliska@suse.cz>
11713 * predict.c (predict_insn_def): Add new assert.
11714 (struct branch_predictor): Change type to signed integer.
11715 (test_prediction_value_range): Amend test to cover
11716 PROB_UNINITIALIZED.
11717 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11718 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11719 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11720 (PRED_LOOP_IV_COMPARE): Likewise.
11721 * predict.h (PROB_UNINITIALIZED): Define new constant.
11723 2018-01-19 Martin Liska <mliska@suse.cz>
11725 * predict.c (dump_prediction): Add new format for
11726 analyze_brprob.py script which is enabled with -details
11728 * profile-count.h (precise_p): New function.
11730 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11732 PR tree-optimization/83922
11733 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11734 there are no statements that need masking.
11735 (vect_active_double_reduction_p): New function.
11736 (vect_analyze_loop_operations): Use it when handling phis that
11737 are not in the loop header.
11739 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11741 PR tree-optimization/83914
11742 * tree-vect-loop.c (vectorizable_induction): Don't convert
11743 init_expr or apply the peeling adjustment for inductions
11744 that are nested within the vectorized loop.
11746 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11748 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11751 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11755 * function.h (gimplify_parameters): Add gimple_seq * argument.
11756 * function.c: Include gimple.h and options.h.
11757 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11758 for the added local temporaries if needed.
11759 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11760 if there are any parameter cleanups, wrap whole body into a
11761 try/finally with the cleanups.
11763 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11766 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11767 Use GET_MODE_CLASS for scalar floating point.
11769 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11773 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11774 Fix call of call_cgraph_insertion_hooks.
11776 2018-01-18 Martin Sebor <msebor@redhat.com>
11778 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11780 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11783 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11786 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11789 * common.opt: (-ffile-prefix-map): New option.
11790 * opts.c (common_handle_option): Defer it.
11791 * opts-global.c (handle_common_deferred_options): Handle it.
11792 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11793 * file-prefix-map.h: New file.
11794 (remap_debug_filename, add_debug_prefix_map): ...here.
11795 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11796 * final.c (debug_prefix_map, add_debug_prefix_map
11797 remap_debug_filename): Move to...
11798 * file-prefix-map.c: New file.
11799 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11800 generalize, get rid of alloca(), use strrchr() instead of strchr().
11801 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11802 Implement in terms of add_prefix_map().
11803 (remap_macro_filename, remap_debug_filename): Implement in term of
11805 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11806 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11807 * dbxout.c: Include file-prefix-map.h.
11808 * varasm.c: Likewise.
11809 * vmsdbgout.c: Likewise.
11810 * xcoffout.c: Likewise.
11811 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11812 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11813 * doc/invoke.texi (-ffile-prefix-map): Document.
11814 (-fdebug-prefix-map): Update description.
11816 2018-01-18 Martin Liska <mliska@suse.cz>
11818 * config/i386/i386.c (indirect_thunk_name): Document that also
11820 (output_indirect_thunk): Document why both instructions
11821 (pause and lfence) are generated.
11823 2018-01-18 Richard Biener <rguenther@suse.de>
11825 PR tree-optimization/83887
11826 * graphite-scop-detection.c
11827 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11828 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11829 (scop_detection::merge_sese): Re-implement with a flood-fill
11830 algorithm that properly finds a SESE region if it exists.
11832 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11835 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11836 pointer_diff optimizations use view_convert instead of convert.
11838 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11840 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11841 Generate different code for -mno-speculate-indirect-jumps.
11842 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11843 (*call_indirect_aix<mode>): Disable for
11844 -mno-speculate-indirect-jumps.
11845 (*call_indirect_aix<mode>_nospec): New define_insn.
11846 (*call_value_indirect_aix<mode>): Disable for
11847 -mno-speculate-indirect-jumps.
11848 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11849 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11850 -mno-speculate-indirect-jumps.
11851 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11853 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11855 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11856 long double type, set the flags for noting the default long double
11857 type, even if we don't pass or return a long double type.
11859 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11862 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11865 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11867 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11868 support for merge[hl].
11869 (fold_mergehl_helper): New helper function.
11870 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11871 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11872 (altivec_vmrglw_direct): Add xxmrglw insn.
11874 2018-01-17 Andrew Waterman <andrew@sifive.com>
11876 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11877 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11879 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11882 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11883 call the lto_location_cache before reading the
11884 DECL_SOURCE_LOCATION of the types.
11886 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11887 Richard Sandiford <richard.sandiford@linaro.org>
11889 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11890 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11891 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11892 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11893 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11895 * config/aarch64/constraints.md (aarch64_movti_operand):
11897 * config/aarch64/predicates.md (Uti): Add new constraint.
11899 2018-01-17 Carl Love <cel@us.ibm.com>
11901 * config/rs6000/vsx.md (define_expand xl_len_r,
11902 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11903 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11905 (define_expand, define_insn): Move the shift left from the
11906 define_insn to the define_expand for lxvl and stxvl instructions.
11907 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11908 and XL_LEN_R definitions to PURE.
11910 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11912 * config/i386/i386.c (indirect_thunk_name): Declare regno
11913 as unsigned int. Compare regno with INVALID_REGNUM.
11914 (output_indirect_thunk): Ditto.
11915 (output_indirect_thunk_function): Ditto.
11916 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11917 in the call to output_indirect_thunk_function.
11919 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11921 PR middle-end/83884
11922 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11923 rather than the size of inner_type to determine the stack slot size
11924 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11926 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
11929 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
11932 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
11934 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
11935 endian Linux systems to optionally enable multilibs for selecting
11936 the long double type if the user configured an explicit type.
11937 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
11938 have no long double multilibs if not defined.
11939 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
11940 warn if the user used -mabi={ieee,ibm}longdouble and we built
11941 multilibs for long double.
11942 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
11943 appropriate multilib option.
11944 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
11946 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
11947 for building long double multilibs.
11948 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
11950 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
11952 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
11955 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
11957 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
11960 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
11963 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
11966 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11968 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
11969 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
11971 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
11973 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
11974 different rtl trees depending on TARGET_64BIT.
11975 (rs6000_gen_lvx): Likewise.
11977 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
11979 * config/visium/visium.md (nop): Tweak comment.
11980 (hazard_nop): Likewise.
11982 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11984 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
11985 -mspeculate-indirect-jumps.
11986 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
11987 for -mno-speculate-indirect-jumps.
11988 (*call_indirect_elfv2<mode>_nospec): New define_insn.
11989 (*call_value_indirect_elfv2<mode>): Disable for
11990 -mno-speculate-indirect-jumps.
11991 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
11992 (indirect_jump): Emit different RTL for
11993 -mno-speculate-indirect-jumps.
11994 (*indirect_jump<mode>): Disable for
11995 -mno-speculate-indirect-jumps.
11996 (*indirect_jump<mode>_nospec): New define_insn.
11997 (tablejump): Emit different RTL for
11998 -mno-speculate-indirect-jumps.
11999 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12000 (tablejumpsi_nospec): New define_expand.
12001 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12002 (tablejumpdi_nospec): New define_expand.
12003 (*tablejump<mode>_internal1): Disable for
12004 -mno-speculate-indirect-jumps.
12005 (*tablejump<mode>_internal1_nospec): New define_insn.
12006 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12009 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12011 * caller-save.c (insert_save): Drop unnecessary parameter. All
12014 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12015 Richard Biener <rguenth@suse.de>
12018 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12019 return early, inline manually is_gimple_sizepos. Make sure if we
12020 call gimplify_expr we don't end up with a gimple constant.
12021 * tree.c (variably_modified_type_p): Don't return true for
12022 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12023 * gimplify.h (is_gimple_sizepos): Remove.
12025 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12027 PR tree-optimization/83857
12028 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12029 vectorizable_live_operation for pure SLP statements.
12030 (vectorizable_live_operation): Handle PHIs.
12032 2018-01-16 Richard Biener <rguenther@suse.de>
12034 PR tree-optimization/83867
12035 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12036 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12038 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12041 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12042 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12043 If off is not INTEGER_CST, issue a may not be aligned warning
12044 rather than isn't aligned. Use isn%'t rather than isn't.
12045 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12047 <case MULT_EXPR>: Improve the case when bottom and one of the
12048 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12049 operand, in that case check if the other operand is multiple of
12050 bottom divided by the INTEGER_CST operand.
12052 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12055 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12056 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12057 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12058 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12059 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12060 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12061 (pa_function_arg_size): New function.
12063 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12065 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12066 in a separate statement.
12068 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12070 PR tree-optimization/83847
12071 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12072 group gathers and scatters.
12074 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12076 PR rtl-optimization/86620
12077 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12079 PR rtl-optimization/83213
12080 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12081 to last if both are JUMP_INSNs.
12083 PR tree-optimization/83843
12084 * gimple-ssa-store-merging.c
12085 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12086 store_immediate_info for bswap/nop orig_stores.
12088 2018-01-15 Andrew Waterman <andrew@sifive.com>
12090 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12092 <UDIV>: Increase cost if !TARGET_DIV.
12094 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12096 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12097 (define_attr "cr_logical_3op"): New.
12098 (cceq_ior_compare): Adjust.
12099 (cceq_ior_compare_complement): Adjust.
12100 (*cceq_rev_compare): Adjust.
12101 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12102 (is_cracked_insn): Adjust.
12103 (insn_must_be_first_in_group): Adjust.
12104 * config/rs6000/40x.md: Adjust.
12105 * config/rs6000/440.md: Adjust.
12106 * config/rs6000/476.md: Adjust.
12107 * config/rs6000/601.md: Adjust.
12108 * config/rs6000/603.md: Adjust.
12109 * config/rs6000/6xx.md: Adjust.
12110 * config/rs6000/7450.md: Adjust.
12111 * config/rs6000/7xx.md: Adjust.
12112 * config/rs6000/8540.md: Adjust.
12113 * config/rs6000/cell.md: Adjust.
12114 * config/rs6000/e300c2c3.md: Adjust.
12115 * config/rs6000/e500mc.md: Adjust.
12116 * config/rs6000/e500mc64.md: Adjust.
12117 * config/rs6000/e5500.md: Adjust.
12118 * config/rs6000/e6500.md: Adjust.
12119 * config/rs6000/mpc.md: Adjust.
12120 * config/rs6000/power4.md: Adjust.
12121 * config/rs6000/power5.md: Adjust.
12122 * config/rs6000/power6.md: Adjust.
12123 * config/rs6000/power7.md: Adjust.
12124 * config/rs6000/power8.md: Adjust.
12125 * config/rs6000/power9.md: Adjust.
12126 * config/rs6000/rs64.md: Adjust.
12127 * config/rs6000/titan.md: Adjust.
12129 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12131 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12132 ix86_indirect_branch_register logic.
12134 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12136 * config/i386/constraints.md (Bs): Update
12137 ix86_indirect_branch_register check. Don't check
12138 ix86_indirect_branch_register with GOT_memory_operand.
12140 * config/i386/predicates.md (GOT_memory_operand): Don't check
12141 ix86_indirect_branch_register here.
12142 (GOT32_symbol_operand): Likewise.
12144 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12146 * config/i386/predicates.md (constant_call_address_operand):
12147 Rewrite ix86_indirect_branch_register logic.
12148 (sibcall_insn_operand): Likewise.
12150 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12152 * config/i386/constraints.md (Bs): Replace
12153 ix86_indirect_branch_thunk_register with
12154 ix86_indirect_branch_register.
12156 * config/i386/i386.md (indirect_jump): Likewise.
12157 (tablejump): Likewise.
12158 (*sibcall_memory): Likewise.
12159 (*sibcall_value_memory): Likewise.
12160 Peepholes of indirect call and jump via memory: Likewise.
12161 * config/i386/i386.opt: Likewise.
12162 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12163 (GOT_memory_operand): Likewise.
12164 (call_insn_operand): Likewise.
12165 (sibcall_insn_operand): Likewise.
12166 (GOT32_symbol_operand): Likewise.
12168 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12170 PR middle-end/83837
12171 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12172 type rather than type addr's type points to.
12173 (expand_omp_atomic_mutex): Likewise.
12174 (expand_omp_atomic): Likewise.
12176 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12179 * config/i386/i386.c (output_indirect_thunk_function): Use
12180 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12181 for __x86_return_thunk.
12183 2018-01-15 Richard Biener <rguenther@suse.de>
12185 PR middle-end/83850
12186 * expmed.c (extract_bit_field_1): Fix typo.
12188 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12191 * config/arm/iterators.md (VF): New mode iterator.
12192 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12193 Remove integer-related logic from pattern.
12194 (neon_vabd<mode>_3): Likewise.
12196 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12198 PR middle-end/82694
12199 * common.opt (fstrict-overflow): No longer an alias.
12200 (fwrapv-pointer): New option.
12201 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12202 also for pointer types based on flag_wrapv_pointer.
12203 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12204 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12205 opts->x_flag_wrapv got set.
12206 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12207 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12208 POINTER_TYPE_OVERFLOW_UNDEFINED.
12209 * match.pd: Likewise in address comparison pattern.
12210 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12212 2018-01-15 Richard Biener <rguenther@suse.de>
12215 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12216 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12217 Reset type names to their identifier if their TYPE_DECL doesn't
12218 have linkage (and thus is used for ODR and devirt).
12219 (save_debug_info_for_decl): Remove.
12220 (save_debug_info_for_type): Likewise.
12221 (add_tree_to_fld_list): Adjust.
12222 * tree-pretty-print.c (dump_generic_node): Make dumping of
12223 type names more robust.
12225 2018-01-15 Richard Biener <rguenther@suse.de>
12227 * BASE-VER: Bump to 8.0.1.
12229 2018-01-14 Martin Sebor <msebor@redhat.com>
12232 * builtins.c (check_access): Avoid warning when the no-warning bit
12235 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12237 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12238 * ira-color (allocno_hard_regs_compare): Likewise.
12240 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12243 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12244 Use .pushsection/.popsection.
12246 2018-01-14 Martin Sebor <msebor@redhat.com>
12249 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12251 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12253 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12254 entry from extra_headers.
12255 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12256 extra_headers, make the list bitwise identical to the i?86-*-* one.
12258 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12260 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12261 -mcmodel=large with -mindirect-branch=thunk,
12262 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12263 -mfunction-return=thunk-extern.
12264 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12265 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12266 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12268 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12270 * config/i386/i386.c (print_reg): Print the name of the full
12271 integer register without '%'.
12272 (ix86_print_operand): Handle 'V'.
12273 * doc/extend.texi: Document 'V' modifier.
12275 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12277 * config/i386/constraints.md (Bs): Disallow memory operand for
12278 -mindirect-branch-register.
12280 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12281 (GOT_memory_operand): Likewise.
12282 (call_insn_operand): Likewise.
12283 (sibcall_insn_operand): Likewise.
12284 (GOT32_symbol_operand): Likewise.
12285 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12286 for -mindirect-branch-register.
12287 (tablejump): Likewise.
12288 (*sibcall_memory): Likewise.
12289 (*sibcall_value_memory): Likewise.
12290 Disallow peepholes of indirect call and jump via memory for
12291 -mindirect-branch-register.
12292 (*call_pop): Replace m with Bw.
12293 (*call_value_pop): Likewise.
12294 (*sibcall_pop_memory): Replace m with Bs.
12295 * config/i386/i386.opt (mindirect-branch-register): New option.
12296 * doc/invoke.texi: Document -mindirect-branch-register option.
12298 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12300 * config/i386/i386-protos.h (ix86_output_function_return): New.
12301 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12302 set function_return_type.
12303 (indirect_thunk_name): Add ret_p to indicate thunk for function
12305 (output_indirect_thunk_function): Pass false to
12306 indirect_thunk_name.
12307 (ix86_output_indirect_branch_via_reg): Likewise.
12308 (ix86_output_indirect_branch_via_push): Likewise.
12309 (output_indirect_thunk_function): Create alias for function
12310 return thunk if regno < 0.
12311 (ix86_output_function_return): New function.
12312 (ix86_handle_fndecl_attribute): Handle function_return.
12313 (ix86_attribute_table): Add function_return.
12314 * config/i386/i386.h (machine_function): Add
12315 function_return_type.
12316 * config/i386/i386.md (simple_return_internal): Use
12317 ix86_output_function_return.
12318 (simple_return_internal_long): Likewise.
12319 * config/i386/i386.opt (mfunction-return=): New option.
12320 (indirect_branch): Mention -mfunction-return=.
12321 * doc/extend.texi: Document function_return function attribute.
12322 * doc/invoke.texi: Document -mfunction-return= option.
12324 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12326 * config/i386/i386-opts.h (indirect_branch): New.
12327 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12328 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12329 with local indirect jump when converting indirect call and jump.
12330 (ix86_set_indirect_branch_type): New.
12331 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12332 (indirectlabelno): New.
12333 (indirect_thunk_needed): Likewise.
12334 (indirect_thunk_bnd_needed): Likewise.
12335 (indirect_thunks_used): Likewise.
12336 (indirect_thunks_bnd_used): Likewise.
12337 (INDIRECT_LABEL): Likewise.
12338 (indirect_thunk_name): Likewise.
12339 (output_indirect_thunk): Likewise.
12340 (output_indirect_thunk_function): Likewise.
12341 (ix86_output_indirect_branch_via_reg): Likewise.
12342 (ix86_output_indirect_branch_via_push): Likewise.
12343 (ix86_output_indirect_branch): Likewise.
12344 (ix86_output_indirect_jmp): Likewise.
12345 (ix86_code_end): Call output_indirect_thunk_function if needed.
12346 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12348 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12349 (ix86_attribute_table): Add indirect_branch.
12350 * config/i386/i386.h (machine_function): Add indirect_branch_type
12351 and has_local_indirect_jump.
12352 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12354 (tablejump): Likewise.
12355 (*indirect_jump): Use ix86_output_indirect_jmp.
12356 (*tablejump_1): Likewise.
12357 (simple_return_indirect_internal): Likewise.
12358 * config/i386/i386.opt (mindirect-branch=): New option.
12359 (indirect_branch): New.
12362 (thunk-inline): Likewise.
12363 (thunk-extern): Likewise.
12364 * doc/extend.texi: Document indirect_branch function attribute.
12365 * doc/invoke.texi: Document -mindirect-branch= option.
12367 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12370 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12372 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12374 * ipa-inline.c (want_inline_small_function_p): Return false if
12375 inlining has already failed with CIF_FINAL_ERROR.
12376 (update_caller_keys): Call want_inline_small_function_p before
12378 (update_callee_keys): Likewise.
12380 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12382 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12384 (rs6000_quadword_masked_address_p): Likewise.
12385 (quad_aligned_load_p): Likewise.
12386 (quad_aligned_store_p): Likewise.
12387 (const_load_sequence_p): Add comment to describe the outer-most loop.
12388 (mimic_memory_attributes_and_flags): New function.
12389 (rs6000_gen_stvx): Likewise.
12390 (replace_swapped_aligned_store): Likewise.
12391 (rs6000_gen_lvx): Likewise.
12392 (replace_swapped_aligned_load): Likewise.
12393 (replace_swapped_load_constant): Capitalize argument name in
12394 comment describing this function.
12395 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12396 and stores that access quad-word aligned addresses and replace
12397 with stvx or lvx instructions when appropriate.
12398 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12399 New function prototype.
12400 (rs6000_quadword_masked_address_p): Likewise.
12401 (rs6000_gen_lvx): Likewise.
12402 (rs6000_gen_stvx): Likewise.
12403 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12404 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12405 when memory address is aligned.
12406 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12407 this split to select lvx instruction when memory address is aligned.
12408 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12409 instruction when memory address is aligned.
12410 (*vsx_le_perm_load_v16qi): Likewise.
12411 (four unnamed splitters): Modify to select the stvx instruction
12412 when memory is aligned.
12414 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12416 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12417 which appears in the queue multiple times.
12419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12420 Alan Hayward <alan.hayward@arm.com>
12421 David Sherwood <david.sherwood@arm.com>
12423 * tree-vectorizer.h (vec_lower_bound): New structure.
12424 (_loop_vec_info): Add check_nonzero and lower_bounds.
12425 (LOOP_VINFO_CHECK_NONZERO): New macro.
12426 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12427 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12428 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12429 fields. Make seg_len the distance travelled, not including the
12431 (dr_direction_indicator): Declare.
12432 (dr_zero_step_indicator): Likewise.
12433 (dr_known_forward_stride_p): Likewise.
12434 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12436 (runtime_alias_check_p): Allow runtime alias checks with
12438 (operator ==): Compare access_size and align.
12439 (prune_runtime_alias_test_list): Rework for new distinction between
12440 the access_size and seg_len.
12441 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12443 (get_segment_min_max): New function.
12444 (create_intersect_range_checks): Use it.
12445 (dr_step_indicator): New function.
12446 (dr_direction_indicator): Likewise.
12447 (dr_zero_step_indicator): Likewise.
12448 (dr_known_forward_stride_p): Likewise.
12449 * tree-loop-distribution.c (data_ref_segment_size): Return
12450 DR_STEP * (niters - 1).
12451 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12453 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12454 (vect_preserves_scalar_order_p): New function, split out from...
12455 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12456 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12457 (vect_vfa_access_size): New function.
12458 (vect_vfa_align): Likewise.
12459 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12460 (dump_lower_bound): New function.
12461 (vect_check_lower_bound): Likewise.
12462 (vect_small_gap_p): Likewise.
12463 (vectorizable_with_step_bound_p): Likewise.
12464 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12465 depencies if the vectorization factor is 1. Convert the checks
12466 for nonzero steps into checks on the bounds of DR_STEP. Try using
12467 a bunds check for variable steps if the minimum required step is
12468 relatively small. Update calls to the dr_with_seg_len
12469 constructor and to vect_compile_time_alias.
12470 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12472 (vect_loop_versioning): Call it.
12473 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12475 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12478 Alan Hayward <alan.hayward@arm.com>
12479 David Sherwood <david.sherwood@arm.com>
12481 * doc/sourcebuild.texi (vect_scatter_store): Document.
12482 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12484 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12486 * genopinit.c (main): Add supports_vec_scatter_store and
12487 supports_vec_scatter_store_cached to target_optabs.
12488 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12489 IFN_MASK_SCATTER_STORE.
12490 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12492 * internal-fn.h (internal_store_fn_p): Declare.
12493 (internal_fn_stored_value_index): Likewise.
12494 * internal-fn.c (scatter_store_direct): New macro.
12495 (expand_scatter_store_optab_fn): New function.
12496 (direct_scatter_store_optab_supported_p): New macro.
12497 (internal_store_fn_p): New function.
12498 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12499 IFN_MASK_SCATTER_STORE.
12500 (internal_fn_mask_index): Likewise.
12501 (internal_fn_stored_value_index): New function.
12502 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12503 for scatter stores.
12504 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12505 * optabs-query.c (supports_vec_scatter_store_p): New function.
12506 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12507 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12508 true for scatter stores.
12509 (vect_gather_scatter_fn_p): Handle scatter stores too.
12510 (vect_check_gather_scatter): Consider using scatter stores if
12511 supports_vec_scatter_store_p.
12512 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12513 scatter stores too.
12514 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12515 internal_fn_stored_value_index.
12516 (check_load_store_masking): Handle scatter stores too.
12517 (vect_get_store_rhs): Make public.
12518 (vectorizable_call): Use internal_store_fn_p.
12519 (vectorizable_store): Handle scatter store internal functions.
12520 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12521 when deciding whether the end of the group has been reached.
12522 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12523 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12524 (mask_scatter_store<mode>): New insns.
12526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12527 Alan Hayward <alan.hayward@arm.com>
12528 David Sherwood <david.sherwood@arm.com>
12530 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12531 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12532 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12534 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12535 Use vect_truncate_gather_scatter_offset if we can't treat the
12536 operation as a normal gather load or scatter store.
12537 (get_group_load_store_type): Take the gather_scatter_info
12538 as argument. Try using a gather load or scatter store for
12539 single-element groups.
12540 (get_load_store_type): Update calls to get_group_load_store_type
12541 and vect_use_strided_gather_scatters_p.
12543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12544 Alan Hayward <alan.hayward@arm.com>
12545 David Sherwood <david.sherwood@arm.com>
12547 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12548 optional tree argument.
12549 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12551 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12552 but continue to use the current value as a fallback.
12553 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12554 to compare the updates.
12555 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12556 (get_load_store_type): Use it when handling a strided access.
12557 (vect_get_strided_load_store_ops): New function.
12558 (vect_get_data_ptr_increment): Likewise.
12559 (vectorizable_load): Handle strided gather loads. Always pass
12560 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12563 Alan Hayward <alan.hayward@arm.com>
12564 David Sherwood <david.sherwood@arm.com>
12566 * doc/md.texi (gather_load@var{m}): Document.
12567 (mask_gather_load@var{m}): Likewise.
12568 * genopinit.c (main): Add supports_vec_gather_load and
12569 supports_vec_gather_load_cached to target_optabs.
12570 * optabs-tree.c (init_tree_optimization_optabs): Use
12571 ggc_cleared_alloc to allocate target_optabs.
12572 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12573 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12575 * internal-fn.h (internal_load_fn_p): Declare.
12576 (internal_gather_scatter_fn_p): Likewise.
12577 (internal_fn_mask_index): Likewise.
12578 (internal_gather_scatter_fn_supported_p): Likewise.
12579 * internal-fn.c (gather_load_direct): New macro.
12580 (expand_gather_load_optab_fn): New function.
12581 (direct_gather_load_optab_supported_p): New macro.
12582 (direct_internal_fn_optab): New function.
12583 (internal_load_fn_p): Likewise.
12584 (internal_gather_scatter_fn_p): Likewise.
12585 (internal_fn_mask_index): Likewise.
12586 (internal_gather_scatter_fn_supported_p): Likewise.
12587 * optabs-query.c (supports_at_least_one_mode_p): New function.
12588 (supports_vec_gather_load_p): Likewise.
12589 * optabs-query.h (supports_vec_gather_load_p): Declare.
12590 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12591 and memory_type field.
12592 (NUM_PATTERNS): Bump to 15.
12593 * tree-vect-data-refs.c: Include internal-fn.h.
12594 (vect_gather_scatter_fn_p): New function.
12595 (vect_describe_gather_scatter_call): Likewise.
12596 (vect_check_gather_scatter): Try using internal functions for
12597 gather loads. Recognize existing calls to a gather load function.
12598 (vect_analyze_data_refs): Consider using gather loads if
12599 supports_vec_gather_load_p.
12600 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12601 (vect_get_gather_scatter_offset_type): Likewise.
12602 (vect_convert_mask_for_vectype): Likewise.
12603 (vect_add_conversion_to_patterm): Likewise.
12604 (vect_try_gather_scatter_pattern): Likewise.
12605 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12606 (vect_vect_recog_func_ptrs): Add it.
12607 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12608 internal_fn_mask_index and internal_gather_scatter_fn_p.
12609 (check_load_store_masking): Take the gather_scatter_info as an
12610 argument and handle gather loads.
12611 (vect_get_gather_scatter_ops): New function.
12612 (vectorizable_call): Check internal_load_fn_p.
12613 (vectorizable_load): Likewise. Handle gather load internal
12615 (vectorizable_store): Update call to check_load_store_masking.
12616 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12617 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12618 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12619 (aarch64_gather_scale_operand_d): New predicates.
12620 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12621 (mask_gather_load<mode>): New insns.
12623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12624 Alan Hayward <alan.hayward@arm.com>
12625 David Sherwood <david.sherwood@arm.com>
12627 * optabs.def (fold_left_plus_optab): New optab.
12628 * doc/md.texi (fold_left_plus_@var{m}): Document.
12629 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12630 * internal-fn.c (fold_left_direct): Define.
12631 (expand_fold_left_optab_fn): Likewise.
12632 (direct_fold_left_optab_supported_p): Likewise.
12633 * fold-const-call.c (fold_const_fold_left): New function.
12634 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12635 * tree-parloops.c (valid_reduction_p): New function.
12636 (gather_scalar_reductions): Use it.
12637 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12638 (vect_finish_replace_stmt): Declare.
12639 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12640 (needs_fold_left_reduction_p): New function, split out from...
12641 (vect_is_simple_reduction): ...here. Accept reductions that
12642 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12643 (vect_force_simple_reduction): Also store the reduction type in
12644 the assignment's STMT_VINFO_REDUC_TYPE.
12645 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12646 (merge_with_identity): New function.
12647 (vect_expand_fold_left): Likewise.
12648 (vectorize_fold_left_reduction): Likewise.
12649 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12650 scalar phi in place for it. Check for target support and reject
12651 cases that would reassociate the operation. Defer the transform
12652 phase to vectorize_fold_left_reduction.
12653 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12654 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12655 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12659 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12660 call to ifc_temp_var.
12662 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12663 Alan Hayward <alan.hayward@arm.com>
12664 David Sherwood <david.sherwood@arm.com>
12666 * target.def (legitimize_address_displacement): Take the original
12667 offset as a poly_int.
12668 * targhooks.h (default_legitimize_address_displacement): Update
12670 * targhooks.c (default_legitimize_address_displacement): Likewise.
12671 * doc/tm.texi: Regenerate.
12672 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12673 as an argument, moving assert of ad->disp == ad->disp_term to...
12674 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12675 Try calling targetm.legitimize_address_displacement before expanding
12676 the address rather than afterwards, and adjust for the new interface.
12677 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12678 Match the new hook interface. Handle SVE addresses.
12679 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12680 new hook interface.
12682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12684 * Makefile.in (OBJS): Add early-remat.o.
12685 * target.def (select_early_remat_modes): New hook.
12686 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12687 * doc/tm.texi: Regenerate.
12688 * targhooks.h (default_select_early_remat_modes): Declare.
12689 * targhooks.c (default_select_early_remat_modes): New function.
12690 * timevar.def (TV_EARLY_REMAT): New timevar.
12691 * passes.def (pass_early_remat): New pass.
12692 * tree-pass.h (make_pass_early_remat): Declare.
12693 * early-remat.c: New file.
12694 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12696 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12699 Alan Hayward <alan.hayward@arm.com>
12700 David Sherwood <david.sherwood@arm.com>
12702 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12703 vfm1 with a bound_epilog parameter.
12704 (vect_do_peeling): Update calls accordingly, and move the prologue
12705 call earlier in the function. Treat the base bound_epilog as 0 for
12706 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12707 this base when peeling for gaps.
12708 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12709 with fully-masked loops.
12710 (vect_estimate_min_profitable_iters): Handle the single peeled
12711 iteration in that case.
12713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12714 Alan Hayward <alan.hayward@arm.com>
12715 David Sherwood <david.sherwood@arm.com>
12717 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12718 single-element interleaving even if the size is not a power of 2.
12719 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12720 accesses for single-element interleaving if the group size is
12723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12724 Alan Hayward <alan.hayward@arm.com>
12725 David Sherwood <david.sherwood@arm.com>
12727 * doc/md.texi (fold_extract_last_@var{m}): Document.
12728 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12729 * optabs.def (fold_extract_last_optab): New optab.
12730 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12731 * internal-fn.c (fold_extract_direct): New macro.
12732 (expand_fold_extract_optab_fn): Likewise.
12733 (direct_fold_extract_optab_supported_p): Likewise.
12734 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12735 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12736 EXTRACT_LAST_REDUCTION.
12737 (get_initial_def_for_reduction): Do not create an initial vector
12738 for EXTRACT_LAST_REDUCTION reductions.
12739 (vectorizable_reduction): Leave the scalar phi in place for
12740 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12741 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12742 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12743 transform phase to vectorizable_condition.
12744 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12746 (vect_finish_stmt_generation): ...here.
12747 (vect_finish_replace_stmt): New function.
12748 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12749 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12751 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12754 Alan Hayward <alan.hayward@arm.com>
12755 David Sherwood <david.sherwood@arm.com>
12757 * doc/md.texi (extract_last_@var{m}): Document.
12758 * optabs.def (extract_last_optab): New optab.
12759 * internal-fn.def (EXTRACT_LAST): New internal function.
12760 * internal-fn.c (cond_unary_direct): New macro.
12761 (expand_cond_unary_optab_fn): Likewise.
12762 (direct_cond_unary_optab_supported_p): Likewise.
12763 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12764 loops using EXTRACT_LAST.
12765 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12766 (extract_last_<mode>): ...this optab.
12767 (vec_extract<mode><Vel>): Update accordingly.
12769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12770 Alan Hayward <alan.hayward@arm.com>
12771 David Sherwood <david.sherwood@arm.com>
12773 * target.def (empty_mask_is_expensive): New hook.
12774 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12775 * doc/tm.texi: Regenerate.
12776 * targhooks.h (default_empty_mask_is_expensive): Declare.
12777 * targhooks.c (default_empty_mask_is_expensive): New function.
12778 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12779 if the target says that empty masks are expensive.
12780 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12782 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12785 Alan Hayward <alan.hayward@arm.com>
12786 David Sherwood <david.sherwood@arm.com>
12788 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12789 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12790 (vect_use_loop_mask_for_alignment_p): New function.
12791 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12792 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12793 niters_skip argument. Make sure that the first niters_skip elements
12794 of the first iteration are inactive.
12795 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12796 Update call to vect_set_loop_masks_directly.
12797 (get_misalign_in_elems): New function, split out from...
12798 (vect_gen_prolog_loop_niters): ...here.
12799 (vect_update_init_of_dr): Take a code argument that specifies whether
12800 the adjustment should be added or subtracted.
12801 (vect_update_init_of_drs): Likewise.
12802 (vect_prepare_for_masked_peels): New function.
12803 (vect_do_peeling): Skip prologue peeling if we're using a mask
12804 instead. Update call to vect_update_inits_of_drs.
12805 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12807 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12808 alignment. Do not include the number of peeled iterations in
12809 the minimum threshold in that case.
12810 (vectorizable_induction): Adjust the start value down by
12811 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12812 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12813 Take the number of skipped iterations into account when calculating
12815 * tree-vect-stmts.c (vect_gen_while_not): New function.
12817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12818 Alan Hayward <alan.hayward@arm.com>
12819 David Sherwood <david.sherwood@arm.com>
12821 * doc/sourcebuild.texi (vect_fully_masked): Document.
12822 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12823 default value to 0.
12824 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12826 (vect_analyze_loop_2): ...here. Don't check the vectorization
12827 factor against the number of loop iterations if the loop is
12830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12831 Alan Hayward <alan.hayward@arm.com>
12832 David Sherwood <david.sherwood@arm.com>
12834 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12835 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12836 (dump_groups): Update accordingly.
12837 (iv_use::mem_type): New member variable.
12838 (address_p): New function.
12839 (record_use): Add a mem_type argument and initialize the new
12841 (record_group_use): Add a mem_type argument. Use address_p.
12842 Remove obsolete null checks of base_object. Update call to record_use.
12843 (find_interesting_uses_op): Update call to record_group_use.
12844 (find_interesting_uses_cond): Likewise.
12845 (find_interesting_uses_address): Likewise.
12846 (get_mem_type_for_internal_fn): New function.
12847 (find_address_like_use): Likewise.
12848 (find_interesting_uses_stmt): Try find_address_like_use before
12849 calling find_interesting_uses_op.
12850 (addr_offset_valid_p): Use the iv mem_type field as the type
12851 of the addressed memory.
12852 (add_autoinc_candidates): Likewise.
12853 (get_address_cost): Likewise.
12854 (split_small_address_groups_p): Use address_p.
12855 (split_address_groups): Likewise.
12856 (add_iv_candidate_for_use): Likewise.
12857 (autoinc_possible_for_pair): Likewise.
12858 (rewrite_groups): Likewise.
12859 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12860 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12861 (get_alias_ptr_type_for_ptr_address): New function.
12862 (rewrite_use_address): Rewrite address uses in calls that were
12863 identified by find_address_like_use.
12865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12866 Alan Hayward <alan.hayward@arm.com>
12867 David Sherwood <david.sherwood@arm.com>
12869 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12871 * gimple-expr.h (is_gimple_addressable: Likewise.
12872 * gimple-expr.c (is_gimple_address): Likewise.
12873 * internal-fn.c (expand_call_mem_ref): New function.
12874 (expand_mask_load_optab_fn): Use it.
12875 (expand_mask_store_optab_fn): Likewise.
12877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12878 Alan Hayward <alan.hayward@arm.com>
12879 David Sherwood <david.sherwood@arm.com>
12881 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12882 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12883 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12884 (cond_umax@var{mode}): Document.
12885 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12886 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12887 (cond_umin_optab, cond_umax_optab): New optabs.
12888 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12889 (COND_IOR, COND_XOR): New internal functions.
12890 * internal-fn.h (get_conditional_internal_fn): Declare.
12891 * internal-fn.c (cond_binary_direct): New macro.
12892 (expand_cond_binary_optab_fn): Likewise.
12893 (direct_cond_binary_optab_supported_p): Likewise.
12894 (get_conditional_internal_fn): New function.
12895 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12896 Cope with reduction statements that are vectorized as calls rather
12898 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12899 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12900 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12901 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12902 (UNSPEC_COND_EOR): New unspecs.
12903 (optab): Add mappings for them.
12904 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12905 (sve_int_op, sve_fp_op): New int attributes.
12907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12908 Alan Hayward <alan.hayward@arm.com>
12909 David Sherwood <david.sherwood@arm.com>
12911 * optabs.def (while_ult_optab): New optab.
12912 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12913 * internal-fn.def (WHILE_ULT): New internal function.
12914 * internal-fn.h (direct_internal_fn_supported_p): New override
12915 that takes two types as argument.
12916 * internal-fn.c (while_direct): New macro.
12917 (expand_while_optab_fn): New function.
12918 (convert_optab_supported_p): Likewise.
12919 (direct_while_optab_supported_p): New macro.
12920 * wide-int.h (wi::udiv_ceil): New function.
12921 * tree-vectorizer.h (rgroup_masks): New structure.
12922 (vec_loop_masks): New typedef.
12923 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12924 and fully_masked_p.
12925 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12926 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
12927 (vect_max_vf): New function.
12928 (slpeel_make_loop_iterate_ntimes): Delete.
12929 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
12930 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
12931 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
12932 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
12933 internal-fn.h, stor-layout.h and optabs-query.h.
12934 (vect_set_loop_mask): New function.
12935 (add_preheader_seq): Likewise.
12936 (add_header_seq): Likewise.
12937 (interleave_supported_p): Likewise.
12938 (vect_maybe_permute_loop_masks): Likewise.
12939 (vect_set_loop_masks_directly): Likewise.
12940 (vect_set_loop_condition_masked): Likewise.
12941 (vect_set_loop_condition_unmasked): New function, split out from
12942 slpeel_make_loop_iterate_ntimes.
12943 (slpeel_make_loop_iterate_ntimes): Rename to..
12944 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
12945 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
12946 (vect_do_peeling): Update call accordingly.
12947 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
12949 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12950 mask_compare_type, can_fully_mask_p and fully_masked_p.
12951 (release_vec_loop_masks): New function.
12952 (_loop_vec_info): Use it to free the loop masks.
12953 (can_produce_all_loop_masks_p): New function.
12954 (vect_get_max_nscalars_per_iter): Likewise.
12955 (vect_verify_full_masking): Likewise.
12956 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
12957 retries, and free the mask rgroups before retrying. Check loop-wide
12958 reasons for disallowing fully-masked loops. Make the final decision
12959 about whether use a fully-masked loop or not.
12960 (vect_estimate_min_profitable_iters): Do not assume that peeling
12961 for the number of iterations will be needed for fully-masked loops.
12962 (vectorizable_reduction): Disable fully-masked loops.
12963 (vectorizable_live_operation): Likewise.
12964 (vect_halve_mask_nunits): New function.
12965 (vect_double_mask_nunits): Likewise.
12966 (vect_record_loop_mask): Likewise.
12967 (vect_get_loop_mask): Likewise.
12968 (vect_transform_loop): Handle the case in which the final loop
12969 iteration might handle a partial vector. Call vect_set_loop_condition
12970 instead of slpeel_make_loop_iterate_ntimes.
12971 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
12972 (check_load_store_masking): New function.
12973 (prepare_load_store_mask): Likewise.
12974 (vectorizable_store): Handle fully-masked loops.
12975 (vectorizable_load): Likewise.
12976 (supportable_widening_operation): Use vect_halve_mask_nunits for
12978 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
12979 (vect_gen_while): New function.
12980 * config/aarch64/aarch64.md (umax<mode>3): New expander.
12981 (aarch64_uqdec<mode>): New insn.
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 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
12988 (reduc_xor_scal_optab): New optabs.
12989 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
12990 (reduc_xor_scal_@var{m}): Document.
12991 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
12992 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
12993 internal functions.
12994 * fold-const-call.c (fold_const_call): Handle them.
12995 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
12996 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
12997 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
12998 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
12999 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13000 (UNSPEC_XORV): New unspecs.
13001 (optab): Add entries for them.
13002 (BITWISEV): New int iterator.
13003 (bit_reduc_op): New int attributes.
13005 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13006 Alan Hayward <alan.hayward@arm.com>
13007 David Sherwood <david.sherwood@arm.com>
13009 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13010 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13011 * optabs.def (vec_shl_insert_optab): New optab.
13012 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13013 (duplicate_and_interleave): Likewise.
13014 * tree-vect-loop.c: Include internal-fn.h.
13015 (neutral_op_for_slp_reduction): New function, split out from
13016 get_initial_defs_for_reduction.
13017 (get_initial_def_for_reduction): Handle option 2 for variable-length
13018 vectors by loading the neutral value into a vector and then shifting
13019 the initial value into element 0.
13020 (get_initial_defs_for_reduction): Replace the code argument with
13021 the neutral value calculated by neutral_op_for_slp_reduction.
13022 Use gimple_build_vector for constant-length vectors.
13023 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13024 but the first group_size elements have a neutral value.
13025 Use duplicate_and_interleave otherwise.
13026 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13027 Update call to get_initial_defs_for_reduction. Handle SLP
13028 reductions for variable-length vectors by creating one vector
13029 result for each scalar result, with the elements associated
13030 with other scalar results stubbed out with the neutral value.
13031 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13032 Require IFN_VEC_SHL_INSERT for double reductions on
13033 variable-length vectors, or SLP reductions that have
13034 a neutral value. Require can_duplicate_and_interleave_p
13035 support for variable-length unchained SLP reductions if there
13036 is no neutral value, such as for MIN/MAX reductions. Also require
13037 the number of vector elements to be a multiple of the number of
13038 SLP statements when doing variable-length unchained SLP reductions.
13039 Update call to vect_create_epilog_for_reduction.
13040 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13041 and remove initial values.
13042 (duplicate_and_interleave): Make public.
13043 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13044 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13047 Alan Hayward <alan.hayward@arm.com>
13048 David Sherwood <david.sherwood@arm.com>
13050 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13051 (can_duplicate_and_interleave_p): New function.
13052 (vect_get_and_check_slp_defs): Take the vector of statements
13053 rather than just the current one. Remove excess parentheses.
13054 Restriction rejectinon of vect_constant_def and vect_external_def
13055 for variable-length vectors to boolean types, or types for which
13056 can_duplicate_and_interleave_p is false.
13057 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13058 (duplicate_and_interleave): New function.
13059 (vect_get_constant_vectors): Use gimple_build_vector for
13060 constant-length vectors and suitable variable-length constant
13061 vectors. Use duplicate_and_interleave for other variable-length
13062 vectors. Don't defer the update when inserting new statements.
13064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13065 Alan Hayward <alan.hayward@arm.com>
13066 David Sherwood <david.sherwood@arm.com>
13068 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13069 min_profitable_iters doesn't go negative.
13071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13072 Alan Hayward <alan.hayward@arm.com>
13073 David Sherwood <david.sherwood@arm.com>
13075 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13076 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13077 * optabs.def (vec_mask_load_lanes_optab): New optab.
13078 (vec_mask_store_lanes_optab): Likewise.
13079 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13080 (MASK_STORE_LANES): Likewise.
13081 * internal-fn.c (mask_load_lanes_direct): New macro.
13082 (mask_store_lanes_direct): Likewise.
13083 (expand_mask_load_optab_fn): Handle masked operations.
13084 (expand_mask_load_lanes_optab_fn): New macro.
13085 (expand_mask_store_optab_fn): Handle masked operations.
13086 (expand_mask_store_lanes_optab_fn): New macro.
13087 (direct_mask_load_lanes_optab_supported_p): Likewise.
13088 (direct_mask_store_lanes_optab_supported_p): Likewise.
13089 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13091 (vect_load_lanes_supported): Likewise.
13092 * tree-vect-data-refs.c (strip_conversion): New function.
13093 (can_group_stmts_p): Likewise.
13094 (vect_analyze_data_ref_accesses): Use it instead of checking
13095 for a pair of assignments.
13096 (vect_store_lanes_supported): Take a masked_p parameter.
13097 (vect_load_lanes_supported): Likewise.
13098 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13099 vect_store_lanes_supported and vect_load_lanes_supported.
13100 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13101 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13102 parameter. Don't allow gaps for masked accesses.
13103 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13104 and vect_load_lanes_supported.
13105 (get_load_store_type): Take a masked_p parameter and update
13106 call to get_group_load_store_type.
13107 (vectorizable_store): Update call to get_load_store_type.
13108 Handle IFN_MASK_STORE_LANES.
13109 (vectorizable_load): Update call to get_load_store_type.
13110 Handle IFN_MASK_LOAD_LANES.
13112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13113 Alan Hayward <alan.hayward@arm.com>
13114 David Sherwood <david.sherwood@arm.com>
13116 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13118 * config/aarch64/aarch64-protos.h
13119 (aarch64_sve_struct_memory_operand_p): Declare.
13120 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13121 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13122 (VPRED, vpred): Handle SVE structure modes.
13123 * config/aarch64/constraints.md (Utx): New constraint.
13124 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13125 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13126 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13127 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13128 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13129 structure modes. Split into pieces after RA.
13130 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13131 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13133 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13134 SVE structure modes.
13135 (aarch64_classify_address): Likewise.
13136 (sizetochar): Move earlier in file.
13137 (aarch64_print_operand): Handle SVE register lists.
13138 (aarch64_array_mode): New function.
13139 (aarch64_sve_struct_memory_operand_p): Likewise.
13140 (TARGET_ARRAY_MODE): Redefine.
13142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13143 Alan Hayward <alan.hayward@arm.com>
13144 David Sherwood <david.sherwood@arm.com>
13146 * target.def (array_mode): New target hook.
13147 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13148 * doc/tm.texi: Regenerate.
13149 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13150 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13151 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13152 targetm.array_mode.
13153 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13156 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13157 Alan Hayward <alan.hayward@arm.com>
13158 David Sherwood <david.sherwood@arm.com>
13160 * fold-const.c (fold_binary_loc): Check the argument types
13161 rather than the result type when testing for a vector operation.
13163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13165 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13166 * doc/tm.texi: Regenerate.
13168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13169 Alan Hayward <alan.hayward@arm.com>
13170 David Sherwood <david.sherwood@arm.com>
13172 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13173 (sve): Document new AArch64 extension.
13174 * doc/md.texi (w): Extend the description of the AArch64
13175 constraint to include SVE vectors.
13176 (Upl, Upa): Document new AArch64 predicate constraints.
13177 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13179 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13180 (msve-vector-bits=): New option.
13181 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13182 SVE when these are disabled.
13183 (sve): New extension.
13184 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13185 modes. Adjust their number of units based on aarch64_sve_vg.
13186 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13187 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13188 aarch64_addr_query_type.
13189 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13190 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13191 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13192 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13193 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13194 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13195 (aarch64_simd_imm_zero_p): Delete.
13196 (aarch64_check_zero_based_sve_index_immediate): Declare.
13197 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13198 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13199 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13200 (aarch64_sve_float_mul_immediate_p): Likewise.
13201 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13202 rather than an rtx.
13203 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13204 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13205 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13206 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13207 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13208 (aarch64_regmode_natural_size): Likewise.
13209 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13210 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13212 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13213 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13214 for VG and the SVE predicate registers.
13215 (V_ALIASES): Add a "z"-prefixed alias.
13216 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13217 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13218 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13219 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13220 (REG_CLASS_NAMES): Add entries for them.
13221 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13222 and the predicate registers.
13223 (aarch64_sve_vg): Declare.
13224 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13225 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13226 (REGMODE_NATURAL_SIZE): Define.
13227 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13229 * config/aarch64/aarch64.c: Include cfgrtl.h.
13230 (simd_immediate_info): Add a constructor for series vectors,
13231 and an associated step field.
13232 (aarch64_sve_vg): New variable.
13233 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13234 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13235 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13236 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13237 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13238 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13239 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13240 (aarch64_get_mask_mode): New functions.
13241 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13242 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13243 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13244 predicate modes and predicate registers. Explicitly restrict
13245 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13246 to store a vector mode if it is recognized by
13247 aarch64_classify_vector_mode.
13248 (aarch64_regmode_natural_size): New function.
13249 (aarch64_hard_regno_caller_save_mode): Return the original mode
13251 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13252 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13253 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13254 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13256 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13257 does not overlap dest if the function is frame-related. Handle
13259 (aarch64_split_add_offset): New function.
13260 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13261 them aarch64_add_offset.
13262 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13263 and update call to aarch64_sub_sp.
13264 (aarch64_add_cfa_expression): New function.
13265 (aarch64_expand_prologue): Pass extra temporary registers to the
13266 functions above. Handle the case in which we need to emit new
13267 DW_CFA_expressions for registers that were originally saved
13268 relative to the stack pointer, but now have to be expressed
13269 relative to the frame pointer.
13270 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13272 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13273 IP0 and IP1 values for SVE frames.
13274 (aarch64_expand_vec_series): New function.
13275 (aarch64_expand_sve_widened_duplicate): Likewise.
13276 (aarch64_expand_sve_const_vector): Likewise.
13277 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13278 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13279 into the register, rather than emitting a SET directly.
13280 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13281 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13282 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13283 (offset_9bit_signed_scaled_p): New functions.
13284 (aarch64_replicate_bitmask_imm): New function.
13285 (aarch64_bitmask_imm): Use it.
13286 (aarch64_cannot_force_const_mem): Reject expressions involving
13287 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13288 (aarch64_classify_index): Handle SVE indices, by requiring
13289 a plain register index with a scale that matches the element size.
13290 (aarch64_classify_address): Handle SVE addresses. Assert that
13291 the mode of the address is VOIDmode or an integer mode.
13292 Update call to aarch64_classify_symbol.
13293 (aarch64_classify_symbolic_expression): Update call to
13294 aarch64_classify_symbol.
13295 (aarch64_const_vec_all_in_range_p): New function.
13296 (aarch64_print_vector_float_operand): Likewise.
13297 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13298 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13299 and the FP immediates 1.0 and 0.5.
13300 (aarch64_print_address_internal): Handle SVE addresses.
13301 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13302 (aarch64_regno_regclass): Handle predicate registers.
13303 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13305 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13306 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13307 (aarch64_convert_sve_vector_bits): New function.
13308 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13309 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13310 rather than an rtx.
13311 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13312 Handle SVE vector and predicate modes. Accept VL-based constants
13313 that need only one temporary register, and VL offsets that require
13314 no temporary registers.
13315 (aarch64_conditional_register_usage): Mark the predicate registers
13316 as fixed if SVE isn't available.
13317 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13318 Return true for SVE vector and predicate modes.
13319 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13320 rather than an unsigned int. Handle SVE modes.
13321 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13323 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13325 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13326 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13327 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13328 (aarch64_sve_float_mul_immediate_p): New functions.
13329 (aarch64_sve_valid_immediate): New function.
13330 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13331 Explicitly reject structure modes. Check for INDEX constants.
13332 Handle PTRUE and PFALSE constants.
13333 (aarch64_check_zero_based_sve_index_immediate): New function.
13334 (aarch64_simd_imm_zero_p): Delete.
13335 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13336 vector modes. Accept constants in the range of CNT[BHWD].
13337 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13338 ask for an Advanced SIMD mode.
13339 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13340 (aarch64_simd_vector_alignment): Handle SVE predicates.
13341 (aarch64_vectorize_preferred_vector_alignment): New function.
13342 (aarch64_simd_vector_alignment_reachable): Use it instead of
13344 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13345 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13347 (MAX_VECT_LEN): Delete.
13348 (expand_vec_perm_d): Add a vec_flags field.
13349 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13350 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13351 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13353 (aarch64_evpc_rev): Rename to...
13354 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13355 (aarch64_evpc_rev_global): New function.
13356 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13357 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13359 (aarch64_evpc_sve_tbl): New function.
13360 (aarch64_expand_vec_perm_const_1): Update after rename of
13361 aarch64_evpc_rev. Handle SVE permutes too, trying
13362 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13363 than aarch64_evpc_tbl.
13364 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13365 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13366 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13367 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13368 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13369 (aarch64_expand_sve_vcond): New functions.
13370 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13371 of aarch64_vector_mode_p.
13372 (aarch64_dwarf_poly_indeterminate_value): New function.
13373 (aarch64_compute_pressure_classes): Likewise.
13374 (aarch64_can_change_mode_class): Likewise.
13375 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13376 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13377 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13378 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13379 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13380 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13381 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13382 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13384 (Dn, Dl, Dr): Accept const as well as const_vector.
13385 (Dz): Likewise. Compare against CONST0_RTX.
13386 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13387 of "vector" where appropriate.
13388 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13389 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13390 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13391 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13392 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13393 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13394 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13395 (v_int_equiv): Extend to SVE modes.
13396 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13398 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13399 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13400 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13401 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13402 (SVE_COND_FP_CMP): New int iterators.
13403 (perm_hilo): Handle the new unpack unspecs.
13404 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13406 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13407 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13408 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13409 (aarch64_equality_operator, aarch64_constant_vector_operand)
13410 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13411 (aarch64_sve_nonimmediate_operand): Likewise.
13412 (aarch64_sve_general_operand): Likewise.
13413 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13414 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13415 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13416 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13417 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13418 (aarch64_sve_float_arith_immediate): Likewise.
13419 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13420 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13421 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13422 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13423 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13424 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13425 (aarch64_sve_float_arith_operand): Likewise.
13426 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13427 (aarch64_sve_float_mul_operand): Likewise.
13428 (aarch64_sve_vec_perm_operand): Likewise.
13429 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13430 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13431 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13432 as well as const_vector.
13433 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13434 in file. Use CONST0_RTX and CONSTM1_RTX.
13435 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13436 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13437 Use aarch64_simd_imm_zero.
13438 * config/aarch64/aarch64-sve.md: New file.
13439 * config/aarch64/aarch64.md: Include it.
13440 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13441 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13442 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13443 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13444 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13445 (sve): New attribute.
13446 (enabled): Disable instructions with the sve attribute unless
13448 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13449 aarch64_expand_mov_immediate.
13450 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13451 CNT[BHSD] immediates.
13452 (movti): Split CONST_POLY_INT moves into two halves.
13453 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13454 Split additions that need a temporary here if the destination
13455 is the stack pointer.
13456 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13457 (*add<mode>3_poly_1): New instruction.
13458 (set_clobber_cc): New expander.
13460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13462 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13463 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13464 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13465 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13466 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13467 Change innermode from fixed_mode_size to machine_mode.
13468 (simplify_subreg): Update call accordingly. Handle a constant-sized
13469 subreg of a variable-length CONST_VECTOR.
13471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13472 Alan Hayward <alan.hayward@arm.com>
13473 David Sherwood <david.sherwood@arm.com>
13475 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13476 (add_offset_to_base): New function, split out from...
13477 (create_mem_ref): ...here. When handling a scale other than 1,
13478 check first whether the address is valid without the offset.
13479 Add it into the base if so, leaving the index and scale as-is.
13481 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13484 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13485 fold_for_warn before checking if arg2 is INTEGER_CST.
13487 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13489 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13490 (store_multiple_operation): Delete.
13491 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13492 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13493 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13494 guarded by TARGET_STRING.
13495 (rs6000_output_load_multiple): Delete.
13496 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13497 OPTION_MASK_STRING / TARGET_STRING handling.
13498 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13499 (const rs6000_opt_masks) <"string">: Change mask to 0.
13500 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13501 (MASK_STRING): Delete.
13502 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13504 (load_multiple): Delete.
13511 (store_multiple): Delete.
13518 (movmemsi_8reg): Delete.
13519 (corresponding unnamed define_insn): Delete.
13520 (movmemsi_6reg): Delete.
13521 (corresponding unnamed define_insn): Delete.
13522 (movmemsi_4reg): Delete.
13523 (corresponding unnamed define_insn): Delete.
13524 (movmemsi_2reg): Delete.
13525 (corresponding unnamed define_insn): Delete.
13526 (movmemsi_1reg): Delete.
13527 (corresponding unnamed define_insn): Delete.
13528 * config/rs6000/rs6000.opt (mno-string): New.
13529 (mstring): Replace by deprecation warning stub.
13530 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13532 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13534 * regrename.c (regrename_do_replace): If replacing the same
13535 reg multiple times, try to reuse last created gen_raw_REG.
13538 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13539 main to workaround a bug in GDB.
13541 2018-01-12 Tom de Vries <tom@codesourcery.com>
13544 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13546 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13548 PR rtl-optimization/80481
13549 * ira-color.c (get_cap_member): New function.
13550 (allocnos_conflict_by_live_ranges_p): Use it.
13551 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13552 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13554 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13557 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13558 (*saddl_se_1): Ditto.
13559 (*ssubsi_1): Ditto.
13560 (*ssubl_se_1): Ditto.
13562 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13564 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13565 rather than wi::to_widest for DR_INITs.
13566 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13567 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13568 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13570 (vect_analyze_group_access_1): Note that here.
13572 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13574 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13575 polynomial type sizes.
13577 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13579 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13580 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13581 (gimple_add_tmp_var): Likewise.
13583 2018-01-12 Martin Liska <mliska@suse.cz>
13585 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13586 (gimple_alloc_sizes): Likewise.
13587 (dump_gimple_statistics): Use PRIu64 in printf format.
13588 * gimple.h: Change uint64_t to int.
13590 2018-01-12 Martin Liska <mliska@suse.cz>
13592 * tree-core.h: Use uint64_t instead of int.
13593 * tree.c (tree_node_counts): Likewise.
13594 (tree_node_sizes): Likewise.
13595 (dump_tree_statistics): Use PRIu64 in printf format.
13597 2018-01-12 Martin Liska <mliska@suse.cz>
13599 * Makefile.in: As qsort_chk is implemented in vec.c, add
13600 vec.o to linkage of gencfn-macros.
13601 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13602 passing the info to record_node_allocation_statistics.
13603 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13605 * ggc-common.c (struct ggc_usage): Add operator== and use
13606 it in operator< and compare function.
13607 * mem-stats.h (struct mem_usage): Likewise.
13608 * vec.c (struct vec_usage): Remove operator< and compare
13609 function. Can be simply inherited.
13611 2018-01-12 Martin Jambor <mjambor@suse.cz>
13614 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13615 * tree-ssa-math-opts.c: Include domwalk.h.
13616 (convert_mult_to_fma_1): New function.
13617 (fma_transformation_info): New type.
13618 (fma_deferring_state): Likewise.
13619 (cancel_fma_deferring): New function.
13620 (result_of_phi): Likewise.
13621 (last_fma_candidate_feeds_initial_phi): Likewise.
13622 (convert_mult_to_fma): Added deferring logic, split actual
13623 transformation to convert_mult_to_fma_1.
13624 (math_opts_dom_walker): New type.
13625 (math_opts_dom_walker::after_dom_children): New method, body moved
13626 here from pass_optimize_widening_mul::execute, added deferring logic
13628 (pass_optimize_widening_mul::execute): Moved most of code to
13629 math_opts_dom_walker::after_dom_children.
13630 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13631 * config/i386/i386.c (ix86_option_override_internal): Added
13632 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13634 2018-01-12 Richard Biener <rguenther@suse.de>
13637 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13638 inline instance vars.
13640 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13643 * config/rx/rx.c (rx_is_restricted_memory_address):
13644 Handle SUBREG case.
13646 2018-01-12 Richard Biener <rguenther@suse.de>
13648 PR tree-optimization/80846
13649 * target.def (split_reduction): New target hook.
13650 * targhooks.c (default_split_reduction): New function.
13651 * targhooks.h (default_split_reduction): Declare.
13652 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13653 target requests first reduce vectors by combining low and high
13655 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13656 (get_vectype_for_scalar_type_and_size): Export.
13657 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13658 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13659 * doc/tm.texi: Regenerate.
13660 * config/i386/i386.c (ix86_split_reduction): Implement
13661 TARGET_VECTORIZE_SPLIT_REDUCTION.
13663 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13666 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13667 in PIC mode except for TARGET_VXWORKS_RTP.
13668 * config/sparc/sparc.c: Include cfgrtl.h.
13669 (TARGET_INIT_PIC_REG): Define.
13670 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13671 (sparc_pic_register_p): New predicate.
13672 (sparc_legitimate_address_p): Use it.
13673 (sparc_legitimize_pic_address): Likewise.
13674 (sparc_delegitimize_address): Likewise.
13675 (sparc_mode_dependent_address_p): Likewise.
13676 (gen_load_pcrel_sym): Remove 4th parameter.
13677 (load_got_register): Adjust call to above. Remove obsolete stuff.
13678 (sparc_expand_prologue): Do not call load_got_register here.
13679 (sparc_flat_expand_prologue): Likewise.
13680 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13681 (sparc_use_pseudo_pic_reg): New function.
13682 (sparc_init_pic_reg): Likewise.
13683 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13684 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13686 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13688 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13689 Add item for branch_cost.
13691 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13693 PR rtl-optimization/83565
13694 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13695 not extend the result to a larger mode for rotate operations.
13696 (num_sign_bit_copies1): Likewise.
13698 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13701 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13703 Use values-Xc.o for -pedantic.
13704 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13706 2018-01-12 Martin Liska <mliska@suse.cz>
13709 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13711 (possible_polymorphic_call_targets): Use it.
13712 (ipa_devirt): Likewise.
13714 2018-01-12 Martin Liska <mliska@suse.cz>
13716 * profile-count.h (enum profile_quality): Use 0 as invalid
13717 enum value of profile_quality.
13719 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13721 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13722 -mext-string options.
13724 2018-01-12 Richard Biener <rguenther@suse.de>
13726 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13727 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13728 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13730 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13732 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13734 * configure.ac (--with-long-double-format): Add support for the
13735 configuration option to change the default long double format on
13737 * config.gcc (powerpc*-linux*-*): Likewise.
13738 * configure: Regenerate.
13739 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13740 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13741 used without modification.
13743 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13745 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13746 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13747 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13748 MISC_BUILTIN_SPEC_BARRIER.
13749 (rs6000_init_builtins): Likewise.
13750 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13752 (speculation_barrier): New define_insn.
13753 * doc/extend.texi: Document __builtin_speculation_barrier.
13755 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13758 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13759 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13760 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13762 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13763 integral modes instead of "ss" and "sd".
13764 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13765 vectors with 32-bit and 64-bit elements.
13766 (vecdupssescalarmodesuffix): New mode attribute.
13767 (vec_dup<mode>): Use it.
13769 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13772 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13773 frame if argument is passed on stack.
13775 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13778 * ree.c (combine_reaching_defs): Optimize also
13779 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13780 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13782 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13784 PR middle-end/83189
13785 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13787 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13789 PR middle-end/83718
13790 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13791 after they are computed.
13793 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13795 PR tree-optimization/83695
13796 * gimple-loop-linterchange.cc
13797 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13798 reset cached scev information after interchange.
13799 (pass_linterchange::execute): Remove call to scev_reset_htab.
13801 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13803 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13804 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13805 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13806 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13807 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13808 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13809 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13810 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13811 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13812 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13813 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13814 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13815 (V_lane_reg): Likewise.
13816 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13818 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13819 (vfmal_lane_low<mode>_intrinsic,
13820 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13821 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13822 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13823 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13824 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13825 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13827 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13829 * config/arm/arm-cpus.in (fp16fml): New feature.
13830 (ALL_SIMD): Add fp16fml.
13831 (armv8.2-a): Add fp16fml as an option.
13832 (armv8.3-a): Likewise.
13833 (armv8.4-a): Add fp16fml as part of fp16.
13834 * config/arm/arm.h (TARGET_FP16FML): Define.
13835 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13837 * config/arm/arm-modes.def (V2HF): Define.
13838 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13839 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13840 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13841 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13842 vfmsl_low, vfmsl_high): New set of builtins.
13843 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13844 (vfml_op): New code attribute.
13845 (VFMLHALVES): New int iterator.
13846 (VFML, VFMLSEL): New mode attributes.
13847 (V_reg): Define mapping for V2HF.
13848 (V_hi, V_lo): New mode attributes.
13849 (VF_constraint): Likewise.
13850 (vfml_half, vfml_half_selector): New int attributes.
13851 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13853 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13854 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13856 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13857 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13858 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13859 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13861 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13862 Document new effective target and option set.
13864 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13866 * config/arm/arm-cpus.in (armv8_4): New feature.
13867 (ARMv8_4a): New fgroup.
13868 (armv8.4-a): New arch.
13869 * config/arm/arm-tables.opt: Regenerate.
13870 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13871 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13872 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13873 Add matching rules for -march=armv8.4-a and extensions.
13874 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13876 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13879 * config/rx/rx.md (BW): New mode attribute.
13880 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13882 2018-01-11 Richard Biener <rguenther@suse.de>
13884 PR tree-optimization/83435
13885 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13886 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13887 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
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_address_info): Add a const_offset
13895 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13896 (aarch64_print_address_internal): Use it to check for a zero offset.
13898 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13899 Alan Hayward <alan.hayward@arm.com>
13900 David Sherwood <david.sherwood@arm.com>
13902 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13903 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13904 Return a poly_int64 rather than a HOST_WIDE_INT.
13905 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13906 rather than a HOST_WIDE_INT.
13907 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13908 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13909 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13910 final_offset from HOST_WIDE_INT to poly_int64.
13911 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13912 to_constant when getting the number of units in an Advanced SIMD
13914 (aarch64_builtin_vectorized_function): Check for a constant number
13916 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13918 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13919 attribute instead of GET_MODE_NUNITS.
13920 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13921 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13922 GET_MODE_SIZE for fixed-size registers.
13923 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13924 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13925 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13926 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
13927 (aarch64_print_operand, aarch64_print_address_internal)
13928 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
13929 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
13930 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
13931 Handle polynomial GET_MODE_SIZE.
13932 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
13933 wider than SImode without modification.
13934 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
13935 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
13936 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
13937 passing and returning SVE modes.
13938 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
13939 rather than GEN_INT.
13940 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
13941 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
13942 (aarch64_allocate_and_probe_stack_space): Likewise.
13943 (aarch64_layout_frame): Cope with polynomial offsets.
13944 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
13945 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
13946 polynomial offsets.
13947 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
13948 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
13949 poly_int64 rather than a HOST_WIDE_INT.
13950 (aarch64_get_separate_components, aarch64_process_components)
13951 (aarch64_expand_prologue, aarch64_expand_epilogue)
13952 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
13953 (aarch64_anchor_offset): New function, split out from...
13954 (aarch64_legitimize_address): ...here.
13955 (aarch64_builtin_vectorization_cost): Handle polynomial
13956 TYPE_VECTOR_SUBPARTS.
13957 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
13959 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
13960 number of elements from the PARALLEL rather than the mode.
13961 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
13962 rather than GET_MODE_BITSIZE.
13963 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
13964 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
13965 (aarch64_expand_vec_perm_const_1): Handle polynomial
13966 d->perm.length () and d->perm elements.
13967 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
13968 Apply to_constant to d->perm elements.
13969 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
13970 polynomial CONST_VECTOR_NUNITS.
13971 (aarch64_move_pointer): Take amount as a poly_int64 rather
13973 (aarch64_progress_pointer): Avoid temporary variable.
13974 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
13975 the mode attribute instead of GET_MODE.
13977 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13978 Alan Hayward <alan.hayward@arm.com>
13979 David Sherwood <david.sherwood@arm.com>
13981 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
13982 x exists before using it.
13983 (aarch64_add_constant_internal): Rename to...
13984 (aarch64_add_offset_1): ...this. Replace regnum with separate
13985 src and dest rtxes. Handle the case in which they're different,
13986 including when the offset is zero. Replace scratchreg with an rtx.
13987 Use 2 additions if there is no spare register into which we can
13988 move a 16-bit constant.
13989 (aarch64_add_constant): Delete.
13990 (aarch64_add_offset): Replace reg with separate src and dest
13991 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
13992 Use aarch64_add_offset_1.
13993 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
13994 an rtx rather than an int. Take the delta as a poly_int64
13995 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
13996 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
13997 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
13998 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
13999 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14000 and aarch64_add_sp.
14001 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14002 aarch64_add_constant.
14004 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14006 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14007 Use scalar_float_mode.
14009 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14011 * config/aarch64/aarch64-simd.md
14012 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14013 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14014 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14015 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14016 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14017 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14018 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14019 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14020 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14021 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14023 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14026 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14027 targ_options->x_arm_arch_string is non NULL.
14029 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14031 * config/aarch64/aarch64.h
14032 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14034 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14037 * expmed.c (emit_store_flag_force): Swap if const op0
14038 and change VOIDmode to mode of op0.
14040 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14042 PR rtl-optimization/83761
14043 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14044 than bytes to mode_for_size.
14046 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14048 PR middle-end/83189
14049 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14050 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14053 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14055 PR middle-end/83575
14056 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14057 when in layout mode.
14058 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14059 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14062 2018-01-10 Michael Collison <michael.collison@arm.com>
14064 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14065 * config/aarch64/aarch64-option-extension.def: Add
14066 AARCH64_OPT_EXTENSION of 'fp16fml'.
14067 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14068 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14069 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14070 * config/aarch64/constraints.md (Ui7): New constraint.
14071 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14072 (VFMLA_SEL_W): Ditto.
14075 (VFMLA16_LOW): New int iterator.
14076 (VFMLA16_HIGH): Ditto.
14077 (UNSPEC_FMLAL): New unspec.
14078 (UNSPEC_FMLSL): Ditto.
14079 (UNSPEC_FMLAL2): Ditto.
14080 (UNSPEC_FMLSL2): Ditto.
14081 (f16mac): New code attribute.
14082 * config/aarch64/aarch64-simd-builtins.def
14083 (aarch64_fmlal_lowv2sf): Ditto.
14084 (aarch64_fmlsl_lowv2sf): Ditto.
14085 (aarch64_fmlalq_lowv4sf): Ditto.
14086 (aarch64_fmlslq_lowv4sf): Ditto.
14087 (aarch64_fmlal_highv2sf): Ditto.
14088 (aarch64_fmlsl_highv2sf): Ditto.
14089 (aarch64_fmlalq_highv4sf): Ditto.
14090 (aarch64_fmlslq_highv4sf): Ditto.
14091 (aarch64_fmlal_lane_lowv2sf): Ditto.
14092 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14093 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14094 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14095 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14096 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14097 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14098 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14099 (aarch64_fmlal_lane_highv2sf): Ditto.
14100 (aarch64_fmlsl_lane_highv2sf): Ditto.
14101 (aarch64_fmlal_laneq_highv2sf): Ditto.
14102 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14103 (aarch64_fmlalq_lane_highv4sf): Ditto.
14104 (aarch64_fmlsl_lane_highv4sf): Ditto.
14105 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14106 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14107 * config/aarch64/aarch64-simd.md:
14108 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14109 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14110 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14111 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14112 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14113 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14114 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14115 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14116 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14117 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14118 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14119 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14120 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14121 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14122 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14123 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14124 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14125 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14126 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14127 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14128 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14129 (vfmlsl_low_u32): Ditto.
14130 (vfmlalq_low_u32): Ditto.
14131 (vfmlslq_low_u32): Ditto.
14132 (vfmlal_high_u32): Ditto.
14133 (vfmlsl_high_u32): Ditto.
14134 (vfmlalq_high_u32): Ditto.
14135 (vfmlslq_high_u32): Ditto.
14136 (vfmlal_lane_low_u32): Ditto.
14137 (vfmlsl_lane_low_u32): Ditto.
14138 (vfmlal_laneq_low_u32): Ditto.
14139 (vfmlsl_laneq_low_u32): Ditto.
14140 (vfmlalq_lane_low_u32): Ditto.
14141 (vfmlslq_lane_low_u32): Ditto.
14142 (vfmlalq_laneq_low_u32): Ditto.
14143 (vfmlslq_laneq_low_u32): Ditto.
14144 (vfmlal_lane_high_u32): Ditto.
14145 (vfmlsl_lane_high_u32): Ditto.
14146 (vfmlal_laneq_high_u32): Ditto.
14147 (vfmlsl_laneq_high_u32): Ditto.
14148 (vfmlalq_lane_high_u32): Ditto.
14149 (vfmlslq_lane_high_u32): Ditto.
14150 (vfmlalq_laneq_high_u32): Ditto.
14151 (vfmlslq_laneq_high_u32): Ditto.
14152 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14153 (AARCH64_FL_FOR_ARCH8_4): New.
14154 (AARCH64_ISA_F16FML): New ISA flag.
14155 (TARGET_F16FML): New feature flag for fp16fml.
14156 (doc/invoke.texi): Document new fp16fml option.
14158 2018-01-10 Michael Collison <michael.collison@arm.com>
14160 * config/aarch64/aarch64-builtins.c:
14161 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14162 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14163 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14164 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14165 (AARCH64_ISA_SHA3): New ISA flag.
14166 (TARGET_SHA3): New feature flag for sha3.
14167 * config/aarch64/iterators.md (sha512_op): New int attribute.
14168 (CRYPTO_SHA512): New int iterator.
14169 (UNSPEC_SHA512H): New unspec.
14170 (UNSPEC_SHA512H2): Ditto.
14171 (UNSPEC_SHA512SU0): Ditto.
14172 (UNSPEC_SHA512SU1): Ditto.
14173 * config/aarch64/aarch64-simd-builtins.def
14174 (aarch64_crypto_sha512hqv2di): New builtin.
14175 (aarch64_crypto_sha512h2qv2di): Ditto.
14176 (aarch64_crypto_sha512su0qv2di): Ditto.
14177 (aarch64_crypto_sha512su1qv2di): Ditto.
14178 (aarch64_eor3qv8hi): Ditto.
14179 (aarch64_rax1qv2di): Ditto.
14180 (aarch64_xarqv2di): Ditto.
14181 (aarch64_bcaxqv8hi): Ditto.
14182 * config/aarch64/aarch64-simd.md:
14183 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14184 (aarch64_crypto_sha512su0qv2di): Ditto.
14185 (aarch64_crypto_sha512su1qv2di): Ditto.
14186 (aarch64_eor3qv8hi): Ditto.
14187 (aarch64_rax1qv2di): Ditto.
14188 (aarch64_xarqv2di): Ditto.
14189 (aarch64_bcaxqv8hi): Ditto.
14190 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14191 (vsha512h2q_u64): Ditto.
14192 (vsha512su0q_u64): Ditto.
14193 (vsha512su1q_u64): Ditto.
14194 (veor3q_u16): Ditto.
14195 (vrax1q_u64): Ditto.
14196 (vxarq_u64): Ditto.
14197 (vbcaxq_u16): Ditto.
14198 * config/arm/types.md (crypto_sha512): New type attribute.
14199 (crypto_sha3): Ditto.
14200 (doc/invoke.texi): Document new sha3 option.
14202 2018-01-10 Michael Collison <michael.collison@arm.com>
14204 * config/aarch64/aarch64-builtins.c:
14205 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14206 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14207 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14208 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14209 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14210 (AARCH64_ISA_SM4): New ISA flag.
14211 (TARGET_SM4): New feature flag for sm4.
14212 * config/aarch64/aarch64-simd-builtins.def
14213 (aarch64_sm3ss1qv4si): Ditto.
14214 (aarch64_sm3tt1aq4si): Ditto.
14215 (aarch64_sm3tt1bq4si): Ditto.
14216 (aarch64_sm3tt2aq4si): Ditto.
14217 (aarch64_sm3tt2bq4si): Ditto.
14218 (aarch64_sm3partw1qv4si): Ditto.
14219 (aarch64_sm3partw2qv4si): Ditto.
14220 (aarch64_sm4eqv4si): Ditto.
14221 (aarch64_sm4ekeyqv4si): Ditto.
14222 * config/aarch64/aarch64-simd.md:
14223 (aarch64_sm3ss1qv4si): Ditto.
14224 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14225 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14226 (aarch64_sm4eqv4si): Ditto.
14227 (aarch64_sm4ekeyqv4si): Ditto.
14228 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14229 (sm3part_op): Ditto.
14230 (CRYPTO_SM3TT): Ditto.
14231 (CRYPTO_SM3PART): Ditto.
14232 (UNSPEC_SM3SS1): New unspec.
14233 (UNSPEC_SM3TT1A): Ditto.
14234 (UNSPEC_SM3TT1B): Ditto.
14235 (UNSPEC_SM3TT2A): Ditto.
14236 (UNSPEC_SM3TT2B): Ditto.
14237 (UNSPEC_SM3PARTW1): Ditto.
14238 (UNSPEC_SM3PARTW2): Ditto.
14239 (UNSPEC_SM4E): Ditto.
14240 (UNSPEC_SM4EKEY): Ditto.
14241 * config/aarch64/constraints.md (Ui2): New constraint.
14242 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14243 * config/arm/types.md (crypto_sm3): New type attribute.
14244 (crypto_sm4): Ditto.
14245 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14246 (vsm3tt1aq_u32): Ditto.
14247 (vsm3tt1bq_u32): Ditto.
14248 (vsm3tt2aq_u32): Ditto.
14249 (vsm3tt2bq_u32): Ditto.
14250 (vsm3partw1q_u32): Ditto.
14251 (vsm3partw2q_u32): Ditto.
14252 (vsm4eq_u32): Ditto.
14253 (vsm4ekeyq_u32): Ditto.
14254 (doc/invoke.texi): Document new sm4 option.
14256 2018-01-10 Michael Collison <michael.collison@arm.com>
14258 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14259 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14260 (AARCH64_FL_FOR_ARCH8_4): New.
14261 (AARCH64_FL_V8_4): New flag.
14262 (doc/invoke.texi): Document new armv8.4-a option.
14264 2018-01-10 Michael Collison <michael.collison@arm.com>
14266 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14267 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14268 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14269 * config/aarch64/aarch64-option-extension.def: Add
14270 AARCH64_OPT_EXTENSION of 'sha2'.
14271 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14272 (crypto): Disable sha2 and aes if crypto disabled.
14273 (crypto): Enable aes and sha2 if enabled.
14274 (simd): Disable sha2 and aes if simd disabled.
14275 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14277 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14278 (TARGET_SHA2): New feature flag for sha2.
14279 (TARGET_AES): New feature flag for aes.
14280 * config/aarch64/aarch64-simd.md:
14281 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14282 conditional on TARGET_AES.
14283 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14284 (aarch64_crypto_sha1hsi): Make pattern conditional
14286 (aarch64_crypto_sha1hv4si): Ditto.
14287 (aarch64_be_crypto_sha1hv4si): Ditto.
14288 (aarch64_crypto_sha1su1v4si): Ditto.
14289 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14290 (aarch64_crypto_sha1su0v4si): Ditto.
14291 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14292 (aarch64_crypto_sha256su0v4si): Ditto.
14293 (aarch64_crypto_sha256su1v4si): Ditto.
14294 (doc/invoke.texi): Document new aes and sha2 options.
14296 2018-01-10 Martin Sebor <msebor@redhat.com>
14298 PR tree-optimization/83781
14299 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14302 2018-01-11 Martin Sebor <msebor@gmail.com>
14303 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14305 PR tree-optimization/83501
14306 PR tree-optimization/81703
14308 * tree-ssa-strlen.c (get_string_cst): Rename...
14309 (get_string_len): ...to this. Handle global constants.
14310 (handle_char_store): Adjust.
14312 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14313 Jim Wilson <jimw@sifive.com>
14315 * config/riscv/riscv-protos.h (riscv_output_return): New.
14316 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14317 (riscv_attribute_table, riscv_output_return),
14318 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14319 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14320 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14322 (riscv_expand_prologue): Add early return for naked function.
14323 (riscv_expand_epilogue): Likewise.
14324 (riscv_function_ok_for_sibcall): Return false for naked function.
14325 (riscv_set_current_function): New.
14326 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14327 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14328 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14329 * doc/extend.texi (RISC-V Function Attributes): New.
14331 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14333 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14334 check for 128-bit long double before checking TCmode.
14335 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14336 128-bit long doubles before checking TFmode or TCmode.
14337 (FLOAT128_IBM_P): Likewise.
14339 2018-01-10 Martin Sebor <msebor@redhat.com>
14341 PR tree-optimization/83671
14342 * builtins.c (c_strlen): Unconditionally return zero for the empty
14344 Use -Warray-bounds for warnings.
14345 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14346 for non-constant array indices with COMPONENT_REF, arrays of
14347 arrays, and pointers to arrays.
14348 (gimple_fold_builtin_strlen): Determine and set length range for
14349 non-constant character arrays.
14351 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14353 PR middle-end/81897
14354 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14357 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14359 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14361 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14364 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14365 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14366 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14367 indexed_or_indirect_operand predicate.
14368 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14369 (*vsx_le_perm_load_v8hi): Likewise.
14370 (*vsx_le_perm_load_v16qi): Likewise.
14371 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14372 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14373 (*vsx_le_perm_store_v8hi): Likewise.
14374 (*vsx_le_perm_store_v16qi): Likewise.
14375 (eight unnamed splitters): Likewise.
14377 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14379 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14380 * config/rs6000/emmintrin.h: Likewise.
14381 * config/rs6000/mmintrin.h: Likewise.
14382 * config/rs6000/xmmintrin.h: Likewise.
14384 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14387 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14389 * tree.c (tree_nop_conversion): Return true for location wrapper
14391 (maybe_wrap_with_location): New function.
14392 (selftest::check_strip_nops): New function.
14393 (selftest::test_location_wrappers): New function.
14394 (selftest::tree_c_tests): Call it.
14395 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14396 (maybe_wrap_with_location): New decl.
14397 (EXPR_LOCATION_WRAPPER_P): New macro.
14398 (location_wrapper_p): New inline function.
14399 (tree_strip_any_location_wrapper): New inline function.
14401 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14404 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14405 stack_realign_offset for the largest alignment of stack slot
14407 (ix86_find_max_used_stack_alignment): New function.
14408 (ix86_finalize_stack_frame_flags): Use it. Set
14409 max_used_stack_alignment if we don't realign stack.
14410 * config/i386/i386.h (machine_function): Add
14411 max_used_stack_alignment.
14413 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14415 * config/arm/arm.opt (-mbranch-cost): New option.
14416 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14419 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14422 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14423 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14425 2018-01-10 Richard Biener <rguenther@suse.de>
14428 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14429 early out so it also covers the case where we have a non-NULL
14432 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14434 PR tree-optimization/83753
14435 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14436 for non-strided grouped accesses if the number of elements is 1.
14438 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14441 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14442 * i386.h (TARGET_USE_GATHER): Define.
14443 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14445 2018-01-10 Martin Liska <mliska@suse.cz>
14448 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14449 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14451 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14452 CLEANUP_NO_PARTITIONING is not set.
14454 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14456 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14457 for vectors, as a partial revert of r254296.
14458 * rtl.h (const_vec_p): Delete.
14459 (const_vec_duplicate_p): Don't test for vector CONSTs.
14460 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14461 * expmed.c (make_tree): Likewise.
14464 * common.md (E, F): Use CONSTANT_P instead of checking for
14466 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14467 checking for CONST_VECTOR.
14469 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14471 PR middle-end/83575
14472 * predict.c (force_edge_cold): Handle in more sane way edges
14473 with no prediction.
14475 2018-01-09 Carl Love <cel@us.ibm.com>
14477 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14479 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14480 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14481 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14482 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14483 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14484 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14485 * config/rs6000/rs6000-protos.h: Add extern defition for
14486 rs6000_generate_float2_double_code.
14487 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14489 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14490 (float2_v2df): Add define_expand.
14492 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14495 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14496 op_mode in the force_to_mode call.
14498 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14500 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14501 instead of checking each element individually.
14502 (aarch64_evpc_uzp): Likewise.
14503 (aarch64_evpc_zip): Likewise.
14504 (aarch64_evpc_ext): Likewise.
14505 (aarch64_evpc_rev): Likewise.
14506 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14507 instead of checking each element individually. Return true without
14509 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14510 whether all selected elements come from the same input, instead of
14511 checking each element individually. Remove calls to gen_rtx_REG,
14512 start_sequence and end_sequence and instead assert that no rtl is
14515 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14517 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14518 order of HIGH and CONST checks.
14520 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14522 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14523 if the destination isn't an SSA_NAME.
14525 2018-01-09 Richard Biener <rguenther@suse.de>
14527 PR tree-optimization/83668
14528 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14530 (canonicalize_loop_form): ... here, renamed from ...
14531 (canonicalize_loop_closed_ssa_form): ... this and amended to
14532 swap successor edges for loop exit blocks to make us use
14533 the RPO order we need for initial schedule generation.
14535 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14537 PR tree-optimization/64811
14538 * match.pd: When optimizing comparisons with Inf, avoid
14539 introducing or losing exceptions from comparisons with NaN.
14541 2018-01-09 Martin Liska <mliska@suse.cz>
14544 * asan.c (shadow_mem_size): Add gcc_assert.
14546 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14548 Don't save registers in main().
14551 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14552 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14553 * config/avr/avr.c (avr_set_current_function): Don't error if
14554 naked, OS_task or OS_main are specified at the same time.
14555 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14557 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14559 * common/config/avr/avr-common.c (avr_option_optimization_table):
14560 Switch on -mmain-is-OS_task for optimizing compilations.
14562 2018-01-09 Richard Biener <rguenther@suse.de>
14564 PR tree-optimization/83572
14565 * graphite.c: Include cfganal.h.
14566 (graphite_transform_loops): Connect infinite loops to exit
14567 and remove fake edges at the end.
14569 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14571 * ipa-inline.c (edge_badness): Revert accidental checkin.
14573 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14576 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14577 symbols; not inline clones.
14579 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14582 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14583 hard registers. Formatting fixes.
14585 PR preprocessor/83722
14586 * gcc.c (try_generate_repro): Pass
14587 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14588 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14591 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14592 Kito Cheng <kito.cheng@gmail.com>
14594 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14595 (riscv_leaf_function_p): Delete.
14596 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14598 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14600 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14602 (do_ifelse): New function.
14603 (do_isel): New function.
14604 (do_sub3): New function.
14605 (do_add3): New function.
14606 (do_load_mask_compare): New function.
14607 (do_overlap_load_compare): New function.
14608 (expand_compare_loop): New function.
14609 (expand_block_compare): Call expand_compare_loop() when appropriate.
14610 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14611 option description.
14612 (-mblock-compare-inline-loop-limit): New option.
14614 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14617 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14618 Reverse order of second and third operands in first alternative.
14619 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14620 of first and second elements in UNSPEC_VPERMR vector.
14621 (altivec_expand_vec_perm_le): Likewise.
14623 2018-01-08 Jeff Law <law@redhat.com>
14625 PR rtl-optimizatin/81308
14626 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14627 (process_switch): If group_case_labels makes a change, then set
14629 (pass_convert_switch::execute): If a switch is converted, then
14630 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14632 PR rtl-optimization/81308
14633 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14636 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14638 PR target/83663 - Revert r255946
14639 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14640 generation for cases where splatting a value is not useful.
14641 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14642 across a vec_duplicate and a paradoxical subreg forming a vector
14643 mode to a vec_concat.
14645 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14647 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14648 -march=armv8.3-a variants.
14649 * config/arm/t-multilib: Likewise.
14650 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14652 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14654 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14656 (cceq_ior_compare_complement): Give it a name so I can use it, and
14657 change boolean_or_operator predicate to boolean_operator so it can
14658 be used to generate a crand.
14659 (eqne): New code iterator.
14660 (bd/bd_neg): New code_attrs.
14661 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14662 a single define_insn.
14663 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14664 decrement (bdnzt/bdnzf/bdzt/bdzf).
14665 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14666 with the new names of the branch decrement patterns, and added the
14667 names of the branch decrement conditional patterns.
14669 2018-01-08 Richard Biener <rguenther@suse.de>
14671 PR tree-optimization/83563
14672 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14675 2018-01-08 Richard Biener <rguenther@suse.de>
14677 PR middle-end/83713
14678 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14680 2018-01-08 Richard Biener <rguenther@suse.de>
14682 PR tree-optimization/83685
14683 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14684 references to abnormals.
14686 2018-01-08 Richard Biener <rguenther@suse.de>
14689 * dwarf2out.c (output_indirect_strings): Handle empty
14690 skeleton_debug_str_hash.
14691 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14693 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14695 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14696 (emit_store_direct): Likewise.
14697 (arc_trampoline_adjust_address): Likewise.
14698 (arc_asm_trampoline_template): New function.
14699 (arc_initialize_trampoline): Use asm_trampoline_template.
14700 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14701 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14702 * config/arc/arc.md (flush_icache): Delete pattern.
14704 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14706 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14707 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14710 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14713 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14714 by not USED_FOR_TARGET.
14715 (make_pass_resolve_sw_modes): Likewise.
14717 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14719 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14722 2018-01-08 Richard Biener <rguenther@suse.de>
14724 PR middle-end/83580
14725 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14727 2018-01-08 Richard Biener <rguenther@suse.de>
14729 PR middle-end/83517
14730 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14732 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14734 PR middle-end/81897
14735 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14736 basic blocks with a small number of successors.
14737 (convert_control_dep_chain_into_preds): Improve handling of
14739 (dump_predicates): Split apart into...
14740 (dump_pred_chain): ...here...
14741 (dump_pred_info): ...and here.
14742 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14743 (can_chain_union_be_invalidated_p): Improve check for invalidation
14745 (uninit_uses_cannot_happen): Avoid unnecessary if
14746 convert_control_dep_chain_into_preds yielded nothing.
14748 2018-01-06 Martin Sebor <msebor@redhat.com>
14750 PR tree-optimization/83640
14751 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14752 subtracting negative offset from size.
14753 (builtin_access::overlap): Adjust offset bounds of the access to fall
14754 within the size of the object if possible.
14756 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14758 PR rtl-optimization/83699
14759 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14760 extract_bit_field_as_subreg to cases in which the extracted
14761 value is also a vector.
14763 * lra-constraints.c (process_alt_operands): Test for the equivalence
14764 substitutions when detecting a possible reload cycle.
14766 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14769 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14770 by default if flag_selective_schedling{,2}. Formatting fixes.
14772 PR rtl-optimization/83682
14773 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14774 if it has non-VECTOR_MODE element mode.
14775 (vec_duplicate_p): Likewise.
14777 PR middle-end/83694
14778 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14779 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14781 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14784 * config/i386/i386-builtin.def
14785 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14786 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14787 Require also OPTION_MASK_ISA_AVX512F in addition to
14788 OPTION_MASK_ISA_GFNI.
14789 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14790 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14791 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14792 to OPTION_MASK_ISA_GFNI.
14793 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14794 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14795 OPTION_MASK_ISA_AVX512BW.
14796 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14797 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14798 addition to OPTION_MASK_ISA_GFNI.
14799 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14800 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14801 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14802 to OPTION_MASK_ISA_GFNI.
14803 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14804 a requirement for all ISAs rather than any of them with a few
14806 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14808 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14809 bitmasks to be enabled with 3 exceptions, instead of requiring any
14810 enabled ISA with lots of exceptions.
14811 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14812 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14813 Change avx512bw in isa attribute to avx512f.
14814 * config/i386/sgxintrin.h: Add license boilerplate.
14815 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14816 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14817 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14818 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14820 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14821 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14822 temporarily sse2 rather than sse if not enabled already.
14825 * config/i386/sse.md (VI248_VLBW): Rename to ...
14826 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14827 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14828 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14829 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14830 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14831 mode iterator instead of VI248_VLBW.
14833 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14835 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14836 (record_modified): Skip clobbers; add debug output.
14837 (param_change_prob): Use sreal frequencies.
14839 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14841 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14842 punt for user-aligned variables.
14844 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14846 * tree-chrec.c (chrec_contains_symbols): Return true for
14849 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14852 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14853 of (x|y) == x for BICS pattern.
14855 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14857 PR tree-optimization/83605
14858 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14859 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14862 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14864 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14865 * config/epiphany/rtems.h: New file.
14867 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14868 Uros Bizjak <ubizjak@gmail.com>
14871 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14872 QIreg_operand instead of register_operand predicate.
14873 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14874 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14875 comments instead of -fmitigate[-_]rop.
14877 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14880 * cgraphunit.c (symbol_table::compile): Switch to text_section
14881 before calling assembly_start debug hook.
14882 * run-rtl-passes.c (run_rtl_passes): Likewise.
14885 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14887 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14888 range_int_cst_p rather than !symbolic_range_p before calling
14889 extract_range_from_multiplicative_op_1.
14891 2018-01-04 Jeff Law <law@redhat.com>
14893 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14894 redundant test in assertion.
14896 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14898 * doc/rtl.texi: Document machine_mode wrapper classes.
14900 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14902 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14903 using tree_to_uhwi.
14905 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14907 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14908 the VEC_PERM_EXPR fold to fail.
14910 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14913 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14914 to switched_sections.
14916 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14919 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14920 test for d.testing.
14922 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14925 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14926 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
14928 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14931 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
14932 is BLKmode and bitpos not zero or mode change is needed.
14934 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14937 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
14940 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
14943 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
14944 instead of MULT rtx. Update all corresponding splitters.
14945 (*saddl_se): Ditto.
14946 (*ssub<modesuffix>): Ditto.
14947 (*ssubl_se): Ditto.
14948 (*cmp_sadd_di): Update split patterns.
14949 (*cmp_sadd_si): Ditto.
14950 (*cmp_sadd_sidi): Ditto.
14951 (*cmp_ssub_di): Ditto.
14952 (*cmp_ssub_si): Ditto.
14953 (*cmp_ssub_sidi): Ditto.
14954 * config/alpha/predicates.md (const23_operand): New predicate.
14955 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
14956 Look for ASHIFT, not MULT inner operand.
14957 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
14959 2018-01-04 Martin Liska <mliska@suse.cz>
14961 PR gcov-profile/83669
14962 * gcov.c (output_intermediate_file): Add version to intermediate
14964 * doc/gcov.texi: Document new field 'version' in intermediate
14965 file format. Fix location of '-k' option of gcov command.
14967 2018-01-04 Martin Liska <mliska@suse.cz>
14970 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
14972 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14974 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
14976 2018-01-03 Martin Sebor <msebor@redhat.com>
14978 PR tree-optimization/83655
14979 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
14980 checking calls with invalid arguments.
14982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14984 * tree-vect-stmts.c (vect_get_store_rhs): New function.
14985 (vectorizable_mask_load_store): Delete.
14986 (vectorizable_call): Return false for masked loads and stores.
14987 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
14988 instead of gimple_assign_rhs1.
14989 (vectorizable_load): Handle IFN_MASK_LOAD.
14990 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
14992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14994 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
14996 (vectorizable_mask_load_store): ...here.
14997 (vectorizable_load): ...and here.
14999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15001 * tree-vect-stmts.c (vect_build_all_ones_mask)
15002 (vect_build_zero_merge_argument): New functions, split out from...
15003 (vectorizable_load): ...here.
15005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15007 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15009 (vectorizable_mask_load_store): ...here.
15010 (vectorizable_store): ...and here.
15012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15014 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15016 (vectorizable_mask_load_store): ...here.
15018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15020 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15021 (vect_model_store_cost): Take a vec_load_store_type instead of a
15023 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15024 (vect_model_store_cost): Take a vec_load_store_type instead of a
15026 (vectorizable_mask_load_store): Update accordingly.
15027 (vectorizable_store): Likewise.
15028 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15032 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15033 IFN_MASK_LOAD calls here rather than...
15034 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15037 Alan Hayward <alan.hayward@arm.com>
15038 David Sherwood <david.sherwood@arm.com>
15040 * expmed.c (extract_bit_field_1): For vector extracts,
15041 fall back to extract_bit_field_as_subreg if vec_extract
15044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15045 Alan Hayward <alan.hayward@arm.com>
15046 David Sherwood <david.sherwood@arm.com>
15048 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15049 they are variable or constant sized.
15050 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15051 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15058 handling COND_EXPRs with boolean comparisons, try to find a better
15059 basis for the mask type than the boolean itself.
15061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15063 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15064 is calculated and how it can be overridden.
15065 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15066 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15068 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15072 Alan Hayward <alan.hayward@arm.com>
15073 David Sherwood <david.sherwood@arm.com>
15075 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15076 Remove the mode argument.
15077 (aarch64_simd_valid_immediate): Remove the mode and inverse
15079 * config/aarch64/iterators.md (bitsize): New iterator.
15080 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15081 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15082 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15083 aarch64_simd_valid_immediate.
15084 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15085 (aarch64_reg_or_bic_imm): Likewise.
15086 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15087 with an insn_type enum and msl with a modifier_type enum.
15088 Replace element_width with a scalar_mode. Change the shift
15089 to unsigned int. Add constructors for scalar_float_mode and
15090 scalar_int_mode elements.
15091 (aarch64_vect_float_const_representable_p): Delete.
15092 (aarch64_can_const_movi_rtx_p)
15093 (aarch64_simd_scalar_immediate_valid_for_move)
15094 (aarch64_simd_make_constant): Update call to
15095 aarch64_simd_valid_immediate.
15096 (aarch64_advsimd_valid_immediate_hs): New function.
15097 (aarch64_advsimd_valid_immediate): Likewise.
15098 (aarch64_simd_valid_immediate): Remove mode and inverse
15099 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15100 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15101 and aarch64_float_const_representable_p on the result.
15102 (aarch64_output_simd_mov_immediate): Remove mode argument.
15103 Update call to aarch64_simd_valid_immediate and use of
15104 simd_immediate_info.
15105 (aarch64_output_scalar_simd_mov_immediate): Update call
15108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15109 Alan Hayward <alan.hayward@arm.com>
15110 David Sherwood <david.sherwood@arm.com>
15112 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15113 (mode_nunits): Likewise CONST_MODE_NUNITS.
15114 * machmode.def (ADJUST_NUNITS): Document.
15115 * genmodes.c (mode_data::need_nunits_adj): New field.
15116 (blank_mode): Update accordingly.
15117 (adj_nunits): New variable.
15118 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15120 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15121 listed in adj_nunits.
15122 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15123 listed in adj_nunits. Don't emit case statements for such modes.
15124 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15125 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15126 nothing if adj_nunits is nonnull.
15127 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15128 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15129 (emit_mode_fbit): Update use of print_maybe_const_decl.
15130 (emit_move_size): Likewise. Treat the array as non-const
15132 (emit_mode_adjustments): Handle adj_nunits.
15134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15136 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15137 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15138 (VECTOR_MODES): Use it.
15139 (make_vector_modes): Take the prefix as an argument.
15141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15142 Alan Hayward <alan.hayward@arm.com>
15143 David Sherwood <david.sherwood@arm.com>
15145 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15146 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15147 for MODE_VECTOR_BOOL.
15148 * machmode.def (VECTOR_BOOL_MODE): Document.
15149 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15150 (make_vector_bool_mode): New function.
15151 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15153 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15154 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15156 * stor-layout.c (int_mode_for_mode): Likewise.
15157 * tree.c (build_vector_type_for_mode): Likewise.
15158 * varasm.c (output_constant_pool_2): Likewise.
15159 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15160 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15161 for MODE_VECTOR_BOOL.
15162 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15163 of mode class checks.
15164 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15165 instead of a list of mode class checks.
15166 (expand_vector_scalar_condition): Likewise.
15167 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15170 Alan Hayward <alan.hayward@arm.com>
15171 David Sherwood <david.sherwood@arm.com>
15173 * machmode.h (mode_size): Change from unsigned short to
15175 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15176 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15177 or if measurement_type is not polynomial.
15178 (fixed_size_mode::includes_p): Check for constant-sized modes.
15179 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15180 return a poly_uint16 rather than an unsigned short.
15181 (emit_mode_size): Change the type of mode_size from unsigned short
15182 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15183 (emit_mode_adjustments): Cope with polynomial vector sizes.
15184 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15186 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15188 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15189 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15190 * caller-save.c (setup_save_areas): Likewise.
15191 (replace_reg_with_saved_mem): Likewise.
15192 * calls.c (emit_library_call_value_1): Likewise.
15193 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15194 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15195 (gen_lowpart_for_combine): Likewise.
15196 * convert.c (convert_to_integer_1): Likewise.
15197 * cse.c (equiv_constant, cse_insn): Likewise.
15198 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15199 (cselib_subst_to_values): Likewise.
15200 * dce.c (word_dce_process_block): Likewise.
15201 * df-problems.c (df_word_lr_mark_ref): Likewise.
15202 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15203 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15204 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15205 (rtl_for_decl_location): Likewise.
15206 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15207 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15208 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15209 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15210 (expand_expr_real_1): Likewise.
15211 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15212 (pad_below): Likewise.
15213 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15214 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15215 * ira.c (get_subreg_tracking_sizes): Likewise.
15216 * ira-build.c (ira_create_allocno_objects): Likewise.
15217 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15218 (ira_sort_regnos_for_alter_reg): Likewise.
15219 * ira-costs.c (record_operand_costs): Likewise.
15220 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15221 (resolve_simple_move): Likewise.
15222 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15223 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15224 (lra_constraints): Likewise.
15225 (CONST_POOL_OK_P): Reject variable-sized modes.
15226 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15227 (add_pseudo_to_slot, lra_spill): Likewise.
15228 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15229 * optabs-query.c (get_best_extraction_insn): Likewise.
15230 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15231 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15232 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15233 * recog.c (offsettable_address_addr_space_p): Likewise.
15234 * regcprop.c (maybe_mode_change): Likewise.
15235 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15236 * regrename.c (build_def_use): Likewise.
15237 * regstat.c (dump_reg_info): Likewise.
15238 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15239 (find_reloads, find_reloads_subreg_address): Likewise.
15240 * reload1.c (eliminate_regs_1): Likewise.
15241 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15242 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15243 (simplify_binary_operation_1, simplify_subreg): Likewise.
15244 * targhooks.c (default_function_arg_padding): Likewise.
15245 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15246 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15247 (verify_gimple_assign_ternary): Likewise.
15248 * tree-inline.c (estimate_move_cost): Likewise.
15249 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15250 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15251 (get_address_cost_ainc): Likewise.
15252 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15253 (vect_supportable_dr_alignment): Likewise.
15254 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15255 (vectorizable_reduction): Likewise.
15256 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15257 (vectorizable_operation, vectorizable_load): Likewise.
15258 * tree.c (build_same_sized_truth_vector_type): Likewise.
15259 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15260 * var-tracking.c (emit_note_insn_var_location): Likewise.
15261 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15262 (ADDR_VEC_ALIGN): Likewise.
15264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15265 Alan Hayward <alan.hayward@arm.com>
15266 David Sherwood <david.sherwood@arm.com>
15268 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15270 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15271 or if measurement_type is polynomial.
15272 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15273 * combine.c (make_extraction): Likewise.
15274 * dse.c (find_shift_sequence): Likewise.
15275 * dwarf2out.c (mem_loc_descriptor): Likewise.
15276 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15277 (extract_bit_field, extract_low_bits): Likewise.
15278 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15279 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15280 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15281 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15282 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15283 * reload.c (find_reloads): Likewise.
15284 * reload1.c (alter_reg): Likewise.
15285 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15286 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15287 * tree-if-conv.c (predicate_mem_writes): Likewise.
15288 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15289 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15290 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15291 * valtrack.c (dead_debug_insert_temp): Likewise.
15292 * varasm.c (mergeable_constant_section): Likewise.
15293 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15296 Alan Hayward <alan.hayward@arm.com>
15297 David Sherwood <david.sherwood@arm.com>
15299 * expr.c (expand_assignment): Cope with polynomial mode sizes
15300 when assigning to a CONCAT.
15302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15303 Alan Hayward <alan.hayward@arm.com>
15304 David Sherwood <david.sherwood@arm.com>
15306 * machmode.h (mode_precision): Change from unsigned short to
15308 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15310 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15311 or if measurement_type is not polynomial.
15312 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15313 in which the mode is already known to be a scalar_int_mode.
15314 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15315 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15317 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15318 for GET_MODE_PRECISION.
15319 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15320 for GET_MODE_PRECISION.
15321 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15323 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15324 (expand_field_assignment, make_extraction): Likewise.
15325 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15326 (get_last_value): Likewise.
15327 * convert.c (convert_to_integer_1): Likewise.
15328 * cse.c (cse_insn): Likewise.
15329 * expr.c (expand_expr_real_1): Likewise.
15330 * lra-constraints.c (simplify_operand_subreg): Likewise.
15331 * optabs-query.c (can_atomic_load_p): Likewise.
15332 * optabs.c (expand_atomic_load): Likewise.
15333 (expand_atomic_store): Likewise.
15334 * ree.c (combine_reaching_defs): Likewise.
15335 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15336 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15337 * tree.h (type_has_mode_precision_p): Likewise.
15338 * ubsan.c (instrument_si_overflow): 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 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15345 polynomial numbers of units.
15346 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15347 (valid_vector_subparts_p): New function.
15348 (build_vector_type): Remove temporary shim and take the number
15349 of units as a poly_uint64 rather than an int.
15350 (build_opaque_vector_type): Take the number of units as a
15351 poly_uint64 rather than an int.
15352 * tree.c (build_vector_from_ctor): Handle polynomial
15353 TYPE_VECTOR_SUBPARTS.
15354 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15355 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15356 (build_vector_from_val): If the number of units is variable,
15357 use build_vec_duplicate_cst for constant operands and
15358 VEC_DUPLICATE_EXPR otherwise.
15359 (make_vector_type): Remove temporary is_constant ().
15360 (build_vector_type, build_opaque_vector_type): Take the number of
15361 units as a poly_uint64 rather than an int.
15362 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15364 * cfgexpand.c (expand_debug_expr): Likewise.
15365 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15366 (store_constructor, expand_expr_real_1): Likewise.
15367 (const_scalar_mask_from_tree): Likewise.
15368 * fold-const-call.c (fold_const_reduction): Likewise.
15369 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15370 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15371 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15372 (fold_relational_const): Likewise.
15373 (native_interpret_vector): Likewise. Change the size from an
15374 int to an unsigned int.
15375 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15376 TYPE_VECTOR_SUBPARTS.
15377 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15378 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15379 duplicating a non-constant operand into a variable-length vector.
15380 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15381 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15382 * ipa-icf.c (sem_variable::equals): Likewise.
15383 * match.pd: Likewise.
15384 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15385 * print-tree.c (print_node): Likewise.
15386 * stor-layout.c (layout_type): Likewise.
15387 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15388 * tree-cfg.c (verify_gimple_comparison): Likewise.
15389 (verify_gimple_assign_binary): Likewise.
15390 (verify_gimple_assign_ternary): Likewise.
15391 (verify_gimple_assign_single): Likewise.
15392 * tree-pretty-print.c (dump_generic_node): Likewise.
15393 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15394 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15395 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15396 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15397 (vect_shift_permute_load_chain): Likewise.
15398 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15399 (expand_vector_condition, optimize_vector_constructor): Likewise.
15400 (lower_vec_perm, get_compute_type): Likewise.
15401 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15402 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15403 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15404 (vect_recog_mask_conversion_pattern): Likewise.
15405 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15406 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15407 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15408 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15409 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15410 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15411 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15412 (supportable_widening_operation): Likewise.
15413 (supportable_narrowing_operation): Likewise.
15414 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15416 * varasm.c (output_constant): Likewise.
15418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15419 Alan Hayward <alan.hayward@arm.com>
15420 David Sherwood <david.sherwood@arm.com>
15422 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15423 so that both the length == 3 and length != 3 cases set up their
15424 own permute vectors. Add comments explaining why we know the
15425 number of elements is constant.
15426 (vect_permute_load_chain): Likewise.
15428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15429 Alan Hayward <alan.hayward@arm.com>
15430 David Sherwood <david.sherwood@arm.com>
15432 * machmode.h (mode_nunits): Change from unsigned char to
15434 (ONLY_FIXED_SIZE_MODES): New macro.
15435 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15436 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15437 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15439 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15440 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15441 or if measurement_type is not polynomial.
15442 * genmodes.c (ZERO_COEFFS): New macro.
15443 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15445 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15446 Use ZERO_COEFFS when emitting initializers.
15447 * data-streamer.h (bp_pack_poly_value): New function.
15448 (bp_unpack_poly_value): Likewise.
15449 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15450 for GET_MODE_NUNITS.
15451 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15452 for GET_MODE_NUNITS.
15453 * tree.c (make_vector_type): Remove temporary shim and make
15454 the real function take the number of units as a poly_uint64
15455 rather than an int.
15456 (build_vector_type_for_mode): Handle polynomial nunits.
15457 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15458 * emit-rtl.c (const_vec_series_p_1): Likewise.
15459 (gen_rtx_CONST_VECTOR): Likewise.
15460 * fold-const.c (test_vec_duplicate_folding): Likewise.
15461 * genrecog.c (validate_pattern): Likewise.
15462 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15463 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15464 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15465 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15466 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15467 * rtlanal.c (subreg_get_info): Likewise.
15468 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15469 (vect_grouped_load_supported): Likewise.
15470 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15471 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15472 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15473 (simplify_const_unary_operation, simplify_binary_operation_1)
15474 (simplify_const_binary_operation, simplify_ternary_operation)
15475 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15476 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15477 instead of CONST_VECTOR_NUNITS.
15478 * varasm.c (output_constant_pool_2): Likewise.
15479 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15480 explicit-encoded elements in the XVEC for variable-length vectors.
15482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15484 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15487 Alan Hayward <alan.hayward@arm.com>
15488 David Sherwood <david.sherwood@arm.com>
15490 * coretypes.h (fixed_size_mode): Declare.
15491 (fixed_size_mode_pod): New typedef.
15492 * builtins.h (target_builtins::x_apply_args_mode)
15493 (target_builtins::x_apply_result_mode): Change type to
15494 fixed_size_mode_pod.
15495 * builtins.c (apply_args_size, apply_result_size, result_vector)
15496 (expand_builtin_apply_args_1, expand_builtin_apply)
15497 (expand_builtin_return): Update accordingly.
15499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15501 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15502 * cselib.c (cselib_hash_rtx): Likewise.
15503 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15504 CONST_VECTOR encoding.
15506 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15507 Jeff Law <law@redhat.com>
15510 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15511 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15512 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15513 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15516 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15517 explicitly probe *sp in a noreturn function if there were any callee
15518 register saves or frame pointer is needed.
15520 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15523 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15524 BLKmode for ternary, binary or unary expressions.
15527 * var-tracking.c (delete_vta_debug_insn): New inline function.
15528 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15529 insns from get_insns () to NULL instead of each bb separately.
15530 Use delete_vta_debug_insn. No longer static.
15531 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15532 delete_vta_debug_insns callers.
15533 * rtl.h (delete_vta_debug_insns): Declare.
15534 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15535 instead of variable_tracking_main.
15537 2018-01-03 Martin Sebor <msebor@redhat.com>
15539 PR tree-optimization/83603
15540 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15541 arguments past the endof the argument list in functions declared
15542 without a prototype.
15543 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15544 Avoid checking when arguments are null.
15546 2018-01-03 Martin Sebor <msebor@redhat.com>
15549 * doc/extend.texi (attribute const): Fix a typo.
15550 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15551 issuing -Wsuggest-attribute for void functions.
15553 2018-01-03 Martin Sebor <msebor@redhat.com>
15555 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15556 offset_int::from instead of wide_int::to_shwi.
15557 (maybe_diag_overlap): Remove assertion.
15558 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15559 * gimple-ssa-sprintf.c (format_directive): Same.
15560 (parse_directive): Same.
15561 (sprintf_dom_walker::compute_format_length): Same.
15562 (try_substitute_return_value): Same.
15564 2018-01-03 Jeff Law <law@redhat.com>
15566 PR middle-end/83654
15567 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15568 non-constant residual for zero at runtime and avoid probing in
15569 that case. Reorganize code for trailing problem to mirror handling
15572 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15574 PR tree-optimization/83501
15575 * tree-ssa-strlen.c (get_string_cst): New.
15576 (handle_char_store): Call get_string_cst.
15578 2018-01-03 Martin Liska <mliska@suse.cz>
15580 PR tree-optimization/83593
15581 * tree-ssa-strlen.c: Include tree-cfg.h.
15582 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15583 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15584 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15586 (strlen_dom_walker::before_dom_children): Call
15587 gimple_purge_dead_eh_edges. Dump tranformation with details
15589 (strlen_dom_walker::before_dom_children): Update call by adding
15590 new argument cleanup_eh.
15591 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15593 2018-01-03 Martin Liska <mliska@suse.cz>
15596 * cif-code.def (VARIADIC_THUNK): New enum value.
15597 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15600 2018-01-03 Jan Beulich <jbeulich@suse.com>
15602 * sse.md (mov<mode>_internal): Tighten condition for when to use
15603 vmovdqu<ssescalarsize> for TI and OI modes.
15605 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15607 Update copyright years.
15609 2018-01-03 Martin Liska <mliska@suse.cz>
15612 * ipa-visibility.c (function_and_variable_visibility): Skip
15613 functions with noipa attribure.
15615 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15617 * gcc.c (process_command): Update copyright notice dates.
15618 * gcov-dump.c (print_version): Ditto.
15619 * gcov.c (print_version): Ditto.
15620 * gcov-tool.c (print_version): Ditto.
15621 * gengtype.c (create_file): Ditto.
15622 * doc/cpp.texi: Bump @copying's copyright year.
15623 * doc/cppinternals.texi: Ditto.
15624 * doc/gcc.texi: Ditto.
15625 * doc/gccint.texi: Ditto.
15626 * doc/gcov.texi: Ditto.
15627 * doc/install.texi: Ditto.
15628 * doc/invoke.texi: Ditto.
15630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15632 * vector-builder.h (vector_builder::m_full_nelts): Change from
15633 unsigned int to poly_uint64.
15634 (vector_builder::full_nelts): Update prototype accordingly.
15635 (vector_builder::new_vector): Likewise.
15636 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15637 (vector_builder::operator ==): Likewise.
15638 (vector_builder::finalize): Likewise.
15639 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15640 Take the number of elements as a poly_uint64 rather than an
15642 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15643 from unsigned int to poly_uint64.
15644 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15645 (vec_perm_indices::new_vector): Likewise.
15646 (vec_perm_indices::length): Likewise.
15647 (vec_perm_indices::nelts_per_input): Likewise.
15648 (vec_perm_indices::input_nelts): Likewise.
15649 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15650 number of elements per input as a poly_uint64 rather than an
15651 unsigned int. Use the original encoding for variable-length
15652 vectors, rather than clamping each individual element.
15653 For the second and subsequent elements in each pattern,
15654 clamp the step and base before clamping their sum.
15655 (vec_perm_indices::series_p): Handle polynomial element counts.
15656 (vec_perm_indices::all_in_range_p): Likewise.
15657 (vec_perm_indices_to_tree): Likewise.
15658 (vec_perm_indices_to_rtx): Likewise.
15659 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15660 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15661 (tree_vector_builder::new_binary_operation): Handle polynomial
15662 element counts. Return false if we need to know the number
15663 of elements at compile time.
15664 * fold-const.c (fold_vec_perm): Punt if the number of elements
15665 isn't known at compile time.
15667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15669 * vec-perm-indices.h (vec_perm_builder): Change element type
15670 from HOST_WIDE_INT to poly_int64.
15671 (vec_perm_indices::element_type): Update accordingly.
15672 (vec_perm_indices::clamp): Handle polynomial element_types.
15673 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15674 (vec_perm_indices::all_in_range_p): Likewise.
15675 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15677 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15678 polynomial vec_perm_indices element types.
15679 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15680 * fold-const.c (fold_vec_perm): Likewise.
15681 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15682 * tree-vect-generic.c (lower_vec_perm): Likewise.
15683 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15684 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15685 element type to HOST_WIDE_INT.
15687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15688 Alan Hayward <alan.hayward@arm.com>
15689 David Sherwood <david.sherwood@arm.com>
15691 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15692 rather than an int. Use plus_constant.
15693 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15694 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15697 Alan Hayward <alan.hayward@arm.com>
15698 David Sherwood <david.sherwood@arm.com>
15700 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15701 a HOST_WIDE_INT to a poly_int64.
15703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15704 Alan Hayward <alan.hayward@arm.com>
15705 David Sherwood <david.sherwood@arm.com>
15707 * calls.c (load_register_parameters): Cope with polynomial
15708 mode sizes. Require a constant size for BLKmode parameters
15709 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15710 forces a parameter to be padded at the lsb end in order to
15711 fill a complete number of words, require the parameter size
15712 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
15719 from unsigned int to poly_uint64_pod.
15720 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
15727 (reload_inner_reg_of_subreg, push_reload): Use it.
15729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15730 Alan Hayward <alan.hayward@arm.com>
15731 David Sherwood <david.sherwood@arm.com>
15733 * lra-constraints.c (process_alt_operands): Reject matched
15734 operands whose sizes aren't ordered.
15735 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15742 that the mode size is in the set {1, 2, 4, 8, 16}.
15744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15745 Alan Hayward <alan.hayward@arm.com>
15746 David Sherwood <david.sherwood@arm.com>
15748 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15749 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15756 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15757 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15758 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15759 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15760 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15761 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15762 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15763 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15764 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15766 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15767 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15768 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15769 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15770 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15771 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15772 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15773 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15774 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15775 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15777 * expr.c (emit_move_resolve_push): Treat the input and result
15778 of PUSH_ROUNDING as a poly_int64.
15779 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15780 (emit_push_insn): Likewise.
15781 * lra-eliminations.c (mark_not_eliminable): Likewise.
15782 * recog.c (push_operand): Likewise.
15783 * reload1.c (elimination_effects): Likewise.
15784 * rtlanal.c (nonzero_bits1): Likewise.
15785 * calls.c (store_one_arg): Likewise. Require the padding to be
15786 known at compile time.
15788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15789 Alan Hayward <alan.hayward@arm.com>
15790 David Sherwood <david.sherwood@arm.com>
15792 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15793 Use plus_constant instead of gen_rtx_PLUS.
15795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15796 Alan Hayward <alan.hayward@arm.com>
15797 David Sherwood <david.sherwood@arm.com>
15799 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15800 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15807 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15808 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
15815 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15816 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15817 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15826 Alan Hayward <alan.hayward@arm.com>
15827 David Sherwood <david.sherwood@arm.com>
15829 * combine.c (can_change_dest_mode): Handle polynomial
15830 REGMODE_NATURAL_SIZE.
15831 * expmed.c (store_bit_field_1): Likewise.
15832 * expr.c (store_constructor): Likewise.
15833 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15834 and polynomial REGMODE_NATURAL_SIZE.
15835 (gen_lowpart_common): Likewise.
15836 * reginfo.c (record_subregs_of_mode): Likewise.
15837 * rtlanal.c (read_modify_subreg_p): Likewise.
15839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15840 Alan Hayward <alan.hayward@arm.com>
15841 David Sherwood <david.sherwood@arm.com>
15843 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15844 numbers of elements.
15846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15847 Alan Hayward <alan.hayward@arm.com>
15848 David Sherwood <david.sherwood@arm.com>
15850 * match.pd: Cope with polynomial numbers of vector elements.
15852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15853 Alan Hayward <alan.hayward@arm.com>
15854 David Sherwood <david.sherwood@arm.com>
15856 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15857 in a POINTER_PLUS_EXPR.
15859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15860 Alan Hayward <alan.hayward@arm.com>
15861 David Sherwood <david.sherwood@arm.com>
15863 * omp-simd-clone.c (simd_clone_subparts): New function.
15864 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15865 (ipa_simd_modify_function_body): Likewise.
15867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15868 Alan Hayward <alan.hayward@arm.com>
15869 David Sherwood <david.sherwood@arm.com>
15871 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15872 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15873 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15874 (expand_vector_condition, vector_element): Likewise.
15875 (subparts_gt): New function.
15876 (get_compute_type): Use subparts_gt.
15877 (count_type_subparts): Delete.
15878 (expand_vector_operations_1): Use subparts_gt instead of
15879 count_type_subparts.
15881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15882 Alan Hayward <alan.hayward@arm.com>
15883 David Sherwood <david.sherwood@arm.com>
15885 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15886 (vect_compile_time_alias): ...this new function. Do the calculation
15887 on poly_ints rather than trees.
15888 (vect_prune_runtime_alias_test_list): Update call accordingly.
15890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15891 Alan Hayward <alan.hayward@arm.com>
15892 David Sherwood <david.sherwood@arm.com>
15894 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15896 (vect_schedule_slp_instance): Likewise.
15898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15899 Alan Hayward <alan.hayward@arm.com>
15900 David Sherwood <david.sherwood@arm.com>
15902 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15903 constant and extern definitions for variable-length vectors.
15904 (vect_get_constant_vectors): Note that the number of units
15905 is known to be constant.
15907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15908 Alan Hayward <alan.hayward@arm.com>
15909 David Sherwood <david.sherwood@arm.com>
15911 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15912 of units as polynomial. Choose between WIDE and NARROW based
15915 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15916 Alan Hayward <alan.hayward@arm.com>
15917 David Sherwood <david.sherwood@arm.com>
15919 * tree-vect-stmts.c (simd_clone_subparts): New function.
15920 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15923 Alan Hayward <alan.hayward@arm.com>
15924 David Sherwood <david.sherwood@arm.com>
15926 * tree-vect-stmts.c (vectorizable_call): Treat the number of
15927 vectors as polynomial. Use build_index_vector for
15928 IFN_GOMP_SIMD_LANE.
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 * tree-vect-stmts.c (get_load_store_type): Treat the number of
15935 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15936 for variable-length vectors.
15937 (vectorizable_mask_load_store): Treat the number of units as
15938 polynomial, asserting that it is constant if the condition has
15939 already been enforced.
15940 (vectorizable_store, vectorizable_load): Likewise.
15942 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15943 Alan Hayward <alan.hayward@arm.com>
15944 David Sherwood <david.sherwood@arm.com>
15946 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
15947 of units as polynomial. Punt if we can't tell at compile time
15948 which vector contains the final result.
15950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15951 Alan Hayward <alan.hayward@arm.com>
15952 David Sherwood <david.sherwood@arm.com>
15954 * tree-vect-loop.c (vectorizable_induction): Treat the number
15955 of units as polynomial. Punt on SLP inductions. Use an integer
15956 VEC_SERIES_EXPR for variable-length integer reductions. Use a
15957 cast of such a series for variable-length floating-point
15960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15961 Alan Hayward <alan.hayward@arm.com>
15962 David Sherwood <david.sherwood@arm.com>
15964 * tree.h (build_index_vector): Declare.
15965 * tree.c (build_index_vector): New function.
15966 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
15967 of units as polynomial, forcibly converting it to a constant if
15968 vectorizable_reduction has already enforced the condition.
15969 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
15970 to create a {1,2,3,...} vector.
15971 (vectorizable_reduction): Treat the number of units as polynomial.
15972 Choose vectype_in based on the largest scalar element size rather
15973 than the smallest number of units. Enforce the restrictions
15976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15977 Alan Hayward <alan.hayward@arm.com>
15978 David Sherwood <david.sherwood@arm.com>
15980 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
15981 number of units as polynomial.
15983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15984 Alan Hayward <alan.hayward@arm.com>
15985 David Sherwood <david.sherwood@arm.com>
15987 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
15988 * target.def (autovectorize_vector_sizes): Return the vector sizes
15989 by pointer, using vector_sizes rather than a bitmask.
15990 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
15991 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
15992 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
15994 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
15995 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
15996 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
15997 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
15998 * omp-general.c (omp_max_vf): Likewise.
15999 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16000 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16001 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16002 * tree-vect-slp.c (vect_slp_bb): Likewise.
16003 * doc/tm.texi: Regenerate.
16004 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16006 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16007 the vector size as a poly_uint64 rather than an unsigned int.
16008 (current_vector_size): Change from an unsigned int to a poly_uint64.
16009 (get_vectype_for_scalar_type): Update accordingly.
16010 * tree.h (build_truth_vector_type): Take the size and number of
16011 units as a poly_uint64 rather than an unsigned int.
16012 (build_vector_type): Add a temporary overload that takes
16013 the number of units as a poly_uint64 rather than an unsigned int.
16014 * tree.c (make_vector_type): Likewise.
16015 (build_truth_vector_type): Take the number of units as a poly_uint64
16016 rather than an unsigned int.
16018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16019 Alan Hayward <alan.hayward@arm.com>
16020 David Sherwood <david.sherwood@arm.com>
16022 * target.def (get_mask_mode): Take the number of units and length
16023 as poly_uint64s rather than unsigned ints.
16024 * targhooks.h (default_get_mask_mode): Update accordingly.
16025 * targhooks.c (default_get_mask_mode): Likewise.
16026 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16027 * doc/tm.texi: Regenerate.
16029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16030 Alan Hayward <alan.hayward@arm.com>
16031 David Sherwood <david.sherwood@arm.com>
16033 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16034 * omp-general.c (omp_max_vf): Likewise.
16035 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16036 (expand_omp_simd): Handle polynomial safelen.
16037 * omp-low.c (omplow_simd_context): Add a default constructor.
16038 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16039 (lower_rec_simd_input_clauses): Update accordingly.
16040 (lower_rec_input_clauses): Likewise.
16042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16043 Alan Hayward <alan.hayward@arm.com>
16044 David Sherwood <david.sherwood@arm.com>
16046 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16047 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16048 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16049 (vect_analyze_slp_cost): Likewise.
16050 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16051 (vect_model_load_cost): Likewise.
16053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16054 Alan Hayward <alan.hayward@arm.com>
16055 David Sherwood <david.sherwood@arm.com>
16057 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16058 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16059 from an unsigned int * to a poly_uint64_pod *.
16060 (calculate_unrolling_factor): New function.
16061 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16064 Alan Hayward <alan.hayward@arm.com>
16065 David Sherwood <david.sherwood@arm.com>
16067 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16068 from an unsigned int to a poly_uint64.
16069 (_loop_vec_info::slp_unrolling_factor): Likewise.
16070 (_loop_vec_info::vectorization_factor): Change from an int
16072 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16073 (vect_get_num_vectors): New function.
16074 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16075 (vect_get_num_copies): Use vect_get_num_vectors.
16076 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16077 to an unsigned int *.
16078 (vect_analyze_data_refs): Change min_vf from an int * to a
16080 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16081 than an unsigned HOST_WIDE_INT.
16082 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16083 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16084 to an unsigned int *.
16085 (vect_analyze_data_ref_dependences): Likewise.
16086 (vect_compute_data_ref_alignment): Handle polynomial vf.
16087 (vect_enhance_data_refs_alignment): Likewise.
16088 (vect_prune_runtime_alias_test_list): Likewise.
16089 (vect_shift_permute_load_chain): Likewise.
16090 (vect_supportable_dr_alignment): Likewise.
16091 (dependence_distance_ge_vf): Take the vectorization factor as a
16092 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16093 (vect_analyze_data_refs): Change min_vf from an int * to a
16095 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16096 vfm1 as a poly_uint64 rather than an int. Make the same change
16097 for the returned bound_scalar.
16098 (vect_gen_vector_loop_niters): Handle polynomial vf.
16099 (vect_do_peeling): Likewise. Update call to
16100 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16101 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16103 * tree-vect-loop.c (vect_determine_vectorization_factor)
16104 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16105 (vect_get_known_peeling_cost): Likewise.
16106 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16107 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16108 (vect_transform_loop): Likewise. Use the lowest possible VF when
16109 updating the upper bounds of the loop.
16110 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16111 rather than an int.
16112 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16113 polynomial unroll factors.
16114 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16115 (vect_make_slp_decision): Likewise.
16116 (vect_supported_load_permutation_p): Likewise, and polynomial
16118 (vect_analyze_slp_cost): Handle polynomial vf.
16119 (vect_slp_analyze_node_operations): Likewise.
16120 (vect_slp_analyze_bb_1): Likewise.
16121 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16122 than an unsigned HOST_WIDE_INT.
16123 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16124 (vectorizable_load): Handle polynomial vf.
16125 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16127 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16130 Alan Hayward <alan.hayward@arm.com>
16131 David Sherwood <david.sherwood@arm.com>
16133 * match.pd: Handle bit operations involving three constants
16134 and try to fold one pair.
16136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16138 * tree-vect-loop-manip.c: Include gimple-fold.h.
16139 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16140 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16141 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16142 Add a path that uses a step of VF instead of 1, but disable it
16144 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16145 and niters_no_overflow parameters. Update calls to
16146 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16147 Create a new SSA name if the latter choses to use a ste other
16148 than zero, and return it via niters_vector_mult_vf_var.
16149 * tree-vect-loop.c (vect_transform_loop): Update calls to
16150 vect_do_peeling, vect_gen_vector_loop_niters and
16151 slpeel_make_loop_iterate_ntimes.
16152 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16153 (vect_gen_vector_loop_niters): Update declarations after above changes.
16155 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16157 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16158 128-bit round to integer instructions.
16159 (ceil<mode>2): Likewise.
16160 (btrunc<mode>2): Likewise.
16161 (round<mode>2): Likewise.
16163 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16165 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16166 unaligned VSX load/store on P8/P9.
16167 (expand_block_clear): Allow the use of unaligned VSX
16168 load/store on P8/P9.
16170 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16172 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16174 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16175 swap associated with both a load and a store.
16177 2018-01-02 Andrew Waterman <andrew@sifive.com>
16179 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16180 * config/riscv/riscv.md (clear_cache): Use it.
16182 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16184 * web.c: Remove out-of-date comment.
16186 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16188 * expr.c (fixup_args_size_notes): Check that any existing
16189 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16190 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16191 (emit_single_push_insn): ...here.
16193 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16195 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16196 (const_vector_encoded_nelts): New function.
16197 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16198 (const_vector_int_elt, const_vector_elt): Declare.
16199 * emit-rtl.c (const_vector_int_elt_1): New function.
16200 (const_vector_elt): Likewise.
16201 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16202 of CONST_VECTOR_ELT.
16204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16206 * expr.c: Include rtx-vector-builder.h.
16207 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16208 directly on the tree encoding.
16209 (const_vector_from_tree): Likewise.
16210 * optabs.c: Include rtx-vector-builder.h.
16211 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16212 sequence of "u" values.
16213 * vec-perm-indices.c: Include rtx-vector-builder.h.
16214 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16215 directly on the vec_perm_indices encoding.
16217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16219 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16220 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16221 * rtx-vector-builder.h: New file.
16222 * rtx-vector-builder.c: Likewise.
16223 * rtl.h (rtx_def::u2): Add a const_vector field.
16224 (CONST_VECTOR_NPATTERNS): New macro.
16225 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16226 (CONST_VECTOR_DUPLICATE_P): Likewise.
16227 (CONST_VECTOR_STEPPED_P): Likewise.
16228 (CONST_VECTOR_ENCODED_ELT): Likewise.
16229 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16230 (unwrap_const_vec_duplicate): Likewise.
16231 (const_vec_series_p): Check for a non-duplicated vector encoding.
16232 Say that the function only returns true for integer vectors.
16233 * emit-rtl.c: Include rtx-vector-builder.h.
16234 (gen_const_vec_duplicate_1): Delete.
16235 (gen_const_vector): Call gen_const_vec_duplicate instead of
16236 gen_const_vec_duplicate_1.
16237 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16238 (gen_const_vec_duplicate): Use rtx_vector_builder.
16239 (gen_const_vec_series): Likewise.
16240 (gen_rtx_CONST_VECTOR): Likewise.
16241 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16242 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16243 Build a new vector rather than modifying a CONST_VECTOR in-place.
16244 (handle_special_swappables): Update call accordingly.
16245 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16246 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16247 Build a new vector rather than modifying a CONST_VECTOR in-place.
16248 (handle_special_swappables): Update call accordingly.
16250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16252 * simplify-rtx.c (simplify_const_binary_operation): Use
16253 CONST_VECTOR_ELT instead of XVECEXP.
16255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16257 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16258 the selector elements to be different from the data elements
16259 if the selector is a VECTOR_CST.
16260 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16261 ssizetype for the selector.
16263 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16265 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16266 before testing each element individually.
16267 * tree-vect-generic.c (lower_vec_perm): Likewise.
16269 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16271 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16272 * selftest-run-tests.c (selftest::run_tests): Call it.
16273 * vector-builder.h (vector_builder::operator ==): New function.
16274 (vector_builder::operator !=): Likewise.
16275 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16276 (vec_perm_indices::all_from_input_p): New function.
16277 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16278 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16279 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16280 instead of reading the VECTOR_CST directly. Detect whether both
16281 vector inputs are the same before constructing the vec_perm_indices,
16282 and update the number of inputs argument accordingly. Use the
16283 utility functions added above. Only construct sel2 if we need to.
16285 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16287 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16288 the broadcast of the low byte.
16289 (expand_mult_highpart): Use an explicit encoding for the permutes.
16290 * optabs-query.c (can_mult_highpart_p): Likewise.
16291 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16292 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16293 (vectorizable_bswap): Likewise.
16294 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16295 explicit encoding for the power-of-2 permutes.
16296 (vect_permute_store_chain): Likewise.
16297 (vect_grouped_load_supported): Likewise.
16298 (vect_permute_load_chain): Likewise.
16300 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16302 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16303 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16304 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16305 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16306 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16307 (vect_gen_perm_mask_any): Likewise.
16309 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16311 * int-vector-builder.h: New file.
16312 * vec-perm-indices.h: Include int-vector-builder.h.
16313 (vec_perm_indices): Redefine as an int_vector_builder.
16314 (auto_vec_perm_indices): Delete.
16315 (vec_perm_builder): Redefine as a stand-alone class.
16316 (vec_perm_indices::vec_perm_indices): New function.
16317 (vec_perm_indices::clamp): Likewise.
16318 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16319 (vec_perm_indices::new_vector): New function.
16320 (vec_perm_indices::new_expanded_vector): Update for new
16321 vec_perm_indices class.
16322 (vec_perm_indices::rotate_inputs): New function.
16323 (vec_perm_indices::all_in_range_p): Operate directly on the
16324 encoded form, without computing elided elements.
16325 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16326 encoding. Update for new vec_perm_indices class.
16327 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16328 the given vec_perm_builder.
16329 (expand_vec_perm_var): Update vec_perm_builder constructor.
16330 (expand_mult_highpart): Use vec_perm_builder instead of
16331 auto_vec_perm_indices.
16332 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16333 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16334 or double series encoding as appropriate.
16335 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16336 vec_perm_indices instead of auto_vec_perm_indices.
16337 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16338 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16339 (vect_permute_store_chain): Likewise.
16340 (vect_grouped_load_supported): Likewise.
16341 (vect_permute_load_chain): Likewise.
16342 (vect_shift_permute_load_chain): Likewise.
16343 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16344 (vect_transform_slp_perm_load): Likewise.
16345 (vect_schedule_slp_instance): Likewise.
16346 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16347 (vectorizable_mask_load_store): Likewise.
16348 (vectorizable_bswap): Likewise.
16349 (vectorizable_store): Likewise.
16350 (vectorizable_load): Likewise.
16351 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16352 vec_perm_indices instead of auto_vec_perm_indices. Use
16353 tree_to_vec_perm_builder to read the vector from a tree.
16354 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16355 vec_perm_builder instead of a vec_perm_indices.
16356 (have_whole_vector_shift): Use vec_perm_builder and
16357 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16358 truncation to calc_vec_perm_mask_for_shift.
16359 (vect_create_epilog_for_reduction): Likewise.
16360 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16361 from auto_vec_perm_indices to vec_perm_indices.
16362 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16363 instead of changing individual elements.
16364 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16365 the vector in d.perm.
16366 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16367 from auto_vec_perm_indices to vec_perm_indices.
16368 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16369 instead of changing individual elements.
16370 (arm_vectorize_vec_perm_const): Use new_vector to install
16371 the vector in d.perm.
16372 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16373 Update vec_perm_builder constructor.
16374 (rs6000_expand_interleave): Likewise.
16375 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16376 (rs6000_expand_interleave): Likewise.
16378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16380 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16381 to qimode could truncate the indices.
16382 * optabs.c (expand_vec_perm_var): Likewise.
16384 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16386 * Makefile.in (OBJS): Add vec-perm-indices.o.
16387 * vec-perm-indices.h: New file.
16388 * vec-perm-indices.c: Likewise.
16389 * target.h (vec_perm_indices): Replace with a forward class
16391 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16392 * optabs.h: Include vec-perm-indices.h.
16393 (expand_vec_perm): Delete.
16394 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16395 (expand_vec_perm_const): Declare.
16396 * target.def (vec_perm_const_ok): Replace with...
16397 (vec_perm_const): ...this new hook.
16398 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16399 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16400 * doc/tm.texi: Regenerate.
16401 * optabs.def (vec_perm_const): Delete.
16402 * doc/md.texi (vec_perm_const): Likewise.
16403 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16404 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16405 expand_vec_perm for constant permutation vectors. Assert that
16406 the mode of variable permutation vectors is the integer equivalent
16407 of the mode that is being permuted.
16408 * optabs-query.h (selector_fits_mode_p): Declare.
16409 * optabs-query.c: Include vec-perm-indices.h.
16410 (selector_fits_mode_p): New function.
16411 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16412 is defined, instead of checking whether the vec_perm_const_optab
16413 exists. Use targetm.vectorize.vec_perm_const instead of
16414 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16415 fit in the vector mode before using a variable permute.
16416 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16417 vec_perm_indices instead of an rtx.
16418 (expand_vec_perm): Replace with...
16419 (expand_vec_perm_const): ...this new function. Take the selector
16420 as a vec_perm_indices rather than an rtx. Also take the mode of
16421 the selector. Update call to shift_amt_for_vec_perm_mask.
16422 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16423 Use vec_perm_indices::new_expanded_vector to expand the original
16424 selector into bytes. Check whether the indices fit in the vector
16425 mode before using a variable permute.
16426 (expand_vec_perm_var): Make global.
16427 (expand_mult_highpart): Use expand_vec_perm_const.
16428 * fold-const.c: Includes vec-perm-indices.h.
16429 * tree-ssa-forwprop.c: Likewise.
16430 * tree-vect-data-refs.c: Likewise.
16431 * tree-vect-generic.c: Likewise.
16432 * tree-vect-loop.c: Likewise.
16433 * tree-vect-slp.c: Likewise.
16434 * tree-vect-stmts.c: Likewise.
16435 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16437 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16438 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16439 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16440 (aarch64_vectorize_vec_perm_const): ...this new function.
16441 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16442 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16443 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16444 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16445 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16446 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16447 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16449 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16450 check for NEON modes.
16451 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16452 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16453 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16454 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16456 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16457 the old VEC_PERM_CONST conditions.
16458 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16459 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16460 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16461 (ia64_vectorize_vec_perm_const_ok): Merge into...
16462 (ia64_vectorize_vec_perm_const): ...this new function.
16463 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16464 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16465 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16466 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16467 * config/mips/mips.c (mips_expand_vec_perm_const)
16468 (mips_vectorize_vec_perm_const_ok): Merge into...
16469 (mips_vectorize_vec_perm_const): ...this new function.
16470 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16471 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16472 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16473 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16474 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16475 (rs6000_expand_vec_perm_const): Delete.
16476 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16478 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16479 (altivec_expand_vec_perm_const_le): Take each operand individually.
16480 Operate on constant selectors rather than rtxes.
16481 (altivec_expand_vec_perm_const): Likewise. Update call to
16482 altivec_expand_vec_perm_const_le.
16483 (rs6000_expand_vec_perm_const): Delete.
16484 (rs6000_vectorize_vec_perm_const_ok): Delete.
16485 (rs6000_vectorize_vec_perm_const): New function.
16486 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16487 an element count and rtx array.
16488 (rs6000_expand_extract_even): Update call accordingly.
16489 (rs6000_expand_interleave): Likewise.
16490 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16491 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16492 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16493 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16494 (rs6000_expand_vec_perm_const): Delete.
16495 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16496 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16497 (altivec_expand_vec_perm_const_le): Take each operand individually.
16498 Operate on constant selectors rather than rtxes.
16499 (altivec_expand_vec_perm_const): Likewise. Update call to
16500 altivec_expand_vec_perm_const_le.
16501 (rs6000_expand_vec_perm_const): Delete.
16502 (rs6000_vectorize_vec_perm_const_ok): Delete.
16503 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16504 reference to the SPE evmerge intructions.
16505 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16506 an element count and rtx array.
16507 (rs6000_expand_extract_even): Update call accordingly.
16508 (rs6000_expand_interleave): Likewise.
16509 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16510 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16512 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16514 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16516 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16517 vector mode and that that mode matches the mode of the data
16519 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16520 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16521 directly using expand_vec_perm_1 when forcing selectors into
16523 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16525 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16527 * optabs-query.h (can_vec_perm_p): Delete.
16528 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16529 * optabs-query.c (can_vec_perm_p): Split into...
16530 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16531 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16532 particular selector is valid.
16533 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16534 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16535 (vect_grouped_load_supported): Likewise.
16536 (vect_shift_permute_load_chain): Likewise.
16537 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16538 (vect_transform_slp_perm_load): Likewise.
16539 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16540 (vectorizable_bswap): Likewise.
16541 (vect_gen_perm_mask_checked): Likewise.
16542 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16543 implementations of variable permutation vectors into account
16544 when deciding which selector to use.
16545 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16546 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16547 with a false third argument.
16548 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16549 to test whether the constant selector is valid and can_vec_perm_var_p
16550 to test whether a variable selector is valid.
16552 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16554 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16555 * optabs-query.c (can_vec_perm_p): Likewise.
16556 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16557 instead of vec_perm_indices.
16558 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16559 (vect_gen_perm_mask_checked): Likewise,
16560 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16561 (vect_gen_perm_mask_checked): Likewise,
16563 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16565 * optabs-query.h (qimode_for_vec_perm): Declare.
16566 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16567 (qimode_for_vec_perm): ...this new function.
16568 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16570 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16572 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16573 does not have a conditional at the top.
16575 2018-01-02 Richard Biener <rguenther@suse.de>
16577 * ipa-inline.c (big_speedup_p): Fix expression.
16579 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16582 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16585 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16589 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16590 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16591 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16592 cond_taken_branch_cost 3->4.
16594 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16596 PR tree-optimization/83581
16597 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16598 TODO_cleanup_cfg if any changes have been made.
16600 PR middle-end/83608
16601 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16602 convert_modes if target mode has the right side, but different mode
16605 PR middle-end/83609
16606 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16607 last argument when extracting from CONCAT. If either from_real or
16608 from_imag is NULL, use expansion through memory. If result is not
16609 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16610 the parts directly to inner mode, if even that fails, use expansion
16613 PR middle-end/83623
16614 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16615 check for bswap in mode rather than HImode and use that in expand_unop
16618 Copyright (C) 2018 Free Software Foundation, Inc.
16620 Copying and distribution of this file, with or without modification,
16621 are permitted in any medium without royalty provided the copyright
16622 notice and this notice are preserved.