1 2018-06-15 Richard Biener <rguenther@suse.de>
4 * tree-cfg.c (move_stmt_op): unshare invariant addresses
5 before adjusting their block.
7 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
10 multilibs for *-*-rtems*.
11 * config/riscv/t-rtems: New file.
13 2018-06-14 Jakub Jelinek <jakub@redhat.com>
16 * match.pd ((A +- CST1) +- CST2): Punt if last resort
17 unsigned_type_for returns NULL.
20 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
21 subregs of multi-word pseudos unless the float mode has word size.
23 2018-06-14 Richard Biener <rguenther@suse.de>
26 * tree-vect-generic.c (build_word_mode_vector_type): Remove
27 duplicate and harmful type_hash_canon.
28 * tree.c (type_hash_canon): Assert we didn't find ourselves.
30 2018-06-14 Richard Biener <rguenther@suse.de>
33 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
36 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
38 * config/rtems.h (STDINT_LONG32): Define.
40 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
41 Prachi Godbole <prachi.godbole@imgtec.com>
43 * config/mips/mips-cpus.def: Define P6600.
44 * config/mips/mips-tables.opt: Regenerate.
45 * config/mips/mips.c (mips_ucbranch_type): New enum.
46 (mips_rtx_cost_data): Add support for P6600.
47 (mips_issue_rate): Likewise.
48 (mips_multipass_dfa_lookahead): Likewise.
49 (mips_avoid_hazard): Likewise.
50 (mips_reorg_process_insns): Likewise.
51 (mips_classify_branch_p6600): New function.
52 * config/mips/mips.h (TUNE_P6600): New define.
53 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
54 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
55 * config/mips/mips.md: Include p6600.md.
56 (processor): Add p6600.
57 * config/mips/p6600.md: New file.
58 * doc/invoke.texi: Add p6600 to supported architectures.
60 2018-06-13 Martin Sebor <msebor@redhat.com>
62 PR tree-optimization/86114
63 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
65 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
67 2018-06-13 Richard Biener <rguenther@suse.de>
69 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
70 Properly set vector type of the intermediate stmt.
71 * tree-vect-stmts.c (vectorizable_operation): The destination
72 var always has vectype_out type.
74 2018-06-13 Jeff Law <law@redhat.com>
76 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
77 integer 0 for argument to print_rtl_with_bb.
78 (rl78_reorg): Likewise.
80 2018-06-13 David Malcolm <dmalcolm@redhat.com>
82 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
83 from rtx to rtx_insn *.
84 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
86 (add_sched_insns_for_speculation): Likewise for local "target",
87 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
88 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
89 from rtx_insn ** to rtx_code_label **.
90 (reorg_emit_nops): Likewise.
91 (c6x_reorg): Likewise for local "call_labels".
92 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
94 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
95 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
96 the loops over LABEL_REFS.
97 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
99 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
100 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
101 (split_branches): Strengthen local "olabel" from rtx to
102 rtx_insn *, adding a safe_as_a cast.
103 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
105 (add_insn_after): Likewise for first two params.
106 (add_insn_before): Likewise.
107 (remove_insn): Likewise for param.
108 (emit_pattern_before_noloc): Likewise for second and third params.
109 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
110 (emit_call_insn_before_noloc): Likewise.
111 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
113 (emit_barrier_before): Likewise.
114 (emit_label_before): Strengthen "label" param from "rtx" to
115 "rtx_code_label *". Strengthen "before" param from "rtx" to
117 (emit_insn_after_1): Strengthen "after" param from "rtx" to
119 (emit_pattern_after_noloc): Likewise.
120 (emit_insn_after_noloc): Likewise.
121 (emit_jump_insn_after_noloc): Likewise.
122 (emit_call_insn_after_noloc): Likewise.
123 (emit_debug_insn_after_noloc): Likewise.
124 (emit_barrier_after): Likewise.
125 (emit_label_after): Likewise for both params.
126 (emit_pattern_after_setloc): Likewise for "after" param. Convert
127 "loc" param from "int" to "location_t".
128 (emit_insn_after_setloc): Likewise.
129 (emit_jump_insn_after_setloc): Likewise.
130 (emit_call_insn_after_setloc): Likewise.
131 (emit_debug_insn_after_setloc): Likewise.
132 (emit_pattern_before_setloc): Likewise for "before" param. Convert
133 "loc" param from "int" to "location_t".
134 (emit_pattern_before): Convert NULL_RTX to NULL.
135 (emit_insn_before_setloc): Convert "loc" param from "int" to
137 (emit_jump_insn_before_setloc): Likewise.
138 (emit_call_insn_before_setloc): Likewise.
139 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
140 rtx_insn *. Convert "loc" param from "int" to "location_t".
141 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
142 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
143 Convert 3rd param from "int" to "location_t".
144 (emit_barrier_before, emit_barrier_after, next_real_insn):
145 Strengthen param from rtx to rtx_insn *.
146 (emit_label_before): Strengthen 1st param from "rtx" to
147 "rtx_code_label *". Strengthen 2nd param from "rtx" to
149 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
150 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
151 Strengthen 2nd param from "rtx" to "rtx_insn *".
152 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
153 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
154 Likewise. Convert 3rd param from "int" to "location_t".
155 (emit_label_after): Strengthen 1st param from "rtx" to
157 (next_real_insn, remove_insn): Strengthen param from "rtx" to
159 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
160 from "rtx" to "rtx_insn *".
162 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
164 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
165 bodies streamed in with -Q.
166 * dumpfile.c (dump_files): Add lto-stream-out dump file.
167 * dumpfile.h (tree_dump_index): Add lto_stream_out.
168 * gimple-streamer-out.c: Include gimple-pretty-print.h
169 (output_bb): Dump stmts streamed.
170 * lto-section-out.c: Include print-tree.h
171 (lto_begin_section): Dump sections created.
172 (lto_output_decl_index): Dump decl encoded.
173 * lto-streamer-out.c: Include print-tree.h
174 (create_output_block): Dump output block created.
175 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
176 (output_function): Dump function output.
177 (output_constructor): Dump constructor streamed.
178 (write_global_stream): Output indexes encoded.
179 (produce_asm_for_decls): Dump streams encoded.
180 * lto-streamer.c (streamer_dump_file): New global var.
181 * lto-streamer.h (streamer_dump_file): Declare.
182 * passes.c (ipa_write_summaries): Initialize streamer dump.
183 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
186 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
189 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
190 offsets for register save directives. Emit a second batch of save
191 directives, if need be, when the function accesses prior frames.
193 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
195 * config/arc/fpu.md (fmasf4): Force operand to register.
198 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
200 * config/arc/arc-protos.h (arc_pad_return): Remove.
201 * config/arc/arc.c (machine_function): Remove force_short_suffix
203 (arc_print_operand): Adjust printing of '&'.
204 (arc_verify_short): Remove conditional printing of short suffix.
205 (arc_final_prescan_insn): Remove reference to size_reason.
206 (pad_return): New function.
207 (arc_reorg): Call pad_return.
208 (arc_pad_return): Remove.
209 (arc_init_machine_status): Remove reference to force_short_suffix.
210 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
211 (attr length): When attribute iscompact is true force to 2
212 regardless; in the case of maybe check if we want to force the
213 instruction to have 4 bytes length.
214 (nopv): Change it to generate 4 byte long nop as well.
215 (blockage): New pattern.
216 (simple_return): Remove call to arc_pad_return.
217 (p_return_i): Likewise.
219 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
221 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
223 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
225 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
228 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
230 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
231 for ARC700 and ARCv2.
233 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
236 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
237 operands[2] instead of operands[1].
240 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
242 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
243 case, check whether the outer register overlaps an unallocatable
244 register, not just whether it fits the required class.
246 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
248 * poly-int.h (can_div_trunc_p): Add new overload in which all values
250 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
251 (memrefs_conflict_p): Likewise.
252 (init_alias_analysis): Likewise.
253 * cfgexpand.c (expand_debug_expr): Likewise.
254 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
255 * cse.c (fold_rtx): Likewise.
256 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
257 * expr.c (emit_block_move_hints): Likewise.
258 (clear_storage_hints, push_block, emit_push_insn): Likewise.
259 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
260 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
261 (emit_group_store): Likewise.
262 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
263 to read the PRE/POST_MODIFY increment.
264 * calls.c (store_one_arg): Use strip_offset.
265 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
267 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
269 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
270 (simplify_binary_operation_1): Extend CONST_INT handling to
272 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
273 than a HOST_WIDE_INT.
274 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
276 (adjust_mems, add_stores): Update accodingly.
277 (vt_canonicalize_addr): Track polynomial offsets.
278 (emit_note_insn_var_location): Likewise.
279 (vt_add_function_parameter): Likewise.
280 (vt_initialize): Likewise.
282 2018-06-12 Jeff Law <law@redhat.com>
284 * config.gcc (alpha*-*-freebsd*): Remove.
285 * config/alpha/freebsd.h: Remove.
287 2018-06-12 David Malcolm <dmalcolm@redhat.com>
290 * spellcheck-tree.c (levenshtein_distance): Rename to...
291 (get_edit_distance): ...this, and update for underlying renaming.
292 * spellcheck-tree.h (levenshtein_distance): Rename to...
293 (get_edit_distance): ...this.
294 * spellcheck.c (levenshtein_distance): Rename to...
295 (get_edit_distance): ...this. Convert from Levenshtein distance
296 to Damerau-Levenshtein distance by supporting transpositions of
297 adjacent characters. Rename "v1" to "v_next" and "v0" to
299 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
300 (selftest::test_edit_distance_unit_test_oneway): ...this, and
301 update for underlying renaming.
302 (selftest::levenshtein_distance_unit_test): Rename to...
303 (selftest::test_get_edit_distance_unit): ...this, and update for
305 (selftest::test_find_closest_string): Add example from PR 69968
306 where transposition helps
307 (selftest::test_metric_conditions): Update for renaming.
308 (selftest::test_metric_conditions): Likewise.
309 (selftest::spellcheck_c_tests): Likewise.
310 * spellcheck.h (levenshtein_distance): Rename both overloads to...
311 (get_edit_distance): ...this.
312 (best_match::consider): Update for renaming.
314 2018-06-12 Martin Sebor <msebor@redhat.com>
316 PR tree-optimization/85259
317 * builtins.c (compute_objsize): Handle constant offsets.
318 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
319 true iff a warning has been issued.
320 * gimple.h (gimple_nonartificial_location): New function.
321 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
322 gimple_nonartificial_location and handle -Wno-system-headers.
323 (handle_builtin_stxncpy): Same.
325 2018-06-12 Martin Sebor <msebor@redhat.com>
328 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
330 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
332 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
333 BUILTIN_VEC_XST entries for pointer to double and long long.
335 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
338 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
340 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
343 2018-06-12 Martin Liska <mliska@suse.cz>
345 * doc/options.texi: Document IntegerRange.
347 2018-06-12 Martin Liska <mliska@suse.cz>
349 * config/i386/i386.opt: Make MPX-related options as Deprecated.
350 * opt-functions.awk: Handle Deprecated flag.
351 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
353 (read_cmdline_option): Report warning for a deprecated option.
354 * opts.h (struct cl_option): Add new field cl_deprecated.
355 (CL_ERR_DEPRECATED): New.
357 2018-06-12 Martin Liska <mliska@suse.cz>
359 * doc/options.texi: Document Deprecated option flag.
361 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
363 * config/arc/arc-arch.h (arc_extras): New enum.
364 (arc_cpu_t):Add field extra.
365 (arc_cpu_types): Consider the extras.
366 * config/arc/arc-cpus.def: Add extras info.
367 * config/arc/arc-opts.h (processor_type): Consider extra field.
368 * config/arc/arc.c (arc_override_options): Handle extra field.
370 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
372 * config/arc/arc-arch.h: Update ARC_OPTX macro.
373 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
375 * config/arc/arc.c (arc_init): Update pic warning.
376 (irq_range): Update irq range parsing warnings.
377 (arc_override_options): Update various warning messages.
378 (arc_handle_aux_attribute): Likewise.
380 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
382 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
384 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
386 * doc/sourcebuild.texi: Document usage of line number 0 in verify
387 compiler messages directives.
389 2018-06-12 Matthew Fortune <mfortune@gmail.com>
391 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
392 * config/mips/mips-tables.opt: Regenerate.
393 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
395 * doc/invoke.texi: Document -march=i6500.
397 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
399 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
400 (i6400_gpmul): Add cpu_unit.
401 (i6400_gpdiv): Likewise.
402 (i6400_msa_add_d): Update reservations.
403 (i6400_msa_int_add) Likewise.
404 (i6400_msa_short_logic3) Likewise.
405 (i6400_msa_short_logic2) Likewise.
406 (i6400_msa_short_logic) Likewise.
407 (i6400_msa_move) Likewise.
408 (i6400_msa_cmp) Likewise.
409 (i6400_msa_short_float2) Likewise.
410 (i6400_msa_div_d) Likewise.
411 (i6400_msa_long_logic1) Likewise.
412 (i6400_msa_long_logic2) Likewise.
413 (i6400_msa_mult) Likewise.
414 (i6400_msa_long_float2) Likewise.
415 (i6400_msa_long_float4) Likewise.
416 (i6400_msa_long_float5) Likewise.
417 (i6400_msa_long_float8) Likewise.
418 (i6400_fpu_fadd): Include frint type.
419 (i6400_fpu_store): New define_insn_reservation.
420 (i6400_fpu_load): Likewise.
421 (i6400_fpu_move): Likewise.
422 (i6400_fpu_fcmp): Likewise.
423 (i6400_fpu_fmadd): Likewise.
424 (i6400_int_mult): Include imul3nc type and update reservation.
425 (i6400_int_div): Include idiv3 type and update reservation.
426 (i6400_int_load): Update to check type not move_type.
427 (i6400_int_store): Likewise.
428 (i6400_int_prefetch): Set zero latency.
430 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
432 * gcc.c: Document new %@{...} sequence.
433 (LINK_COMMAND_SPEC): Use it for the -L switches.
434 (cpp_unique_options): Use it for the -I switches.
435 (at_file_argbuf): New global variable.
436 (in_at_file): Likewise.
437 (alloc_args): Create at_file_argbuf.
438 (clear_args): Truncate at_file_argbuf.
439 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
440 (open_at_file): New function.
441 (close_at_file): Likewise.
442 (create_at_file): Delete.
443 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
446 (validate_switches_from_spec): Deal with %@{...} sequence.
447 (validate_switches): Likewise.
448 (driver::finalize): Call clear_args.
450 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
452 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
454 2018-06-11 Martin Sebor <msebor@redhat.com>
456 * doc/invoke.texi (-Wall): List -Wc++17-compat.
457 (Wno-class-memaccess): Add @opindex.
458 (Wno-templates, Wno-multiple-inheritance): Same.
459 (Wno-virtual-inheritance, Wno-namespaces): Same.
460 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
461 (Wno-format-overflow, Wno-format-truncation): Same.
462 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
463 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
464 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
465 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
466 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
467 (Wno-misspelled-isr): Same.
469 2018-06-11 Martin Sebor <msebor@redhat.com>
471 * PR tree-optimization/86083
472 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
474 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
476 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
478 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
481 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
482 on the correct operand.
483 (*movdi_internal64): Ditto.
485 2018-06-11 Martin Liska <mliska@suse.cz>
487 PR tree-optimization/86089
488 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
490 2018-06-11 Julia Koval <julia.koval@intel.com>
492 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
493 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
494 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
496 2018-06-11 Olivier Hainque <hainque@adacore.com>
498 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
499 for Ada with strict dwarf2.
501 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
504 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
507 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
509 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
511 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
513 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
516 2018-06-08 Martin Liska <mliska@suse.cz>
518 * tree-cfg.h (debug_function): Fix argument type to match
521 2018-06-08 Martin Liska <mliska@suse.cz>
523 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
524 Remove usage of MPX-related (and removed) fields.
525 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
527 2018-06-08 David Malcolm <dmalcolm@redhat.com>
529 * cfg.c (debug): Use TDF_NONE rather than 0.
530 * cfghooks.c (debug): Likewise.
531 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
532 (struct dump_option_value_info): Convert to...
533 (struct kv_pair): ...this template type.
534 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
536 (optinfo_verbosity_options): Likewise.
537 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
539 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
540 than int for "optgroup_flags" param.
541 (dump_generic_expr_loc): Use dump_flags_t rather than int for
543 (dump_dec): Likewise.
544 (dump_finish): Use TDF_NONE rather than 0.
545 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
546 rather than int for "optgroup_flags" param. Use TDF_NONE rather
547 than 0. Update for change to option_ptr.
548 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
549 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
550 0. Update for changes to optinfo_verbosity_options and
552 (opt_info_switch_p): Convert optgroup_flags from int to
554 (dump_basic_block): Use dump_flags_t rather than int
555 for "dump_kind" param.
556 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
557 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
558 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
559 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
560 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
561 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
562 TDF_NONE): Convert from macros to...
563 (enum dump_flag): ...this new enum.
564 (dump_flags_t): Update to use enum.
565 (operator|, operator&, operator~, operator|=, operator&=):
566 Implement for dump_flags_t.
567 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
568 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
569 Convert from macros to...
570 (enum optgroup_flag): ...this new enum.
571 (optgroup_flags_t): New typedef.
572 (operator|, operator|=): Implement for optgroup_flags_t.
573 (struct dump_file_info): Convert field "alt_flags" to
574 dump_flags_t. Convert field "optgroup_flags" to
576 (dump_basic_block): Use dump_flags_t rather than int for param.
577 (dump_generic_expr_loc): Likewise.
578 (dump_dec): Likewise.
579 (dump_register): Convert param "optgroup_flags" to
581 (opt_info_enable_passes): Likewise.
582 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
584 * gimple-pretty-print.c (debug): Likewise.
585 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
586 (merged_store_group::apply_stores): Likewise.
587 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
588 * gimple.c (verify_gimple_pp): Likewise.
589 * graphite-poly.c (print_pbb_body): Likewise.
590 * passes.c (pass_manager::register_one_dump_file): Convert
591 local "optgroup_flags" to optgroup_flags_t.
592 * print-tree.c (print_node): Use TDF_NONE rather than 0.
594 (debug_body): Likewise.
595 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
597 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
599 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
600 (convert_mult_to_fma): Likewise.
601 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
602 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
603 * tree-vect-data-refs.c (dump_lower_bound): Convert param
604 "dump_kind" to dump_flags_t.
606 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
608 * config/rs6000/rs6000.c (min, max): Delete.
610 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
612 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
613 -mabi=spe and -mabi=no-spe.
615 2018-06-08 Martin Liska <mliska@suse.cz>
617 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
618 where we expect an existing summary.
620 2018-06-08 Martin Liska <mliska@suse.cz>
622 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
623 * ipa-inline.h (estimate_edge_growth): Likewise.
625 2018-06-08 Martin Liska <mliska@suse.cz>
627 * cgraph.c (function_version_hasher::hash): Use
628 cgraph_node::get_uid ().
629 (function_version_hasher::equal):
630 * cgraph.h (cgraph_node::get_uid): New method.
631 * ipa-inline.c (update_caller_keys): Use
632 cgraph_node::get_uid ().
633 (update_callee_keys): Likewise.
634 * ipa-utils.c (searchc): Likewise.
635 (ipa_reduced_postorder): Likewise.
636 * lto-cgraph.c (input_node): Likewise.
637 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
638 * symbol-summary.h (symtab_insertion): Likewise.
639 (symtab_removal): Likewise.
640 (symtab_duplication): Likewise.
641 * tree-pretty-print.c (dump_function_header): Likewise.
642 * tree-sra.c (convert_callers_for_node): Likewise.
644 2018-06-08 Martin Liska <mliska@suse.cz>
646 * cgraph.c (symbol_table::create_edge): Always assign a new
648 (symbol_table::free_edge): Do not recycle numbers.
649 * cgraph.h (cgraph_edge::get): New method.
650 * symbol-summary.h (symtab_removal): Use it.
651 (symtab_duplication): Likewise.
652 (call_summary::hashable_uid): Remove.
654 2018-06-08 Martin Liska <mliska@suse.cz>
656 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
657 (initialize_growth_caches): Remove.
658 (free_growth_caches): Likewise.
659 (do_estimate_edge_time): Use edge_growth_cache.
660 (do_estimate_edge_size): Likewise.
661 (do_estimate_edge_hints): Likewise.
662 * ipa-inline.c (reset_edge_caches): Likewise.
663 (recursive_inlining): Likewise.
664 (inline_small_functions): Likewise.
665 * ipa-inline.h (initialize_growth_caches): Remove.
666 (estimate_edge_size): Likewise.
667 (estimate_edge_time): Likewise.
668 (estimate_edge_hints): Likewise.
669 (reset_edge_growth_cache): Likewise.
670 * symbol-summary.h (call_summary::remove): New method.
672 2018-06-08 Martin Liska <mliska@suse.cz>
674 * ipa-cp.c (class edge_clone_summary): New summary.
675 (grow_edge_clone_vectors): Remove.
676 (ipcp_edge_duplication_hook): Remove.
677 (class edge_clone_summary_t): New call_summary class.
678 (ipcp_edge_removal_hook): Remove.
679 (edge_clone_summary_t::duplicate): New function.
680 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
681 (create_specialized_node): Likewise.
682 (ipcp_driver): Initialize edge_clone_summaries and do not
685 2018-06-08 Martin Liska <mliska@suse.cz>
687 * symbol-summary.h (get): New function.
688 (call_summary::m_initialize_when_cloning): New class member.
690 2018-06-08 Martin Liska <mliska@suse.cz>
692 * cgraph.c (cgraph_node::remove): Do not recycle uid.
693 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
694 (symbol_table::allocate_cgraph_symbol): Do not set uid.
695 * passes.c (uid_hash_t): Record removed_nodes by their uids.
696 (remove_cgraph_node_from_order): Use the removed_nodes set.
697 (do_per_function_toporder): Likwise.
698 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
699 instead of summary_uid.
700 (symtab_removal): Likewise.
701 (symtab_duplication): Likewise.
703 2018-06-08 Martin Liska <mliska@suse.cz>
705 * ipa-cp.c (ipcp_store_bits_results): Use
706 ipcp_transformation_sum.
707 (ipcp_store_vr_results): Likewise.
708 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
710 (ipcp_transformation_initialize): ... this.
711 (ipa_set_node_agg_value_chain):
712 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
713 (write_ipcp_transformation_info): Likewise.
714 (read_ipcp_transformation_info): Likewise.
715 (ipcp_update_bits): Likewise.
716 (ipcp_update_vr): Likewise.
717 (ipcp_transform_function): Likewise.
718 * ipa-prop.h: Rename ipcp_transformation_summary to
720 (class ipcp_transformation_t): New function summary.
721 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
722 (ipa_get_agg_replacements_for_node): Likewise.
724 2018-06-08 Martin Liska <mliska@suse.cz>
726 * ipa-pure-const.c (struct funct_state_d): Do it class instead
728 (class funct_state_summary_t): New function_summary class.
729 (has_function_state): Remove.
730 (get_function_state): Likewise.
731 (set_function_state): Likewise.
732 (add_new_function): Likewise.
733 (funct_state_summary_t::insert): New function.
734 (duplicate_node_data): Remove.
735 (remove_node_data): Remove.
736 (funct_state_summary_t::duplicate): New function.
737 (register_hooks): Create new funct_state_summaries.
738 (pure_const_generate_summary): Use it.
739 (pure_const_write_summary): Likewise.
740 (pure_const_read_summary): Likewise.
741 (propagate_pure_const): Likewise.
742 (propagate_nothrow): Likewise.
743 (dump_malloc_lattice): Likewise.
744 (propagate_malloc): Likewise.
745 (execute): Do not register hooks, just remove summary
747 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
750 2018-06-08 Martin Liska <mliska@suse.cz>
752 * ipa-reference.c (remove_node_data): Remove.
753 (duplicate_node_data): Likewise.
754 (class ipa_ref_var_info_summary_t): New class.
755 (class ipa_ref_opt_summary_t): Likewise.
756 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
757 (get_reference_optimization_summary): Use
758 ipa_ref_opt_sum_summaries.
759 (set_reference_vars_info): Remove.
760 (set_reference_optimization_summary): Likewise.
761 (ipa_init): Create summaries.
762 (init_function_info): Use function summary.
763 (ipa_ref_opt_summary_t::duplicate): New function.
764 (ipa_ref_opt_summary_t::remove): New function.
765 (get_read_write_all_from_node): Fix GNU coding style.
766 (propagate): Use function summary.
767 (write_node_summary_p): Fix GNU coding style.
768 (stream_out_bitmap): Likewise.
769 (ipa_reference_read_optimization_summary): Use function summary.
770 (ipa_reference_c_finalize): Do not release hooks.
772 2018-06-08 Martin Liska <mliska@suse.cz>
774 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
775 (analyze_function_body): Extract multiple calls of get_create.
776 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
777 * ipa-inline.c (recursive_inlining): Use ::get method.
778 * ipa-inline.h (estimate_edge_growth): Likewise.
780 2018-06-08 Martin Liska <mliska@suse.cz>
782 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
784 (hsa_function_summary::hsa_function_summary): Use the new enum
786 (hsa_gpu_implementation_p): Use hsa_summaries::get.
787 * hsa-gen.c (hsa_get_host_function): Likewise.
788 (get_brig_function_name): Likewise.
789 * ipa-hsa.c (process_hsa_functions): Likewise.
790 (ipa_hsa_write_summary): Likewise.
791 * symbol-summary.h (symtab_duplication): Use ::get function/
794 2018-06-08 Martin Liska <mliska@suse.cz>
796 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
798 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
799 (hsa_register_kernel): Likewise.
800 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
801 * hsa-gen.c (hsa_get_host_function): Likewise.
802 (get_brig_function_name): Likewise.
803 (generate_hsa): Likewise.
804 (pass_gen_hsail::execute): Likewise.
805 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
806 (devirtualization_time_bonus): Likewise.
807 (ipcp_propagate_stage): Likewise.
808 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
809 (edge_set_predicate): Likewise.
810 (evaluate_conditions_for_known_args): Likewise.
811 (evaluate_properties_for_edge): Likewise.
812 (ipa_fn_summary::reset): Likewise.
813 (ipa_fn_summary_t::duplicate): Likewise.
814 (dump_ipa_call_summary): Likewise.
815 (ipa_dump_fn_summary): Likewise.
816 (analyze_function_body): Likewise.
817 (compute_fn_summary): Likewise.
818 (estimate_edge_devirt_benefit): Likewise.
819 (estimate_edge_size_and_time): Likewise.
820 (estimate_calls_size_and_time): Likewise.
821 (estimate_node_size_and_time): Likewise.
822 (inline_update_callee_summaries): Likewise.
823 (remap_edge_change_prob): Likewise.
824 (remap_edge_summaries): Likewise.
825 (ipa_merge_fn_summary_after_inlining): Likewise.
826 (ipa_update_overall_fn_summary): Likewise.
827 (read_ipa_call_summary): Likewise.
828 (inline_read_section): Likewise.
829 (write_ipa_call_summary): Likewise.
830 (ipa_fn_summary_write): Likewise.
831 (ipa_free_fn_summary): Likewise.
832 * ipa-hsa.c (process_hsa_functions): Likewise.
833 (ipa_hsa_write_summary): Likewise.
834 (ipa_hsa_read_section): Likewise.
835 * ipa-icf.c (sem_function::merge): Likewise.
836 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
837 (do_estimate_edge_time): Likewise.
838 (estimate_size_after_inlining): Likewise.
839 (estimate_growth): Likewise.
840 (growth_likely_positive): Likewise.
841 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
842 (inline_call): Likewise.
843 * ipa-inline.c (caller_growth_limits): Likewise.
844 (can_inline_edge_p): Likewise.
845 (can_inline_edge_by_limits_p): Likewise.
846 (compute_uninlined_call_time): Likewise.
847 (compute_inlined_call_time): Likewise.
848 (want_inline_small_function_p): Likewise.
849 (edge_badness): Likewise.
850 (update_caller_keys): Likewise.
851 (update_callee_keys): Likewise.
852 (recursive_inlining): Likewise.
853 (inline_small_functions): Likewise.
854 (inline_to_all_callers_1): Likewise.
855 (dump_overall_stats): Likewise.
856 (early_inline_small_functions): Likewise.
857 (early_inliner): Likewise.
858 * ipa-inline.h (estimate_edge_growth): Likewise.
859 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
860 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
861 * ipa-prop.h (IPA_NODE_REF): Likewise.
862 (IPA_EDGE_REF): Likewise.
863 * ipa-pure-const.c (malloc_candidate_p): Likewise.
864 (propagate_malloc): Likewise.
865 * ipa-split.c (execute_split_functions): Likewise.
866 * symbol-summary.h: Rename get to get_create.
868 (get_create): Likewise.
869 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
871 2018-06-08 Martin Liska <mliska@suse.cz>
873 * symbol-summary.h (release): Move definition out of class
875 (symtab_removal): Likewise.
876 (symtab_duplication): Likewise.
878 2018-06-08 Martin Liska <mliska@suse.cz>
880 * symbol-summary.h (function_summary): Move constructor
881 implementation out of class declaration.
883 (symtab_insertion): Likewise.
884 (symtab_removal): Likewise.
885 (symtab_duplication): Likewise.
888 2018-06-08 Martin Liska <mliska@suse.cz>
890 * Makefile.in: Remove support for MPX (macros, related functions,
891 fields in cgraph_node, ...).
892 * builtin-types.def (BT_BND): Likewise.
893 (BT_FN_BND_CONST_PTR): Likewise.
894 (BT_FN_CONST_PTR_BND): Likewise.
895 (BT_FN_VOID_PTR_BND): Likewise.
896 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
897 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
898 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
899 (expand_builtin_mempcpy_with_bounds): Likewise.
900 (expand_builtin_memset_with_bounds): Likewise.
901 (expand_builtin_memset_args): Likewise.
902 (std_expand_builtin_va_start): Likewise.
903 (expand_builtin): Likewise.
904 (expand_builtin_with_bounds): Likewise.
905 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
906 (DEF_LIB_BUILTIN_CHKP): Likewise.
907 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
908 (DEF_CHKP_BUILTIN): Likewise.
909 (BUILT_IN_MEMCPY): Likewise.
910 (BUILT_IN_MEMMOVE): Likewise.
911 (BUILT_IN_MEMPCPY): Likewise.
912 (BUILT_IN_MEMSET): Likewise.
913 (BUILT_IN_STPCPY): Likewise.
914 (BUILT_IN_STRCAT): Likewise.
915 (BUILT_IN_STRCHR): Likewise.
916 (BUILT_IN_STRCPY): Likewise.
917 (BUILT_IN_STRLEN): Likewise.
918 (BUILT_IN_MEMCPY_CHK): Likewise.
919 (BUILT_IN_MEMMOVE_CHK): Likewise.
920 (BUILT_IN_MEMPCPY_CHK): Likewise.
921 (BUILT_IN_MEMSET_CHK): Likewise.
922 (BUILT_IN_STPCPY_CHK): Likewise.
923 (BUILT_IN_STRCAT_CHK): Likewise.
924 (BUILT_IN_STRCPY_CHK): Likewise.
925 * calls.c (store_bounds): Likewise.
926 (emit_call_1): Likewise.
927 (special_function_p): Likewise.
928 (maybe_warn_nonstring_arg): Likewise.
929 (initialize_argument_information): Likewise.
930 (finalize_must_preallocate): Likewise.
931 (compute_argument_addresses): Likewise.
932 (expand_call): Likewise.
933 * cfgexpand.c (expand_call_stmt): Likewise.
934 (expand_return): Likewise.
935 (expand_gimple_stmt_1): Likewise.
936 (pass_expand::execute): Likewise.
937 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
938 (cgraph_node::remove): Likewise.
939 (cgraph_node::dump): Likewise.
940 (cgraph_node::verify_node): Likewise.
941 * cgraph.h (chkp_function_instrumented_p): Likewise.
942 (symtab_node::get_alias_target): Likewise.
943 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
944 (cgraph_local_p): Likewise.
945 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
946 (cgraph_edge::rebuild_references): Likewise.
947 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
948 (walk_polymorphic_call_targets): Likewise.
949 (cgraph_node::expand_thunk): Likewise.
950 (symbol_table::output_weakrefs): Likewise.
951 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
952 (ix86_handle_option): Likewise.
953 * config/i386/constraints.md: Likewise.
954 * config/i386/i386-builtin-types.def (BND): Likewise.
958 * config/i386/i386-builtin.def (BDESC_END): Likewise.
959 (BDESC_FIRST): Likewise.
961 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
962 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
963 * config/i386/i386.c (enum reg_class): Likewise.
964 (ix86_target_string): Likewise.
965 (ix86_option_override_internal): Likewise.
966 (ix86_conditional_register_usage): Likewise.
967 (ix86_valid_target_attribute_inner_p): Likewise.
968 (ix86_set_indirect_branch_type): Likewise.
969 (ix86_set_current_function): Likewise.
970 (ix86_function_arg_regno_p): Likewise.
971 (init_cumulative_args): Likewise.
972 (ix86_function_arg_advance): Likewise.
973 (ix86_function_arg): Likewise.
974 (ix86_pass_by_reference): Likewise.
975 (ix86_function_value_regno_p): Likewise.
976 (ix86_function_value_1): Likewise.
977 (ix86_function_value_bounds): Likewise.
978 (ix86_return_in_memory): Likewise.
979 (ix86_setup_incoming_vararg_bounds): Likewise.
980 (ix86_va_start): Likewise.
981 (indirect_thunk_need_prefix): Likewise.
982 (print_reg): Likewise.
983 (ix86_print_operand): Likewise.
984 (ix86_expand_call): Likewise.
985 (ix86_output_function_return): Likewise.
986 (reg_encoded_number): Likewise.
987 (BDESC_VERIFYS): Likewise.
988 (ix86_init_mpx_builtins): Likewise.
989 (ix86_init_builtins): Likewise.
990 (ix86_emit_cmove): Likewise.
991 (ix86_emit_move_max): Likewise.
992 (ix86_expand_builtin): Likewise.
993 (ix86_builtin_mpx_function): Likewise.
994 (ix86_get_arg_address_for_bt): Likewise.
995 (ix86_load_bounds): Likewise.
996 (ix86_store_bounds): Likewise.
997 (ix86_load_returned_bounds): Likewise.
998 (ix86_store_returned_bounds): Likewise.
999 (ix86_class_likely_spilled_p): Likewise.
1000 (ix86_hard_regno_mode_ok): Likewise.
1001 (x86_order_regs_for_local_alloc): Likewise.
1002 (ix86_mitigate_rop): Likewise.
1003 (ix86_bnd_prefixed_insn_p): Likewise.
1004 (ix86_mpx_bound_mode): Likewise.
1005 (ix86_make_bounds_constant): Likewise.
1006 (ix86_initialize_bounds): Likewise.
1007 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1008 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1009 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1010 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1011 (TARGET_CHKP_BOUND_MODE): Likewise.
1012 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1013 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1014 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1015 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1016 * config/i386/i386.h (TARGET_MPX): Likewise.
1017 (TARGET_MPX_P): Likewise.
1018 (VALID_BND_REG_MODE): Likewise.
1019 (FIRST_BND_REG): Likewise.
1020 (LAST_BND_REG): Likewise.
1021 (enum reg_class): Likewise.
1022 (BND_REG_P): Likewise.
1023 (BND_REGNO_P): Likewise.
1024 (BNDmode): Likewise.
1025 (ADJUST_INSN_LENGTH): Likewise.
1026 * config/i386/i386.md: Likewise.
1027 * config/i386/i386.opt: Likewise.
1028 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1029 (defined): Likewise.
1030 (LINK_MPX): Likewise.
1031 (MPX_SPEC): Likewise.
1032 (LIBMPX_SPEC): Likewise.
1033 (LIBMPXWRAPPERS_SPEC): Likewise.
1034 (CHKP_SPEC): Likewise.
1035 * config/i386/predicates.md: Likewise.
1036 * dbxout.c (dbxout_type): Likewise.
1037 * doc/extend.texi: Likewise.
1038 * doc/invoke.texi: Likewise.
1039 * doc/md.texi: Likewise.
1040 * doc/tm.texi: Likewise.
1041 * doc/tm.texi.in: Likewise.
1042 * dwarf2out.c (is_base_type): Likewise.
1043 (gen_formal_types_die): Likewise.
1044 (gen_subprogram_die): Likewise.
1045 (gen_type_die_with_usage): Likewise.
1046 (gen_decl_die): Likewise.
1047 (dwarf2out_late_global_decl): Likewise.
1048 * expr.c (expand_assignment): Likewise.
1049 (emit_storent_insn): Likewise.
1050 (store_expr_with_bounds): Likewise.
1051 (store_expr): Likewise.
1052 (expand_expr_real_1): Likewise.
1053 * expr.h (store_expr_with_bounds): Likewise.
1054 * function.c (use_register_for_decl): Likewise.
1055 (struct bounds_parm_data): Likewise.
1056 (assign_parms_augmented_arg_list): Likewise.
1057 (assign_parm_find_entry_rtl): Likewise.
1058 (assign_parm_is_stack_parm): Likewise.
1059 (assign_parm_load_bounds): Likewise.
1060 (assign_bounds): Likewise.
1061 (assign_parms): Likewise.
1062 (expand_function_start): Likewise.
1063 * gcc.c (CHKP_SPEC): Likewise.
1064 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1065 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1066 (wrestrict_dom_walker::check_call): Likewise.
1067 * gimple.c (gimple_build_call_from_tree): Likewise.
1068 * gimple.h (enum gf_mask): Likewise.
1069 (gimple_call_with_bounds_p): Likewise.
1070 (gimple_call_set_with_bounds): Likewise.
1071 * gimplify.c (gimplify_init_constructor): Likewise.
1072 * ipa-cp.c (initialize_node_lattices): Likewise.
1073 (propagate_constants_across_call): Likewise.
1074 (find_more_scalar_values_for_callers_subset): Likewise.
1075 * ipa-hsa.c (process_hsa_functions): Likewise.
1076 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1077 * ipa-icf.c (sem_function::merge): Likewise.
1078 * ipa-inline.c (early_inliner): Likewise.
1079 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1080 (warn_function_cold): Likewise.
1081 (propagate_pure_const): Likewise.
1082 * ipa-ref.h (enum GTY): Likewise.
1083 * ipa-split.c (find_retbnd): Likewise.
1084 (consider_split): Likewise.
1085 (split_function): Likewise.
1086 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1087 * ipa.c (walk_polymorphic_call_targets): Likewise.
1088 (symbol_table::remove_unreachable_nodes): Likewise.
1089 (process_references): Likewise.
1090 (cgraph_build_static_cdtor_1): Likewise.
1091 * lto-cgraph.c (lto_output_node): Likewise.
1092 (output_refs): Likewise.
1093 (compute_ltrans_boundary): Likewise.
1094 (input_overwrite_node): Likewise.
1095 (input_node): Likewise.
1096 (input_cgraph_1): Likewise.
1097 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1098 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1099 (class pass_chkp_instrumentation_passes): Likewise.
1100 (make_pass_chkp_instrumentation_passes): Likewise.
1101 * passes.def: Likewise.
1102 * rtl.h (struct GTY): Likewise.
1103 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1104 * stor-layout.c (layout_type): Likewise.
1105 * symtab.c: Likewise.
1106 * target.def: Likewise.
1107 * targhooks.c (default_chkp_bound_type): Likewise.
1108 (default_chkp_bound_mode): Likewise.
1109 (default_builtin_chkp_function): Likewise.
1110 (default_chkp_function_value_bounds): Likewise.
1111 (default_chkp_make_bounds_constant): Likewise.
1112 (default_chkp_initialize_bounds): Likewise.
1113 * targhooks.h (default_chkp_bound_type): Likewise.
1114 (default_chkp_bound_mode): Likewise.
1115 (default_builtin_chkp_function): Likewise.
1116 (default_chkp_function_value_bounds): Likewise.
1117 (default_chkp_make_bounds_constant): Likewise.
1118 (default_chkp_initialize_bounds): Likewise.
1119 * toplev.c (compile_file): Likewise.
1120 (process_options): Likewise.
1121 * tree-core.h (DEF_BUILTIN): Likewise.
1122 (DEF_BUILTIN_CHKP): Likewise.
1123 * tree-inline.c (declare_return_variable): Likewise.
1124 (remap_gimple_stmt): Likewise.
1125 (copy_bb): Likewise.
1126 (initialize_inlined_parameters): Likewise.
1127 (expand_call_inline): Likewise.
1128 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1129 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1130 (make_pass_ipa_chkp_produce_thunks): Likewise.
1131 (make_pass_chkp): Likewise.
1132 (make_pass_chkp_opt): Likewise.
1133 (make_pass_chkp_instrumentation_passes): Likewise.
1134 * tree-pretty-print.c (dump_generic_node): Likewise.
1135 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1136 * tree-ssa-dce.c (propagate_necessity): Likewise.
1137 (eliminate_unnecessary_stmts): Likewise.
1138 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1139 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1140 * tree-ssa-sccvn.h: Likewise.
1141 * tree-ssa-strlen.c (get_string_length): Likewise.
1142 (valid_builtin_call): Likewise.
1143 (adjust_last_stmt): Likewise.
1144 (handle_builtin_strchr): Likewise.
1145 (handle_builtin_strcpy): Likewise.
1146 (handle_builtin_stxncpy): Likewise.
1147 (handle_builtin_memcpy): Likewise.
1148 (handle_builtin_strcat): Likewise.
1149 (strlen_check_and_optimize_stmt): Likewise.
1150 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1151 * tree-streamer-in.c: Likewise.
1152 * tree-streamer.c (record_common_node): Likewise.
1153 * tree.c (tree_code_size): Likewise.
1154 (wide_int_to_tree_1): Likewise.
1155 (type_contains_placeholder_1): Likewise.
1156 (build_common_tree_nodes): Likewise.
1157 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1158 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1159 (POINTER_BOUNDS_P): Likewise.
1160 (BOUNDED_TYPE_P): Likewise.
1161 (BOUNDED_P): Likewise.
1162 (CALL_WITH_BOUNDS_P): Likewise.
1163 (pointer_bounds_type_node): Likewise.
1164 * value-prof.c (gimple_ic): Likewise.
1165 * var-tracking.c (vt_add_function_parameters): Likewise.
1166 * varasm.c (make_decl_rtl): Likewise.
1167 (assemble_start_function): Likewise.
1168 (output_constant): Likewise.
1169 (maybe_assemble_visibility): Likewise.
1170 * varpool.c (ctor_for_folding): Likewise.
1171 * chkp-builtins.def: Remove.
1172 * ipa-chkp.c: Remove.
1173 * ipa-chkp.h: Remove.
1174 * rtl-chkp.c: Remove.
1175 * rtl-chkp.h: Remove.
1176 * tree-chkp-opt.c: Remove.
1177 * tree-chkp.c: Remove.
1178 * tree-chkp.h: Remove.
1180 2018-06-07 Carl Love <cel@us.ibm.com>
1182 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1183 vextract_fp_from_shortl): Add BE support.
1185 2018-06-07 Paul Koning <ni1d@arrl.net>
1187 * compare-elim.c (try_merge_compare): Don't merge compare if
1188 address contains a side effect.
1189 (try_eliminate_compare): Likewise.
1191 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1193 * config.gcc: Support "tremont".
1194 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1195 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1197 * config/i386/i386.c (m_TREMONT): Define.
1198 (processor_target_table): Add "tremont".
1199 (PTA_TREMONT): Define.
1200 (ix86_lea_outperforms): Add TARGET_TREMONT.
1201 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1202 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1203 and M_INTEL_GOLDMONT_PLUS.
1204 (fold_builtin_cpu): Add "tremont".
1205 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1206 (ix86_option_override_internal): Add "tremont".
1207 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1208 (processor_type): Add PROCESSOR_TREMONT.
1209 * config/i386/x86-tune.def: Add m_TREMONT.
1210 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1212 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1214 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1215 symbol defined for msp430i* devices to be lower case.
1217 gcc/testsuite/gcc.target/msp430/
1218 * msp430i-device-symbol.c: New test.
1219 * msp430f-device-symbol.c: New test.
1220 * msp430.h: New test header file.
1222 2018-06-07 Richard Biener <rguenther@suse.de>
1224 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1225 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1226 Properly wrap signed arithmetic if overflow wraps.
1228 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1230 PR tree-optimization/69615
1231 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1232 of a cast from a same precision integral SSA_NAME in a bb dominated
1233 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1234 cast to utype if rhs2 has already a compatible type.
1236 2018-06-07 Richard Biener <rguenther@suse.de>
1238 PR tree-optimization/85935
1239 * graphite-scop-detection.c (find_params_in_bb): Analyze
1240 condition operands with respect to the correct loop. Assert
1241 the analysis doesn't fail.
1243 2018-06-04 Carl Love <cel@us.ibm.com>
1245 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1246 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1247 as it is slightly cheaper.
1248 (first_match_or_eos_index_<mode>):
1249 Calculate index using natural element order.
1250 (first_match_index_<mode>):
1251 Calculate index using natural element order.
1252 (first_match_or_eos_index_<mode>):
1253 Calculate index using natural order.
1254 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1255 for BE and LE modes.
1256 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1257 P9V_BUILTIN_VCLZLSBB_V16QI.
1258 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1261 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1263 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1264 indentation and line wrap for many prototypes. Add missing
1265 @smallexample directives around block of prototypes for vec_xl and
1268 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1270 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1271 track if we pass or return IEEE 128-bit floating point.
1272 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1273 C++ mangling that is compatible with GCC 8.1.
1274 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1275 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1276 floating point types.
1277 (rs6000_function_arg_advance_1): Likewise.
1278 (rs6000_mangle_type): Optionally generate mangled names that match
1279 what GCC 8.1 generated for IEEE 128-bit floating point types.
1280 (rs6000_globalize_decl_name): If we have an external function that
1281 passes or returns IEEE 128-bit types, generate a weak reference
1282 from the mangled name used in GCC 8.1 to the current mangled
1284 (rs6000_init_builtins): Make __ibm128 use the long double type if
1285 long double is IBM extended double. Make __float128 use the long
1286 double type if long double is IEEE 128-bit.
1289 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1290 macro for __ibm128 built-in functions.
1291 (PACK_IF): Add __ibm128 pack/unpack functions.
1292 (UNPACK_IF): Likewise.
1293 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1294 enable long double built-in functions if long double is IEEE
1295 128-bit floating point.
1296 (rs6000_invalid_builtin): Update long double built-in function
1298 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1299 functions, adjust the built-in function to use the long double
1300 built-in function if __ibm128 and long double are the same type.
1301 * doc/extend.texi (PowerPC builtins): Update documention for
1302 __builtin_{,un}pack_longdouble. Add documentation for
1303 __builtin_{,un}pack_ibm128.
1305 2018-06-06 Jim Wilson <jimw@sifive.com>
1307 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1308 (struct machine_function): New field interrupt_mode.
1309 (riscv_handle_type_attribute): New function. Add forward declaration.
1310 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1311 (riscv_expand_epilogue): Check interrupt_mode field.
1312 (riscv_set_current_function): Check interrupt attribute args and
1313 set interrupt_mode field.
1314 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1315 (riscv_sret, riscv_uret): New.
1316 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1317 new arguments to interrupt attribute.
1319 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1322 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1323 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1325 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1327 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1328 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1329 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1330 several redundant entries.
1332 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1334 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1335 type from "rtx" to "rtx_insn *".
1336 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1337 for local "call_insn", removing cast.
1338 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1340 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1342 PR tree-optimization/86066
1343 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1344 for BIT_INSERT_EXPR stores.
1346 2018-06-06 Richard Biener <rguenther@suse.de>
1348 PR tree-optimization/86062
1349 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1350 component refs ontop
1351 of to be offsetted base.
1353 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1355 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1356 to be static and remove check on interrupt attribute name.
1358 gcc/testsuite/gcc.target/msp430/
1359 * function-attributes-4.c: New test.
1360 * static-interrupts.c: New test.
1362 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1364 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1365 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1367 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1370 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1372 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1373 Remove second argument, change how error is called.
1374 (aarch64_layout_arg): Remove second argument from
1375 aarch64_err_no_fpadvsimd call.
1376 (aarch64_init_cumulative_args): Ditto.
1377 (aarch64_gimplify_va_arg_expr): Ditto.
1378 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1380 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1382 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1383 (*simple_return_indirect_internal<mode>): Rename from
1384 simple_return_indirect_internal. Use W mode iterator.
1385 (rstorssp): New expander.
1386 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1387 (clrssbsy): New expander.
1388 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1390 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1392 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1394 (cmse_check_pointed_object): Likewise.
1396 2018-06-05 Martin Liska <mliska@suse.cz>
1398 PR gcov-profile/47618
1399 * doc/invoke.texi: Document how -fprofile-dir format
1402 2018-06-05 Richard Biener <rguenther@suse.de>
1404 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1405 removal pretend DOM info isn't available so we do not update
1406 it and only remove edges, not dominated blocks. Actually free
1407 DOM info in case we removed something. Remove unreachable blocks.
1408 (mfb_keep_latches): Work with either DOM info or marked backedges.
1409 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1410 first. Mark backedges if DOM info isn't available.
1411 (Re-)compute DOM info after cleanup_control_flow_pre.
1413 2018-06-05 Richard Biener <rguenther@suse.de>
1415 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1416 (locus_discrim_hasher::hash): Adjust.
1417 (locus_discrim_hasher::equal): Likewise.
1418 (next_discriminator_for_locus): Work on line directly.
1419 (same_line_p): Pass in expanded locus1 as well.
1420 (assign_discriminators): Avoid redundant location expansions.
1422 2018-06-05 Richard Biener <rguenther@suse.de>
1424 PR tree-optimization/86046
1425 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1426 if required after clearing TREE_ADDRESSABLE.
1428 2018-06-05 Richard Biener <rguenther@suse.de>
1430 PR tree-optimization/86047
1431 * tree-ssa-loop.c (for_each_index): Glob handling of all
1432 decls and constants and really handle all of them.
1434 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1437 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1438 qualifier_void_pointer and qualifier_const_void_pointer.
1439 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1440 (arm_init_builtins): Handle the above.
1441 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1442 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1443 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1446 2018-06-05 Martin Liska <mliska@suse.cz>
1448 * auto-profile.c (read_autofdo_file): Do not use
1449 gcov_ctr_summary struct.
1450 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1451 * coverage.c (struct counts_entry): Likewise.
1452 (read_counts_file): Read just single summary entry.
1453 (get_coverage_counts): Use gcov_summary struct.
1454 * coverage.h (get_coverage_counts): Likewise.
1455 * gcov-dump.c (dump_working_sets): Likewise.
1456 (tag_summary): Dump just single summary.
1457 * gcov-io.c (gcov_write_summary): Write just histogram
1459 (gcov_read_summary): Read just single summary.
1460 (compute_working_sets): Use gcov_summary struct.
1461 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1462 of GCOV_COUNTERS_SUMMABLE.
1463 (GCOV_COUNTERS_SUMMABLE): Remove.
1464 (GCOV_FIRST_VALUE_COUNTER): Replace with
1465 GCOV_COUNTER_V_INTERVAL.
1466 (struct gcov_ctr_summary): Remove.
1467 (struct gcov_summary): Directly use fields of former
1469 (compute_working_sets): Use gcov_summary struct.
1470 * gcov.c (read_count_file): Do not use ctrs fields.
1471 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1473 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1475 * profile.c: Likewise.
1476 * profile.h: Likewise.
1478 2018-06-05 Martin Liska <mliska@suse.cz>
1480 PR gcov-profile/84846
1481 * gcov.c (output_lines): Print working directory only
1482 in intermediate format.
1484 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1486 * config/s390/s390-builtin-types.def: Add void function type.
1487 * config/s390/s390-builtins.def: Use the function type for the
1490 2018-06-04 Jim Wilson <jimw@sifive.com>
1492 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1494 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1495 and maybe_eh_return. Change regno to unsigned int. Use new args to
1496 handle EH_RETURN_DATA_REGNO registers properly.
1497 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1498 (riscv_expand_epilogue): Update comment. Change argument name and
1499 type. Update code to use new name and type. Pass new args to
1500 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1502 * config/riscv/riscv.md (NORMAL_RETURN): New.
1503 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1504 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1505 (eh_return): Call gen_eh_return_internal and emit barrier.
1506 (eh_return_internal): Call riscv_expand_epilogue.
1508 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1510 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1511 bit_insertion field and declare can_be_merged_into method.
1512 (merged_store_group::can_be_merged_into): New method.
1513 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1514 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1515 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1517 2018-06-04 Richard Biener <rguenther@suse.de>
1519 PR tree-optimization/85955
1520 * builtins.c (fold_builtin_sincos): Convert pointers to
1521 destination to appropriate type before dereferencing.
1523 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1525 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1527 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1529 * expr.c (expand_expr_real_1): Force the operand into memory if
1530 its TYPE_MODE is BLKmode and if there is no integer mode for
1531 the number of bits being extracted.
1533 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1537 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1538 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1540 2018-06-04 Richard Biener <rguenther@suse.de>
1542 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1543 (cleanup_tree_cfg_noloop): ... single caller. Do
1544 start_recording_case_labels later.
1546 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1548 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1549 to _IMMINTRIN_H_INCLUDED.
1550 * config/i386/pconfigintrin.h: Ditto.
1551 * config/i386/waitpkgintrin.h: Ditto.
1552 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1553 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1554 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1555 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1556 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1557 waitpkgintrin.h and cldemoteintrin.h.
1559 2018-06-04 Richard Biener <rguenther@suse.de>
1561 PR tree-optimization/86038
1562 * tracer.c (find_best_successor): Check probability for
1563 being initialized, bail out if not.
1565 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1568 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1569 of bits to ignore when comparing architectures.
1571 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1573 PR tree-optimization/69615
1574 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1575 maximum or minimum of the type, try to merge it also as if
1576 range1 is + [-, x - 1] or + [x + 1, -].
1579 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1581 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1583 PR tree-optimization/86034
1584 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1585 the unsigned bitfield type in a bit insertion sequence if it does not
1586 have a larger precision than the bitfield size.
1587 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1589 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1591 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1593 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1595 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1596 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1597 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1598 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1600 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1602 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1603 Disable -fdelete-null-pointer-checks for ELF toolchain.
1605 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1606 Kito Cheng <kito.cheng@gmail.com>
1608 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1609 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1610 (nds32*-*-*): Add glibc and uclibc conditions.
1611 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1612 (TARGET_EXCEPT_UNWIND_INFO): Define.
1613 * config/nds32/elf.h: New file.
1614 * config/nds32/linux.h: New file.
1615 * config/nds32/nds32-elf.opt: New file.
1616 * config/nds32/nds32-linux.opt: New file.
1617 * config/nds32/nds32-fp-as-gp.c
1618 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1619 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1621 (nds32_asm_file_end): Ditto.
1622 (nds32_print_operand): Ditto.
1623 (nds32_insert_attributes): Ditto.
1624 (nds32_init_libfuncs): New function.
1625 (TARGET_HAVE_TLS): Define.
1626 (TARGET_INIT_LIBFUNCS): Define.
1627 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1629 (TARGET_ELF): Apply different mcmodel setting.
1630 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1631 been migrated into elf.h and linux.h files.
1632 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1633 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1634 (mcmodel): The content has been migrated into nds32-elf.opt and
1635 nds32-linux.opt files.
1636 * config/nds32/t-elf: New file.
1637 * config/nds32/t-linux: New file.
1639 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1640 Shiva Chen <shiva0217@gmail.com>
1642 * config/nds32/constants.md (unspec_volatile_element): Add
1643 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1644 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1646 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1647 (make_pass_nds32_fp_as_gp): Declare.
1648 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1650 (nds32_asm_function_end_prologue): Remove unused asm output.
1651 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1652 (nds32_asm_file_start): Output necessary fp_as_gp information.
1653 (nds32_option_override): Adjust register usage.
1654 (nds32_expand_prologue): Consider fp_as_gp situation.
1655 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1656 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1659 (simple_return): Ditto.
1660 (omit_fp_begin): Output special directive for fp_as_gp.
1661 (omit_fp_end): Output special directive for fp_as_gp.
1662 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1663 mforbid-fp-as-gp): New options.
1665 2018-06-01 Mark Wielaard <mark@klomp.org>
1667 * dwarf2out.c (dwarf2out_finish): Remove generation of
1668 DW_AT_loclists_base.
1670 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1672 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1673 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1674 (struct merged_store_group): Add bit_insertion field.
1675 (dump_char_array): Use standard hexadecimal format.
1676 (merged_store_group::merged_store_group): Set bit_insertion to false.
1677 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1678 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1679 also print the mask in the dump file.
1680 (pass_store_merging::gate): Minor tweak.
1681 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1682 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1683 stores with INTEGER_CST stores.
1684 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1685 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1686 and use it throughout. Generate bit insertion sequences if need be.
1687 (pass_store_merging::process_store): Remove redundant condition.
1688 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1690 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1692 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1693 the 128-bit floating point types. Fix function comment.
1695 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1697 * config/aarch64/aarch64-simd.md
1698 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1700 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1703 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1705 PR tree-optimization/85989
1706 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1708 (backprop::intersect_uses): Check it when deciding whether this
1709 is a backedge reference.
1710 (backprop::process_block): Add each phi to m_visited_phis
1711 after visiting it, then clear it at the end.
1713 2018-06-01 Richard Biener <rguenther@suse.de>
1715 * tree-vectorizer.h (vect_dr_stmt): New function.
1716 (vect_get_load_cost): Adjust.
1717 (vect_get_store_cost): Likewise.
1718 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1719 Use vect_dr_stmt instead of DR_SMTT.
1720 (vect_record_base_alignments): Likewise.
1721 (vect_calculate_target_alignment): Likewise.
1722 (vect_compute_data_ref_alignment): Likewise and make static.
1723 (vect_update_misalignment_for_peel): Likewise.
1724 (vect_verify_datarefs_alignment): Likewise.
1725 (vector_alignment_reachable_p): Likewise.
1726 (vect_get_data_access_cost): Likewise. Pass down
1727 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1728 (vect_get_peeling_costs_all_drs): Likewise.
1729 (vect_peeling_hash_get_lowest_cost): Likewise.
1730 (vect_enhance_data_refs_alignment): Likewise.
1731 (vect_find_same_alignment_drs): Likewise.
1732 (vect_analyze_data_refs_alignment): Likewise.
1733 (vect_analyze_group_access_1): Likewise.
1734 (vect_analyze_group_access): Likewise.
1735 (vect_analyze_data_ref_access): Likewise.
1736 (vect_analyze_data_ref_accesses): Likewise.
1737 (vect_vfa_segment_size): Likewise.
1738 (vect_small_gap_p): Likewise.
1739 (vectorizable_with_step_bound_p): Likewise.
1740 (vect_prune_runtime_alias_test_list): Likewise.
1741 (vect_analyze_data_refs): Likewise.
1742 (vect_supportable_dr_alignment): Likewise.
1743 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1744 (vect_gen_prolog_loop_niters): Likewise.
1745 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1746 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1748 (vect_recog_mask_conversion_pattern): Likewise.
1749 (vect_try_gather_scatter_pattern): Likewise.
1750 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1751 to vect_get_store_cost.
1752 (vect_get_store_cost): Get stmt_info instead of DR.
1753 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1754 (vect_get_load_cost): Get stmt_info instead of DR.
1756 2018-06-01 Richard Biener <rguenther@suse.de>
1759 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1760 vars wrapped in MEM_REFs.
1762 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1764 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1765 Fix subreg tests so that we only return a choice between
1766 GENERAL_REGS and FP_REGS if the original classes included both.
1768 2018-06-01 Richard Biener <rguenther@suse.de>
1771 * tree-ssa-structalias.c (get_function_part_constraint):
1772 Handle NULL fi->decl.
1773 (find_func_aliases_for_call): Properly handle indirect
1774 fi from direct call.
1775 (find_func_clobbers): Likewise.
1776 (ipa_pta_execute): Likewise.
1777 (create_variable_info_for): For functions that are ifunc_resolver
1778 resolve to a varinfo that contains the result of the resolver call.
1779 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1782 2018-05-31 Michael Collison <michael.collison@arm.com>
1784 * config/aarch64/aarch64.md:
1785 (*fix_to_zero_extenddfdi2): New pattern.
1786 * gcc.target/aarch64/fix_extend1.c: New testcase.
1788 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1792 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1793 and BUILT_IN_STRNCMP_EQ.
1794 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1795 BUILT_IN_STRNCMP_EQ.
1796 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1797 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1798 (gimple_fold_builtin): Likewise.
1799 * tree-ssa-strlen.c (compute_string_length): New function.
1800 (determine_min_obsize): New function.
1801 (handle_builtin_string_cmp): New function to handle calls to
1802 string compare functions.
1803 (strlen_optimize_stmt): Add handling to builtin string compare
1805 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1806 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1807 * tree.c (build_common_builtin_nodes): Add new defines of
1808 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1810 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1813 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1814 functions with naked attribute.
1816 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1818 * config/i386/sse.md (avx_vec_concat<mode>):
1819 Substitute concat_tg_mode mode attribute with xtg_mode.
1820 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1821 (concat_tg_mode): Remove mode attribute.
1823 2018-05-31 Martin Sebor <msebor@redhat.com>
1826 * calls.c (alloc_max_size): Correct a logic error/typo.
1827 Treat excessive arguments as infinite. Warn for invalid arguments.
1828 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1830 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1833 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1834 and movx for Haswell.
1836 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1837 Cesar Philippidis <cesar@codesourcery.com>
1840 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1841 when emitting error on private/firstprivate reductions.
1842 * omp-low.c (lower_omp_target): Avoid reference-type processing
1843 on pointers for firstprivate clause.
1845 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1847 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1850 * config/aarch64/aarch64-simd.md
1851 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1852 (aarch64_ld1_x3_<mode>): Likewise
1853 (aarch64_st1x2<VALLDIF:mode>): Likewise
1854 (aarch64_st1_x2_<mode>): Likewise
1855 (aarch64_st1x3<VALLDIF:mode>): Likewise
1856 (aarch64_st1_x3_<mode>): Likewise
1857 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1858 (vld1_s8_x3): Likewise.
1859 (vld1_u16_x3): Likewise.
1860 (vld1_s16_x3): Likewise.
1861 (vld1_u32_x3): Likewise.
1862 (vld1_s32_x3): Likewise.
1863 (vld1_u64_x3): Likewise.
1864 (vld1_s64_x3): Likewise.
1865 (vld1_f16_x3): Likewise.
1866 (vld1_f32_x3): Likewise.
1867 (vld1_f64_x3): Likewise.
1868 (vld1_p8_x3): Likewise.
1869 (vld1_p16_x3): Likewise.
1870 (vld1_p64_x3): Likewise.
1871 (vld1q_u8_x3): Likewise.
1872 (vld1q_s8_x3): Likewise.
1873 (vld1q_u16_x3): Likewise.
1874 (vld1q_s16_x3): Likewise.
1875 (vld1q_u32_x3): Likewise.
1876 (vld1q_s32_x3): Likewise.
1877 (vld1q_u64_x3): Likewise.
1878 (vld1q_s64_x3): Likewise.
1879 (vld1q_f16_x3): Likewise.
1880 (vld1q_f32_x3): Likewise.
1881 (vld1q_f64_x3): Likewise.
1882 (vld1q_p8_x3): Likewise.
1883 (vld1q_p16_x3): Likewise.
1884 (vld1q_p64_x3): Likewise.
1885 (vst1_s64_x2): Likewise.
1886 (vst1_u64_x2): Likewise.
1887 (vst1_f64_x2): Likewise.
1888 (vst1_s8_x2): Likewise.
1889 (vst1_p8_x2): Likewise.
1890 (vst1_s16_x2): Likewise.
1891 (vst1_p16_x2): Likewise.
1892 (vst1_s32_x2): Likewise.
1893 (vst1_u8_x2): Likewise.
1894 (vst1_u16_x2): Likewise.
1895 (vst1_u32_x2): Likewise.
1896 (vst1_f16_x2): Likewise.
1897 (vst1_f32_x2): Likewise.
1898 (vst1_p64_x2): Likewise.
1899 (vst1q_s8_x2): Likewise.
1900 (vst1q_p8_x2): Likewise.
1901 (vst1q_s16_x2): Likewise.
1902 (vst1q_p16_x2): Likewise.
1903 (vst1q_s32_x2): Likewise.
1904 (vst1q_s64_x2): Likewise.
1905 (vst1q_u8_x2): Likewise.
1906 (vst1q_u16_x2): Likewise.
1907 (vst1q_u32_x2): Likewise.
1908 (vst1q_u64_x2): Likewise.
1909 (vst1q_f16_x2): Likewise.
1910 (vst1q_f32_x2): Likewise.
1911 (vst1q_f64_x2): Likewise.
1912 (vst1q_p64_x2): Likewise.
1913 (vst1_s64_x3): Likewise.
1914 (vst1_u64_x3): Likewise.
1915 (vst1_f64_x3): Likewise.
1916 (vst1_s8_x3): Likewise.
1917 (vst1_p8_x3): Likewise.
1918 (vst1_s16_x3): Likewise.
1919 (vst1_p16_x3): Likewise.
1920 (vst1_s32_x3): Likewise.
1921 (vst1_u8_x3): Likewise.
1922 (vst1_u16_x3): Likewise.
1923 (vst1_u32_x3): Likewise.
1924 (vst1_f16_x3): Likewise.
1925 (vst1_f32_x3): Likewise.
1926 (vst1_p64_x3): Likewise.
1927 (vst1q_s8_x3): Likewise.
1928 (vst1q_p8_x3): Likewise.
1929 (vst1q_s16_x3): Likewise.
1930 (vst1q_p16_x3): Likewise.
1931 (vst1q_s32_x3): Likewise.
1932 (vst1q_s64_x3): Likewise.
1933 (vst1q_u8_x3): Likewise.
1934 (vst1q_u16_x3): Likewise.
1935 (vst1q_u32_x3): Likewise.
1936 (vst1q_u64_x3): Likewise.
1937 (vst1q_f16_x3): Likewise.
1938 (vst1q_f32_x3): Likewise.
1939 (vst1q_f64_x3): Likewise.
1940 (vst1q_p64_x3): Likewise.
1942 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1944 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1945 user_label_prefix to name.
1947 * tree-core.h: Update comment about the format of NAME string
1948 passed to handler in attribute_spec.
1950 * config/msp430/msp430.md: Remove erroneous subreg expression from
1951 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1952 zero_extend{q,h}isi2.
1954 2018-05-30 Borislav Petkov <bp@suse.de>
1956 * doc/extend.texi: Document some architecture specific
1957 constraints and sort entries.
1959 2018-05-30 Martin Sebor <msebor@redhat.com>
1962 * builtins.c (expand_builtin_stpcpy_1): New function.
1963 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1964 only if the former succeeds.
1966 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1968 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1971 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1973 * doc/invoke.texi (-flinker-output): Document
1975 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1977 * passes.c (ipa_write_summaries): Only modify statements if body
1979 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1980 incrementally linking.
1981 (ipa_passes): Likewise.
1982 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1983 pass down resolution info.
1984 * common.opt (flag_incremental_link): Update info.
1985 * gcc.c (plugin specs): Turn flinker-output=* to
1986 -plugin-opt=-linker-output-known
1987 * toplev.c (compile_file): Also cut compilation when doing incremental
1989 * flag-types. (enum lto_partition_model): Add
1990 LTO_LINKER_OUTPUT_NOLTOREL.
1991 (invoke.texi): Add -flinker-output docs.
1992 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
1993 link same way as WPA; do not stream in dead initializers.
1995 * dwarf2out.c (dwarf2out_die_ref_for_decl,
1996 darf2out_register_external_decl): Support incremental link.
1998 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2000 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2002 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2004 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2005 it down to simple_object_copy_lto_debug_sections.
2006 (run_gcc): Determine incremental LTO link time and configure
2007 lto1 into non-wpa mode, disable renaming of debug sections.
2009 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2011 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2012 descriptions of various incorrectly documented functions.
2014 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2017 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2018 address check not strict.
2020 2018-05-30 Richard Biener <rguenther@suse.de>
2022 PR tree-optimization/85964
2023 * tracer.c (better_p): Drop initialized count check, we only
2024 call the function with initialized counts now.
2025 (find_best_successor): Do find a best edge if one
2026 has uninitialized count.
2027 (find_best_predecessor): Likewise. Do BB frequency check only
2028 if count is initialized.
2030 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2032 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2033 (aarch64_ldrstr_offset_compare): New.
2034 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2035 load/store orderings.
2036 (aarch64_gen_adjusted_ldpstp): Likewise.
2038 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2040 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2041 Check for subset of GENERAL_REGS and FP_REGS.
2042 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2045 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2047 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2048 and wi::to_poly_offset. Add the current offset and then check
2049 whether the sum fits, rather than using an unchecked addition of
2050 a checked term. Check for a shwi rather than a uhwi.
2051 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2052 (store_constructor): Use poly_int_tree_p.
2053 (expand_expr_real_1): Likewise.
2054 * function.c (assign_temp): Likewise.
2055 * fold-const.c (const_binop): Use poly_int_tree_p and
2057 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2059 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2060 to_poly_offset for MEM offsets.
2061 * ipa-icf.c (sem_variable::equals): Likewise.
2062 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2063 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2064 wi::to_poly_offset for BIT_FIELD_REF offsets.
2065 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2067 * var-tracking.c (emit_note_insn_var_location): Use
2068 tree_to_poly_uint64.
2070 2018-05-29 Jim Wilson <jimw@sifive.com>
2072 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2074 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2077 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2078 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2080 (sse4_1_round<mode>2): Use nonimmediate_operand
2081 for operand 1 predicate.
2083 2018-05-29 Martin Sebor <msebor@redhat.com>
2084 Richard Biener <rguenther@suse.de>
2087 * calls.c (get_size_range): Call determine_value_range instead
2088 of get_value_range..
2089 * tree-vrp.h (determine_value_range): Declared new function.
2090 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2092 2018-05-29 Richard Biener <rguenther@suse.de>
2094 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2095 sure to use non-pattern stmts for get_earlier_stmt arguments.
2096 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2097 called on pattern stmts.
2098 (get_later_stmt): Likewise.
2100 2018-05-29 Martin Liska <mliska@suse.cz>
2102 PR gcov-profile/85759
2103 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2106 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2108 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2110 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2114 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2115 VEC_PACK_FLOAT_EXPR): New tree codes.
2116 * tree-pretty-print.c (op_code_prio): Handle
2117 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2118 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2119 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2120 * tree-inline.c (estimate_operator_cost): Likewise.
2121 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2122 * fold-const.c (const_binop): Likewise.
2123 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2124 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2125 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2126 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2127 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2128 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2129 * expr.c (expand_expr_real_2): Likewise.
2130 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2131 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2132 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2134 * optabs.c (expand_widen_pattern_expr): For
2135 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2136 sign from result type rather than operand's type.
2137 (expand_binop_directly): For vec_packu_float_optab and
2138 vec_packs_float_optab allow result type to be different from operand's
2140 * optabs-tree.c (optab_for_tree_code): Handle
2141 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2142 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2143 * tree-vect-generic.c (expand_vector_operations_1): Handle
2144 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2145 VEC_PACK_FLOAT_EXPR.
2146 * tree-vect-stmts.c (supportable_widening_operation): Handle
2148 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2149 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2150 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2151 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2152 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2154 (vec_pack<floatprefix>_float_<mode>): New expander.
2155 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2157 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2158 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2159 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2160 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2161 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2163 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2164 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2165 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2166 VEC_PACK_FLOAT_EXPR): Document.
2168 2018-05-29 Richard Biener <rguenther@suse.de>
2170 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2172 (stmt_vec_info_vec): Make pointer.
2173 (init_stmt_vec_info_vec): Remove.
2174 (free_stmt_vec_info_vec): Likewise.
2175 (set_stmt_vec_info_vec): New function.
2176 (free_stmt_vec_infos): Likewise.
2177 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2178 (set_vinfo_for_stmt): Likewise.
2179 (get_earlier_stmt): Likewise.
2180 (get_later_stmt): Likewise.
2181 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2182 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2183 (vec_info::~vec_info): Free stmt_vec_infos.
2184 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2185 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2186 (pass_slp_vectorize::execute): Likewise.
2187 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2188 (free_stmt_vec_info_vec): Likewise.
2189 (set_stmt_vec_info_vec): New function.
2190 (free_stmt_vec_infos): Likewise.
2191 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2192 the global stmt_vec_info_vec.
2193 * tree-parloops.c (gather_scalar_reductions): Use
2194 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2197 2018-05-29 Richard Biener <rguenther@suse.de>
2199 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2201 2018-05-29 Martin Liska <mliska@suse.cz>
2202 David Malcolm <dmalcolm@redhat.com>
2204 * vec.c (test_reverse): New.
2205 (vec_c_tests): Add new test.
2206 * vec.h (vl_ptr>::reverse): New function.
2208 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2210 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2212 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2215 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2217 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2219 2018-05-28 Richard Biener <rguenther@suse.de>
2221 PR tree-optimization/85933
2222 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2223 look at stmts marked as vectorizable.
2225 2018-05-28 Richard Biener <rguenther@suse.de>
2227 PR tree-optimization/85934
2228 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2229 vector boolean check before scalar optimization.
2231 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2233 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2236 2018-05-28 Mark Wielaard <mark@klomp.org>
2238 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2239 if it is an expression containing a minus sign.
2241 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2243 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2245 2018-05-27 Paul Koning <ni1d@arrl.net>
2247 * config/pdp11/pdp11.md (truncsihi2): Remove.
2249 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2250 Chung-Ju Wu <jasonwucj@gmail.com>
2252 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2254 (unaligned_store_dw): Ditto.
2255 * config/nds32/nds32-memory-manipulation.c
2256 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2257 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2258 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2259 (emit_setmem_word_loop): Rename to ...
2260 (emit_setmem_doubleword_loop): ... this.
2261 (nds32_gen_dup_4_byte_to_word_value): New function.
2262 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2263 (nds32_expand_setmem_loop): Refine implementation.
2264 (nds32_expand_setmem_loop_v3m): Ditto.
2265 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2268 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2270 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2272 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2274 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2275 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2276 machine->attr_no_prologue_p.
2277 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2278 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2279 (nds32_expand_epilogue): Consider attr_naked_p.
2280 (nds32_expand_epilogue_v3pop): Likewise.
2281 (nds32_can_use_return_insn): Likewise.
2282 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2283 attr_no_prologue_p fields.
2284 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2286 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2289 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2291 * config/i386/sse.md
2292 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2294 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2296 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2298 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2300 (*<floatsuffix>floatv2div2sf2): Rename to ...
2301 (*float<floatunssuffix>v2div2sf2): ... this.
2302 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2303 (float<floatunssuffix>v2div2sf2_mask): ... this.
2304 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2305 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2306 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2308 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2310 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2312 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2314 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2316 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2318 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2319 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2320 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2321 gen_ufix_truncv8dfv8si2.
2322 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2323 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2324 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2325 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2326 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2327 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2328 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2329 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2331 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2335 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2337 2018-05-25 Jim Wilson <jimw@sifive.com>
2339 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2340 * config/riscv/riscv.c (struct machine_function): Add
2341 interrupt_handler_p and attribute_checked_p fields.
2342 (riscv_attribute_table): Add interrupt.
2343 (riscv_interrupt_type_p): New.
2344 (riscv_save_reg_p): Save extra regs for interrupt handler.
2345 (riscv_use_save_libcall): Return false for interrupt handler.
2346 (riscv_first_stack_step): Add forward declaration.
2347 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2348 for interrupt handler with large frame. Use it for saved reg list.
2349 (riscv_expand_prologue): Move flag_stack_usage_info support to
2350 eliminate duplication.
2351 (riscv_expand_epilogue): Generate mret for interrupt handler.
2352 (riscv_epilogue_uses): New.
2353 (riscv_can_use_return_insn): Return false for interrupt handler.
2354 (riscv_function_ok_for_sibcall): Likewise.
2355 (riscv_set_current_function): Add interrupt handler support.
2356 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2357 * config/riscv/riscv.md (UNSPECV_MRET): New.
2359 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2361 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2363 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2365 PR tree-optimization/85712
2366 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2367 this candidate has already been replaced in-situ by a copy.
2369 2018-05-25 Jason Merrill <jason@redhat.com>
2371 PR c++/80485 - inline function non-zero address.
2372 * symtab.c (nonzero_address): Check DECL_COMDAT.
2374 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2377 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2378 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2379 extension of SImode operation. Use const123_operand predicate.
2380 (*saddsi_1): Remove.
2381 (*saddl_se_1): Ditto.
2383 (*ssubl_se_1): Ditto.
2384 * config/alpha/predicates.md (const123_operand): New predicate.
2385 * config/alpha/constraints.md (P): Use IN_RANGE.
2387 2018-05-25 Richard Biener <rguenther@suse.de>
2389 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2391 (ref_maybe_used_by_stmt_p): Likewise.
2392 (stmt_may_clobber_ref_p): Likewise.
2393 (stmt_may_clobber_ref_p_1): Likewise.
2394 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2396 (ref_maybe_used_by_stmt_p): Likewise.
2397 (stmt_may_clobber_ref_p): Likewise.
2398 (stmt_may_clobber_ref_p_1): Likewise.
2399 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2400 the alias oracle to disambiguate DRs with stmts DR analysis
2402 (vect_analyze_data_refs): Do not give up on not analyzable
2403 DRs for BB vectorization. Remove code truncating the dataref
2406 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2409 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2410 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2411 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2413 2018-05-25 Richard Biener <rguenther@suse.de>
2415 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2416 function, combining stmt data ref gathering and fatal analysis
2418 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2419 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2420 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2421 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2422 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2424 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2426 PR tree-optimization/85720
2427 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2428 SCC if all partitions are builtins.
2429 (version_loop_by_alias_check): New parameter. Generate cancelable
2430 runtime alias check if all partitions are builtins.
2431 (distribute_loop): Update call to above function.
2433 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2435 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2436 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2437 (parm_default_def_partition_arg): Ditto.
2438 (set_parm_default_def_partition): Ditto.
2439 (get_parm_default_def_partitions): Ditto and make it static.
2440 (get_undefined_value_partitions): Ditto and make it static.
2441 (remove_ssa_form): Refactor call to init_var_map here.
2442 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2443 computation for loop region.
2444 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2445 (register_default_def): Delete.
2446 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2447 (parm_default_def_partition_arg): Ditto.
2448 (set_parm_default_def_partition): Ditto.
2449 (get_parm_default_def_partitions): Ditto and make it static.
2450 (get_undefined_value_partitions): Ditto and make it static.
2451 (coalesce_with_default, coalesce_with_default): Update comment.
2452 (create_coalesce_list_for_region): New func factored out from
2453 create_outofssa_var_map.
2454 (populate_coalesce_list_for_outofssa): New func factored out from
2455 create_outofssa_var_map and coalesce_ssa_name.
2456 (create_outofssa_var_map): Delete.
2457 (coalesce_ssa_name): Refactor to support live range computation.
2458 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2459 (get_parm_default_def_partitions): Delete.
2460 (get_undefined_value_partitions): Ditto.
2461 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2462 computation for loop region.
2463 (new_tree_live_info, loe_visit_block): Ditto.
2464 (live_worklist, set_var_live_on_entry): Ditto.
2465 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2466 * tree-ssa-live.h (struct _var_map): New fields.
2467 (init_var_map): Change decl.
2468 (region_contains_p): New.
2470 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2472 * tree-ssa-live.h (live_merge_and_clear): Delete.
2474 2018-05-25 Richard Biener <rguenther@suse.de>
2477 * tree-dump.c (dequeue_and_dump): Remove access to removed
2478 operand 2 of a SWITCH_EXPR.
2480 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2482 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2483 multiplication and division.
2484 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2485 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2486 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2487 (cond_udiv_optab, cond_umod_optab): New optabs.
2488 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2489 (IFN_COND_RDIV): New internal functions.
2490 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2491 TRUNC_MOD_EXPR and RDIV_EXPR.
2492 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2493 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2495 (SVE_INT_BINARY): Include mult.
2496 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2497 (optab, sve_int_op): Handle mult.
2498 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2500 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2501 for SVE_INT_BINARY_SD.
2503 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2505 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2506 (optab, sve_int_op): Handle div and udiv.
2507 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2508 for SVE_INT_BINARY_SD.
2509 (*<optab><mode>3): New insn for the same.
2511 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2513 * tree-vect-patterns.c: Include predict.h.
2514 (vect_recog_divmod_pattern): Restrict check for division support
2515 to when optimizing for size.
2517 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2519 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2520 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2521 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2522 (gimple_match_op::set_op): Likewise.
2523 (gimple_resimplify4): Declare.
2524 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2525 (expr::gen_transform): Likewise.
2526 (decision_tree::gen): Generate a simplification routine for 4 operands.
2527 * gimple-match-head.c (gimple_simplify): Add an overload for
2528 4 operands. In the top-level function, handle up to 4 call
2529 arguments and call gimple_resimplify4.
2530 (gimple_resimplify4): New function.
2531 (build_call_internal): Pass a fourth operand.
2532 (maybe_push_to_seq): Likewise.
2533 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2534 Fold VEC_COND_EXPRs of an operation and a default value into
2535 an IFN_COND_* function if possible.
2536 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2538 (SVE_COND_FP_BINARY): Include them.
2539 (optab, sve_fp_op): Handle them.
2540 (SVE_INT_BINARY_REV): New code iterator.
2541 (SVE_COND_FP_BINARY_REV): New int iterator.
2542 (commutative): New int attribute.
2543 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2545 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2547 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2548 (*cond_<optab><mode>): New patterns for reversed operands.
2550 2018-05-25 Richard Biener <rguenther@suse.de>
2552 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2553 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2554 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2555 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2556 * tree-vect-data-refs.c (everywhere): Adjust users.
2557 * tree-vect-loop.c (everywhere): Likewise.
2558 * tree-vect-slp.c (everywhere): Likewise.
2559 * tree-vect-stmts.c (everywhere): Likewise.
2560 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2562 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2564 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2566 (gcc_cv_as_section_exclude): ... this.
2567 Try Solaris as #exclude syntax.
2568 * configure: Regenerate.
2569 * config.in: Regenerate.
2570 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2572 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2573 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2575 * varasm.c (default_elf_asm_named_section): Don't check if
2576 HAVE_GAS_SECTION_EXCLUDE is defined.
2578 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2580 * doc/md.texi: Update the documentation of the cond_* optabs
2581 to mention the new final operand. Fix GET_MODE_NUNITS call.
2582 Describe the scalar case too.
2583 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2584 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2586 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2587 (get_conditional_internal_fn): Update comment.
2588 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2589 accumulator value as a final argument to conditional functions.
2590 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2591 a define_expand and add an "else" operand. Assert for now that
2592 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2593 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2594 (*cond_<optab><mode>): New patterns.
2595 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2596 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2597 (UNSPEC_COND_EOR): Delete.
2598 (optab): Remove associated mappings.
2599 (SVE_INT_BINARY): New code iterator.
2600 (sve_int_op): Remove int attribute and add "minus" to the code
2602 (SVE_COND_INT_OP): Delete.
2603 (SVE_COND_FP_OP): Rename to...
2604 (SVE_COND_FP_BINARY): ...this.
2606 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2608 * optabs.c (can_reuse_operands_p): New function.
2609 (maybe_legitimize_operands): Try to reuse the results for
2612 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2614 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2615 Add {q} suffix to insn mnemonic.
2617 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2619 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2620 (msp430_warn_func_return): New.
2622 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2624 * fold-const.c (tree_nonzero_bits): New function.
2625 * fold-const.h (tree_nonzero_bits): Likewise.
2626 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2627 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2629 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2633 * varasm.c (assemble_alias): Check ifunc_resolver only on
2636 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2639 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2640 when memory input operand is handled.
2642 2018-05-24 Luis Machado <luis.machado@linaro.org>
2644 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2646 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2648 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2650 * match.pd: Delay FMA folds until after vectorization.
2652 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2655 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2656 address check not strict.
2658 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2660 * gimple-match.h (gimple_match_op): New class.
2661 (mprts_hook): Replace parameters with a gimple_match_op *.
2662 (maybe_build_generic_op): Likewise.
2663 (gimple_simplified_result_is_gimple_val): Replace parameters with
2664 a const gimple_match_op *.
2665 (gimple_simplify): Replace code_helper * and tree * parameters with
2666 a gimple_match_op * parameter.
2667 (gimple_resimplify1): Replace code_helper *, tree and tree *
2668 parameters with a gimple_match_op * parameter.
2669 (gimple_resimplify2): Likewise.
2670 (gimple_resimplify3): Likewise.
2671 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2672 parameters with a gimple_match_op * parameter.
2673 * gimple-match-head.c (gimple_simplify): Change prototypes of
2674 auto-generated functions to take a gimple_match_op * instead of
2675 separate code_helper * and tree * parameters. Make the same
2676 change in the top-level overload and update calls to the
2677 gimple_resimplify routines. Update calls to the auto-generated
2678 functions and to maybe_push_res_to_seq in the publicly-facing
2679 operation-specific gimple_simplify overloads.
2680 (gimple_match_op::MAX_NUM_OPS): Define.
2681 (gimple_resimplify1): Replace rcode and ops with a single res_op
2682 parameter. Update call to gimple_simplify.
2683 (gimple_resimplify2): Likewise.
2684 (gimple_resimplify3): Likewise.
2685 (mprts_hook): Replace parameters with a gimple_match_op *.
2686 (maybe_build_generic_op): Likewise.
2687 (build_call_internal): Replace type, nargs and ops with
2688 a gimple_match_op *.
2689 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2690 with a single gimple_match_op *. Update calls to mprts_hook,
2691 build_call_internal and gimple_simplified_result_is_gimple_val.
2692 Factor out code that is common to the tree_code and combined_fn cases.
2693 * genmatch.c (expr::gen_transform): Replace tem_code and
2694 tem_ops with a gimple_match_op called tem_op. Update calls
2695 to the gimple_resimplify functions and maybe_push_res_to_seq.
2696 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2697 res_ops. Update call to the gimple_resimplify functions.
2698 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2699 (decision_tree::gen): Make the functions take a gimple_match_op *
2700 called res_op instead of separate res_code and res_ops parameters.
2701 Update call accordingly.
2702 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2703 and ops with a single res_op parameter. Update calls to
2704 maybe_build_generic_op and maybe_push_res_to_seq.
2705 (fold_stmt_1): Update calls to gimple_simplify and
2706 replace_stmt_with_simplification.
2707 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2708 and gimple_simplified_result_is_gimple_val.
2709 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2711 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2712 with a gimple_match_op *.
2713 (vn_nary_build_or_lookup): Likewise. Update call to
2714 vn_nary_build_or_lookup_1.
2715 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2716 gimple_match_op *. Update calls to the gimple_resimplify routines
2717 and to gimple_simplified_result_is_gimple_val.
2718 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2719 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2720 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2721 (visit_nary_op): Likewise.
2722 (visit_reference_op_load): Likewise.
2724 2018-05-23 Luis Machado <luis.machado@linaro.org>
2726 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2727 modifier for printing the step amount.
2729 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2732 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2735 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2737 * doc/sourcebuild.texi (Endianness): New subsubsection.
2739 2018-05-23 Luis Machado <luis.machado@linaro.org>
2741 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2742 <prefetch_dynamic_strides>: New const bool field.
2743 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2744 prefetch_dynamic_strides.
2745 (exynosm1_prefetch_tune): Likewise.
2746 (thunderxt88_prefetch_tune): Likewise.
2747 (thunderx_prefetch_tune): Likewise.
2748 (thunderx2t99_prefetch_tune): Likewise.
2749 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2751 (aarch64_override_options_internal): Update to set
2752 PARAM_PREFETCH_DYNAMIC_STRIDES.
2753 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2754 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2755 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2756 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2757 prefetch-dynamic-strides setting.
2759 2018-05-23 Luis Machado <luis.machado@linaro.org>
2761 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2762 <minimum_stride>: New const int field.
2763 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2764 minimum_stride field defaulting to -1.
2765 (exynosm1_prefetch_tune): Likewise.
2766 (thunderxt88_prefetch_tune): Likewise.
2767 (thunderx_prefetch_tune): Likewise.
2768 (thunderx2t99_prefetch_tune): Likewise.
2769 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2770 <default_opt_level>: Set to 3.
2771 (aarch64_override_options_internal): Update to set
2772 PARAM_PREFETCH_MINIMUM_STRIDE.
2773 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2774 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2775 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2776 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2777 stride is constant and is below the minimum stride threshold.
2779 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2781 * config/arm/arm-cpus.in (mode26): Delete.
2782 (armv4): Delete mode26 reference.
2783 * config/arm/arm.c (arm_configure_build_target): Delete use of
2786 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2788 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2790 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2791 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2793 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2794 (floatunsdidf2): Ditto.
2796 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2798 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2799 (fixuns_trunc<mode>si2_avx512f): Ditto.
2800 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2801 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2802 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2804 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2806 PR rtl-optimization/79985
2807 * df-scan.c (df_insn_refs_collect): Remove special case for
2808 global registers and asm statements.
2810 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2812 * extend.texi (Global Register Variables): Rewrite the bullet list.
2813 Note that the register is available for allocation. Note that access
2814 via inline asm must use constraints. Add note about async-signal
2815 handlers. Remove paragraph about automagic register selection.
2817 2018-05-23 Richard Biener <rguenther@suse.de>
2819 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2820 of fixed offset from memset VN.
2822 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2824 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2826 (alloc_cand_and_find_basis): Initialize first_interp field.
2827 (slsr_process_mul): Modify first_interp field.
2828 (slsr_process_add): Likewise.
2829 (slsr_process_cast): Modify first_interp field for each new
2831 (slsr_process_copy): Likewise.
2832 (dump_candidate): Dump first_interp field.
2833 (replace_mult_candidate): Process all interpretations, not just
2835 (replace_rhs_if_not_dup): Likewise.
2836 (replace_one_candidate): Likewise.
2838 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2840 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2842 (aarch64_needs_frame_chain): New function.
2843 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2845 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2848 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2849 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2850 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2851 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2852 as true for strict-align.
2853 (aarch64_can_inline_p): Perform checks even when callee has no
2854 attributes to check for strict alignment.
2855 * doc/extend.texi (AArch64 Function Attributes): Document
2857 * doc/invoke.texi: (AArch64 Options): Likewise.
2859 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2861 PR tree-optimization/85853
2862 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2863 the handling of the root of the node to...
2864 (vect_slp_analyze_node_operations_1): ...this new function,
2865 and run the whole thing with the child nodes' def types
2866 set according to their SLP node's def type.
2868 2018-05-23 Richard Biener <rguenther@suse.de>
2871 * tree-data-ref.c (create_runtime_alias_checks): Defer
2872 and ignore overflow warnings.
2874 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2876 PR tree-optimization/85822
2877 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2880 2018-05-23 Richard Biener <rguenther@suse.de>
2882 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2883 memset constants via native_interpret_expr.
2885 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2888 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2890 (cgraph_node::create_alias): Likewise.
2891 (cgraph_node::get_availability): Check ifunc_resolver instead
2892 of looking up ifunc attribute.
2893 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2894 * varasm.c (do_assemble_alias): Likewise.
2895 (assemble_alias): Likewise.
2896 (default_binds_local_p_3): Likewise.
2897 * cgraph.h (cgraph_node): Add ifunc_resolver.
2898 (cgraph_node::only_called_directly_or_aliased_p): Return false
2900 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2902 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2903 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2904 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2905 instead of looking up ifunc attribute.
2907 2018-05-22 Luis Machado <luis.machado@linaro.org>
2909 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2911 2018-05-22 Martin Sebor <msebor@redhat.com>
2914 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2915 only when expasion succeeds.
2916 (expand_builtin_strcmp): Same.
2917 (expand_builtin_strncmp): Same.
2919 2018-05-22 Martin Sebor <msebor@redhat.com>
2921 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2923 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2924 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2926 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2927 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2929 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2930 Add check for legitimate_address.
2931 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2932 (aarch64_swap_ldrstr_operands): New.
2933 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2936 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2937 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2939 * config/aarch64/aarch64.md: New patterns to generate stp
2941 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2942 single words and double words.
2943 (load_pair_sw, load_pair_dw): Likewise.
2944 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2946 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2948 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2949 for different mode ldpstp and add peephole for merged zero stores.
2951 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2953 (aarch64_gen_store_pair): Rename calls to match new patterns.
2954 (aarch64_gen_load_pair): Rename calls to match new patterns.
2955 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2956 (load_pair<DREG:mode><DREG2:mode>): ... This.
2957 (store_pair<mode>): Rename to...
2958 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2959 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2961 (V_INT_EQUIV): Handle SImode.
2962 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2965 2018-05-22 Martin Sebor <msebor@redhat.com>
2968 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2969 or ajust the presumed bound on an operation to avoid unnecessary
2972 2018-05-22 Martin Sebor <msebor@redhat.com>
2974 PR tree-optimization/85826
2975 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2976 assuming that a DECL necesarily has a constant size.
2978 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2981 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2983 2018-05-22 Richard Biener <rguenther@suse.de>
2985 PR tree-optimization/85834
2986 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2987 non-constant and non-zero memset arguments.
2989 2018-05-22 Martin Liska <mliska@suse.cz>
2992 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
2994 2018-05-22 Richard Biener <rguenther@suse.de>
2996 PR tree-optimization/85863
2997 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
2998 comparisons when vectype is specified.
2999 (vectorizable_condition): Do not specify vectype for
3000 vect_is_simple_cond when SLP vectorizing.
3002 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3005 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3006 define __ibm128 as long double.
3007 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3008 as a distinct type when IEEE 128-bit support is enabled.
3009 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3010 128-bit types to use the correct functions.
3011 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3012 convert between 128-bit floating point types that have different
3013 modes but the same representation, instead of using gen_lowpart to
3015 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3017 (IFKF_reg): New attributes to give the register constraints for
3019 (extend<mode>tf2_internal): New insns to mark an explicit
3020 conversion between 128-bit floating point types that have a
3021 different mode but share the same representation.
3023 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3025 PR tree-optimization/85814
3026 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3027 a null return from get_strinfo when unsharing the next
3028 strinfo in the chain.
3030 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3033 * varasm.c (weak_finish): Clean up weak_decls.
3035 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3037 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3038 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3039 UNSPEC_UADALP values.
3040 * config/aarch64/iterators.md (ABAL): New int iterator.
3043 (sur): Add mappings for the above.
3044 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3046 (aarch64_<sur>abal<mode>_4): Likewise.
3047 (aarch64_<sur>adalp<mode>_3): Likewise.
3048 (<sur>sadv16qi): New define_expand.
3050 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3052 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3053 (*movdf_internal): Ditto.
3054 (*rcpsf2_sse): Ditto.
3055 (*rsqrtsf2_sse): Ditto.
3056 (*sqrt<mode>2_sse): Ditto.
3058 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3060 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3062 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3063 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3064 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3065 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3067 * config/aarch64/arm_neon.h: Likewise.
3068 * config/aarch64/iterators.md (VQ_I): New.
3070 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3072 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3074 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3077 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3079 * config/nds32/constraints.md (S): New constraint.
3080 * config/nds32/nds32.md (call_internal): Use constraint S.
3081 (call_value_internal): Likewise.
3082 (sibcall_internal): Likewise.
3083 (sibcall_value_internal): Likewise.
3085 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3086 Chung-Ju Wu <jasonwucj@gmail.com>
3088 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3091 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3092 Chung-Ju Wu <jasonwucj@gmail.com>
3094 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3095 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3096 (nds32_rtx_costs_impl): Simplify.
3097 (nds32_address_cost_impl): Simplify.
3098 (nds32_init_rtx_costs): New function.
3099 (nds32_rtx_costs_speed_prefer): Likewise.
3100 (nds32_rtx_costs_size_prefer): Likewise.
3101 (nds32_address_cost_speed_prefer): Likewise.
3102 (nds32_address_cost_speed_fwprop): Likewise.
3103 (nds32_address_cost_size_prefer): Likewise.
3104 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3105 * config/nds32/nds32.c (nds32_option_override): Use
3106 nds32_init_rtx_costs function.
3108 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3110 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3111 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3112 (TARGET_PIPELINE_N8): Likewise.
3113 (TARGET_PIPELINE_N10): Likewise.
3114 (TARGET_PIPELINE_N13): Likewise.
3115 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3117 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3119 * config/nds32/nds32-fpu.md: Update copyright year.
3121 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3123 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3125 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3127 * config/nds32/nds32.c
3128 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3129 * config/nds32/nds32.opt (minline-asm-r15): New option.
3131 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3133 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3135 * config/nds32/nds32.md (abssi2): New pattern.
3137 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3139 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3140 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3141 Merge insn pattern from sse_cvtsi2ss<round_name> and
3142 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3143 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3144 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3145 using SWI48 mode iterator.
3146 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3147 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3148 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3149 pattern from sse_cvttss2si<round_saeonly_name>
3150 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3151 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3152 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3153 using SWI48 mode iterator.
3154 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3155 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3156 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3157 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3158 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3159 using SWI48 mode iterator.
3160 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3161 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3162 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3163 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3164 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3165 SWI48 mode iterator.
3166 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3167 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3168 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3169 pattern from sse_cvttsd2si<round_saeonly_name>
3170 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3172 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3174 * config/nds32/nds32-md-auxiliary.c
3175 (nds32_valid_smw_lwm_base_p): Refine.
3176 (nds32_output_smw_single_word): Refine.
3177 (nds32_output_smw_double_word): New.
3178 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3180 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3182 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3183 (nds32_output_stack_pop): Refine.
3184 (nds32_expand_unaligned_load): Refine.
3185 (nds32_expand_unaligned_store): Refine.
3187 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3188 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config/nds32/constants.md: Add TP_REGNUM constant.
3191 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3192 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3194 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3195 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3196 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3197 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3198 and PIC code generation.
3199 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3201 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3203 * config/nds32/nds32.md: Support TLS and PIC.
3204 * config/nds32/nds32.c: Support TLS and PIC.
3205 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3206 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3209 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3211 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3212 mode with E_ prefix.
3214 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3215 Chung-Ju Wu <jasonwucj@gmail.com>
3217 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3218 * config/nds32/nds32-md-auxiliary.c
3219 (symbolic_reference_mentioned_p): New.
3220 (nds32_legitimize_ict_address): New.
3221 (nds32_expand_ict_move): New.
3222 (nds32_indirect_call_referenced_p): New.
3223 (nds32_symbol_binds_local_p): Delete.
3224 (nds32_long_call_p): Modify.
3225 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3226 * config/nds32/nds32-protos.h
3227 (symbolic_reference_mentioned_p): Declare.
3228 (nds32_legitimize_ict_address): Declare.
3229 (nds32_expand_ict_move): Declare.
3230 (nds32_indirect_call_referenced_p): Declare.
3231 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3232 (nds32_relax_group): Use nds32_ict_const_p as condition.
3233 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3234 (nds32_asm_file_start): Output ict_model directive in asm code.
3235 (nds32_legitimate_address_p): Consider indirect call.
3236 (nds32_print_operand): Consider indirect call.
3237 (nds32_print_operand_address): Consider indirect call.
3238 (nds32_insert_attributes): Handle "indirect_call" attribute.
3239 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3240 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3241 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3242 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3243 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3244 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3245 (TARGET_ICT_MODEL_SMALL): Define.
3246 (TARGET_ICT_MODEL_LARGE): Define.
3247 * config/nds32/nds32.md (movsi): Consider ict model.
3248 (call, call_value): Consider ict model.
3249 (sibcall, sibcall_value): Consider ict model.
3250 * config/nds32/nds32.opt (mict-model): New option.
3251 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3254 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3255 Monk Chiang <sh.chiang04@gmail.com>
3256 Jim Wilson <jimw@sifive.com>
3258 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3259 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3260 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3261 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3262 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3263 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3264 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3265 compute save_libcall_adjustment properly.
3266 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3267 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3268 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3269 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3270 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3271 (ABI_SPEC): Handle mabi=ilp32e.
3272 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3273 (RVE): Add RVE mask.
3274 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3275 <-march>: Add rv32e as an example.
3277 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3280 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3281 (intra_create_variable_infos): Handle C++ constructors.
3283 2018-05-18 Martin Liska <mliska@suse.cz>
3285 * passes.def: Remove a redundant pass.
3287 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3290 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3292 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3294 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3296 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3298 (armv2, armv2a, armv3, armv3m): Delete architectures.
3299 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3300 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3301 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3303 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3304 (*mulsidi3adddi): Likewise.
3305 (mulsidi3): Likewise.
3306 (*mulsidi3_nov6): Likewise.
3307 (umulsidi3): Likewise.
3308 (umulsidi3_nov6): Likewise.
3309 (umaddsidi4): Likewise.
3310 (*umulsidi3adddi): Likewise.
3311 (smulsi3_highpart): Likewise.
3312 (*smulsi3_highpart_nov6): Likewise.
3313 (umulsi3_highpart): Likewise.
3314 (*umulsi3_highpart_nov6): Likewise.
3315 * config/arm/arm.h (arm_arch3m): Delete.
3316 * config/arm/arm.c (arm_arch3m): Delete.
3317 (arm_option_override_internal): Update armv3-related comment.
3318 (arm_configure_build_target): Delete use of isa_bit_mode32.
3319 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3320 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3321 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3323 (mulusa3): Likewise.
3324 * config/arm/arm-protos.h (arm_arch3m): Delete.
3325 * config/arm/arm-tables.opt: Regenerate.
3326 * config/arm/arm-tune.md: Likewise.
3327 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3328 deleted architectures.
3330 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3332 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3333 (armv5t, armv5te): New features.
3334 (ARMv5, ARMv5e): Delete fgroups.
3335 (ARMv5t, ARMv5te): Adjust for above changes.
3337 (armv5, armv5e): Delete arches.
3338 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3340 (*call_reg_arm): Likewise.
3341 (*call_value_reg_armv5): Likewise.
3342 (*call_value_reg_arm): Likewise.
3343 (*call_symbol): Likewise.
3344 (*call_value_symbol): Likewise.
3345 (*sibcall_insn): Likewise.
3346 (*sibcall_value_insn): Likewise.
3348 (prefetch): Likewise.
3349 (define_split and define_peephole2 dependent on arm_arch5):
3351 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3353 (TARGET_ARM_QBIT): Likewise.
3354 (TARGET_DSP_MULTIPLY): Likewise.
3355 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3356 (arm_arch5, arm_arch5e): Delete.
3357 (arm_arch5t, arm_arch5te): Declare.
3358 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3359 (arm_arch5t): Declare.
3360 (arm_option_reconfigure_globals): Update for the above.
3361 (arm_options_perform_arch_sanity_checks): Update comment, replace
3362 use of arm_arch5 with arm_arch5t.
3363 (use_return_insn): Likewise.
3364 (arm_emit_call_insn): Likewise.
3365 (output_return_instruction): Likewise.
3366 (arm_final_prescan_insn): Likewise.
3367 (arm_coproc_builtin_available): Likewise.
3368 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3369 arm_arch5e with arm_arch5t and arm_arch5te.
3370 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3371 (arm_arch5t, arm_arch5te): Declare.
3372 * config/arm/arm-tables.opt: Regenerate.
3373 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3374 * config/arm/t-multilib: Likewise.
3375 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3376 instead of arm_arch5.
3377 (*call_reg_thumb1): Likewise.
3378 (*call_value_reg_thumb1_v5): Likewise.
3379 (*call_value_reg_thumb1): Likewise.
3380 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3382 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3384 2018-05-18 Martin Liska <mliska@suse.cz>
3386 PR gcov-profile/84846
3387 * doc/gcov.texi: Document -t option of gcov tool.
3389 2018-05-18 Martin Liska <mliska@suse.cz>
3391 PR gcov-profile/84846
3392 * gcov.c (print_usage): Add new -t option.
3393 (process_args): Handle the option.
3394 (generate_results): Use stdout as output when requested by
3397 2018-05-18 Martin Liska <mliska@suse.cz>
3399 PR gcov-profile/84846
3400 * coverage.c (coverage_init): Write PWD to .gcno file.
3401 * doc/gcov.texi: Document how working directory is printed.
3402 * gcov-dump.c (dump_gcov_file): Print PWD.
3403 * gcov.c (output_intermediate_file): Likewise.
3404 (read_graph_file): Read PWD string.
3405 (output_lines): Print PWD.
3407 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3410 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3411 for retval and return false if all args to phi are zero.
3413 2018-05-18 Richard Biener <rguenther@suse.de>
3415 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3417 (evrp_dom_walker::before_dom_children): Call it.
3419 2018-05-18 Richard Biener <rguenther@suse.de>
3421 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3422 results when processing array refs with variable index.
3424 2018-05-18 Toon Moene <toon@moene.org>
3426 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3427 directly after that of -floop-interchange. Indicate that both
3428 options are enabled by default when specifying -O3.
3430 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3432 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3433 iterator. Delete separate integer-mode vec_set<mode> expander.
3434 (aarch64_simd_vec_setv2di): Delete.
3435 (vec_setv2di): Delete.
3436 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3437 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3438 the "w, r" alternative.
3440 2018-05-18 Martin Liska <mliska@suse.cz>
3442 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3443 * tree-pass.h (make_pass_lower_switch_O0): New function.
3444 * tree-switch-conversion.c (node_has_low_bound): Remove.
3445 (node_has_high_bound): Likewise.
3446 (node_is_bounded): Likewise.
3447 (class pass_lower_switch): Make it a template type and create
3449 (pass_lower_switch::execute): Add template argument.
3450 (make_pass_lower_switch): New function.
3451 (make_pass_lower_switch_O0): New function.
3452 (do_jump_if_equal): Remove.
3453 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3454 all the hard work to tree optimization passes.
3456 2018-05-18 Martin Liska <mliska@suse.cz>
3458 * dbgcnt.c (limit_low): Renamed from limit.
3459 (limit_high): New variable.
3460 (dbg_cnt_is_enabled): Check for upper limit.
3461 (dbg_cnt): Adjust dumping.
3462 (dbg_cnt_set_limit_by_index): Add new argument for high
3464 (dbg_cnt_set_limit_by_name): Likewise.
3465 (dbg_cnt_process_single_pair): Parse new format.
3466 (dbg_cnt_process_opt): Use strtok.
3467 (dbg_cnt_list_all_counters): Remove 'value' and add
3469 * doc/invoke.texi: Document changes.
3471 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3473 * doc/sourcebuild.texi (scalar_all_fma): Document.
3474 * tree.def (FMA_EXPR): Delete.
3475 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3476 * internal-fn.c (ternary_direct): New macro.
3477 (expand_ternary_optab_fn): Likewise.
3478 (direct_ternary_optab_supported_p): Likewise.
3479 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3480 * builtins.c (fold_builtin_fma): Delete.
3481 (fold_builtin_3): Don't call it.
3482 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3483 * expr.c (expand_expr_real_2): Likewise.
3484 * fold-const.c (operand_equal_p): Likewise.
3485 (fold_ternary_loc): Likewise.
3486 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3487 * gimple.c (DEFTREECODE): Likewise.
3488 * gimplify.c (gimplify_expr): Likewise.
3489 * optabs-tree.c (optab_for_tree_code): Likewise.
3490 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3491 * tree-eh.c (operation_could_trap_p): Likewise.
3492 (stmt_could_throw_1_p): Likewise.
3493 * tree-inline.c (estimate_operator_cost): Likewise.
3494 * tree-pretty-print.c (dump_generic_node): Likewise.
3495 (op_code_prio): Likewise.
3496 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3497 * tree-ssa-operands.c (get_expr_operands): Likewise.
3498 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3499 * fold-const-call.h (fold_fma): Delete.
3500 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3501 CFN_FNMA and CFN_FNMS.
3503 * genmatch.c (combined_fn): New enum.
3504 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3505 (commutative_op): New function.
3506 (commutate): Use it. Handle more than 2 operands.
3507 (dt_operand::gen_gimple_expr): Use commutative_op.
3508 (parser::parse_expr): Allow :c to be used with non-binary
3509 operators if the commutative operand is known.
3510 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3511 CFN_FMS, CFN_FNMA and CFN_FNMS.
3512 (backprop::process_assign_use): Remove FMA_EXPR handling.
3513 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3514 (gen_hsa_fma): New function.
3515 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3516 IFN_FNMA and IFN_FNMS.
3517 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3518 * gimple-fold.h (follow_all_ssa_edges): Declare.
3519 * gimple-fold.c (follow_all_ssa_edges): New function.
3520 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3521 gimple_build interface and use follow_all_ssa_edges to fold the result.
3522 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3523 instead of checking for optabs directly.
3524 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3525 rather than FMA_EXPRs.
3526 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3527 call to IFN_FMA instead of an FMA_EXPR.
3529 2018-05-17 Jim Wilson <jimw@sifive.com>
3531 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3532 sign extended promoted subreg, and the range does not have the sign bit
3533 set, then do a sign extend.
3535 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3536 test, check for sign extended subreg and/or constant operands, and
3537 do a sign extend in that case.
3539 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3541 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3542 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3544 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3545 Change logics_shift_reg to logics_shift_imm.
3546 (thunderx2t99_fp_loadpair_basic): Delete.
3547 (thunderx2t99_fp_storepair_basic): Delete.
3548 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3549 (thunderx2t99_asimd_polynomial): Delete.
3550 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3551 and neon_fp_mul_d_scalar_q.
3552 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3553 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3554 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3555 (thunderx2t99_asimd_lut): Add missing tbl types.
3556 (thunderx2t99_asimd_ext): Delete.
3557 (thunderx2t99_asimd_load1_1_mult): Delete.
3558 (thunderx2t99_asimd_load1_2_mult): Delete.
3559 (thunderx2t99_asimd_load1_ldp): New.
3560 (thunderx2t99_asimd_load1): New.
3561 (thunderx2t99_asimd_load2): Add missing *load2* types.
3562 (thunderx2t99_asimd_load3): New.
3563 (thunderx2t99_asimd_load4): New.
3564 (thunderx2t99_asimd_store1_1_mult): Delete.
3565 (thunderx2t99_asimd_store1_2_mult): Delete.
3566 (thunderx2t99_asimd_store2_mult): Delete.
3567 (thunderx2t99_asimd_store2_onelane): Delete.
3568 (thunderx2t99_asimd_store_stp): New.
3569 (thunderx2t99_asimd_store1): New.
3570 (thunderx2t99_asimd_store2): New.
3571 (thunderx2t99_asimd_store3): New.
3572 (thunderx2t99_asimd_store4): New.
3574 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3576 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3577 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3579 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3580 Segher Boessenkool <segher@kernel.crashing.org>
3583 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3586 2018-05-17 Richard Biener <rguenther@suse.de>
3588 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3589 for pruning loop and prune defs feeding only already visited PHIs.
3591 2018-05-17 Richard Biener <rguenther@suse.de>
3593 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3595 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3596 Richard Biener <rguenther@suse.de>
3598 PR tree-optimization/85793
3599 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3600 for VMAT_ELEMENTWISE.
3602 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3604 * internal-fn.h (lookup_internal_fn): Declare
3605 * internal-fn.c (lookup_internal_fn): New function.
3606 * gimple.c (gimple_build_call_from_tree): Handle calls to
3608 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3609 internal function names.
3610 * tree-pretty-print.c (dump_generic_node): Likewise.
3611 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3613 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3615 * gimple-fold.h (gimple_build): Make the function forms take
3616 combined_fn rather than built_in_function.
3617 (gimple_simplify): Likewise.
3618 * gimple-match-head.c (gimple_simplify): Likewise.
3619 * gimple-fold.c (gimple_build): Likewise.
3620 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3621 rather than gimple_build_call_internal.
3622 (get_initial_defs_for_reduction): Likewise.
3623 (vect_create_epilog_for_reduction): Likewise.
3624 (vectorizable_live_operation): Likewise.
3626 2018-05-17 Martin Liska <mliska@suse.cz>
3628 * gimple-ssa-sprintf.c (format_directive): Do not use
3629 space in between 'G_' and '('.
3631 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3634 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3635 even if the mask is not all ones.
3638 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3640 (ix86_gimple_fold_builtin): Likewise.
3643 * config/i386/i386.c: Include tree-vector-builder.h.
3644 (ix86_vector_shift_count): New function.
3645 (ix86_fold_builtin): Fold shift builtins by scalar count.
3646 (ix86_gimple_fold_builtin): Likewise.
3648 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3649 _mm512_setzero): New intrinsics.
3651 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3652 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3654 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3655 code generation for cases where splatting a value is not useful.
3656 * simplify-rtx.c (simplify_ternary_operation): Simplify
3657 vec_merge across a vec_duplicate and a paradoxical subreg forming
3658 a vector mode to a vec_concat.
3660 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3662 * config.gcc: Support "goldmont-plus".
3663 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3665 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3666 PROCESSOR_GOLDMONT_PLUS.
3667 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3668 (processor_target_table): Add "goldmont-plus".
3669 (PTA_GOLDMONT_PLUS): Define.
3670 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3671 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3672 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3673 (fold_builtin_cpu): Add "goldmont-plus".
3674 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3675 (ix86_option_override_internal): Add "goldmont-plus".
3676 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3677 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3678 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3679 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3681 2018-05-17 Richard Biener <rguenther@suse.de>
3683 PR tree-optimization/85757
3684 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3685 remove defs that only feed that PHI from further processing.
3687 2018-05-16 Jim Wilson <jimw@sifive.com>
3689 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3691 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3693 2018-05-16 Mark Wielaard <mark@klomp.org>
3695 * dwarf2out.c (count_index_strings): New function.
3696 (output_indirect_strings): Call count_index_strings and generate
3697 header for dwarf_version >= 5.
3699 2018-05-16 Mark Wielaard <mark@klomp.org>
3701 * dwarf2out.c (dwarf_FORM): New function.
3702 (set_indirect_string): Use dwarf_FORM.
3703 (reset_indirect_string): Likewise.
3704 (size_of_die): Likewise.
3705 (value_format): Likewise.
3706 (output_die): Likewise.
3707 (add_skeleton_AT_string): Likewise.
3708 (output_macinfo_op): Likewise.
3709 (index_string): Likewise.
3710 (output_index_string_offset): Likewise.
3711 (output_index_string): Likewise.
3712 (count_index_strings): Likewise.
3714 2018-05-16 Carl Love <cel@us.ibm.com>
3716 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3717 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3719 2018-05-16 Martin Jambor <mjambor@suse.cz>
3721 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3722 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3724 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3726 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3727 (fnma<mode>4): Likewise.
3728 (fms<mode>4): Likewise.
3729 (fnms<mode>4): Likewise.
3730 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3731 (aarch64_fnma<mode>4): Likewise.
3732 (aarch64_fms<mode>4): Likewise.
3733 (aarch64_fnms<mode>4): Likewise.
3734 (aarch64_fnmadd<mode>4): Likewise.
3736 2018-05-16 Jason Merrill <jason@redhat.com>
3738 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3740 2018-05-16 Richard Biener <rguenther@suse.de>
3742 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3743 (dump_stmt_cost): Declare.
3744 (add_stmt_cost): Dump cost we add.
3745 (add_stmt_costs): New function.
3746 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3748 (vect_analyze_stmt): Adjust prototype.
3749 (vectorizable_condition): Likewise.
3750 (vectorizable_live_operation): Likewise.
3751 (vectorizable_reduction): Likewise.
3752 (vectorizable_induction): Likewise.
3753 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3754 cost vector to pass to vectorizable_ and record afterwards.
3755 (vect_model_reduction_cost): Take cost vector argument and adjust.
3756 (vect_model_induction_cost): Likewise.
3757 (vectorizable_reduction): Likewise.
3758 (vectorizable_induction): Likewise.
3759 (vectorizable_live_operation): Likewise.
3760 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3761 SLP_TREE_NUMBER_OF_VEC_STMTS.
3762 (vect_analyze_slp_cost_1): Remove.
3763 (vect_analyze_slp_cost): Likewise.
3764 (vect_slp_analyze_node_operations): Take visited args and
3765 a target cost vector. Avoid processing already visited stmt sets.
3766 (vect_slp_analyze_operations): Use a local cost vector to gather
3767 costs and register those of non-discarded instances.
3768 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3769 (vect_schedule_slp_instance): Remove copying of
3770 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3772 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3773 adding cost. Record cost entry location.
3774 (vect_prologue_cost_for_slp_op): Function to compute cost of
3775 a constant or invariant generated for SLP vect in the prologue,
3776 split out from vect_analyze_slp_cost_1.
3777 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3778 (vect_model_promotion_demotion_cost): Likewise.
3779 (vect_model_store_cost): Likewise, make static.
3780 (vect_model_load_cost): Likewise.
3781 (vectorizable_bswap): Add cost vector arg and adjust.
3782 (vectorizable_call): Likewise.
3783 (vectorizable_simd_clone_call): Likewise.
3784 (vectorizable_conversion): Likewise.
3785 (vectorizable_assignment): Likewise.
3786 (vectorizable_shift): Likewise.
3787 (vectorizable_operation): Likewise.
3788 (vectorizable_store): Likewise.
3789 (vectorizable_load): Likewise.
3790 (vectorizable_condition): Likewise.
3791 (vectorizable_comparison): Likewise.
3792 (can_vectorize_live_stmts): Likewise.
3793 (vect_analyze_stmt): Likewise.
3794 (vect_transform_stmt): Adjust calls to vectorizable_*.
3795 * tree-vectorizer.c: Include gimple-pretty-print.h.
3796 (dump_stmt_cost): New function.
3798 2018-05-16 Richard Biener <rguenther@suse.de>
3800 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3801 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3802 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3803 (check_name): New callback.
3804 (dse_classify_store): Track cycles via a visited bitmap of PHI
3805 defs and simplify handling of in-loop and across loop dead stores
3806 and properly fail for loop-variant refs. Handle byte-tracking with
3807 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3810 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3812 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3813 (vect_get_mask_type_for_stmt): Likewise.
3814 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3816 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3817 to determine the statement's vector type and the vector type that
3818 should be used for calculating nunits. Deal with cases in which
3819 the type has to be deferred.
3820 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3821 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3822 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3823 (vect_determine_vf_for_stmt): New functions, split out from...
3824 (vect_determine_vectorization_factor): ...here.
3825 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3826 (vect_get_mask_type_for_stmt): New functions, split out from
3827 vect_determine_vectorization_factor.
3829 2018-05-16 Richard Biener <rguenther@suse.de>
3831 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3832 verify the [VEC_]COND_EXPR embedded comparison.
3834 2018-05-15 Martin Sebor <msebor@redhat.com>
3836 PR tree-optimization/85753
3837 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3838 RECORD_TYPE in addition to ARRAY_TYPE.
3840 2018-05-15 Martin Sebor <msebor@redhat.com>
3843 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3845 2018-05-15 Richard Biener <rguenther@suse.de>
3847 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3848 add by_clobber_p one. Change algorithm to collect all defs
3849 representing uses we need to walk and try reducing them to
3850 a single one before failing.
3851 (dse_dom_walker::dse_optimize_stmt): Adjust.
3853 2018-05-13 Mark Wielaard <mark@klomp.org>
3855 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3856 (size_of_loc_descr): Likewise.
3857 (output_loc_operands): Likewise.
3858 (output_loc_operands_raw): Likewise.
3859 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3860 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3861 (hash_loc_operands): Likewise.
3862 (compare_loc_operands): Likewise.
3864 2018-05-14 Mark Wielaard <mark@klomp.org>
3866 * dwarf2out.c (count_index_addrs): New function.
3867 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3869 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3871 PR tree-optimization/83648
3872 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3873 return value as malloc candidate.
3875 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3878 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3879 param as true in call to suggest_attribute.
3881 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3883 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3884 -mreadonly-in-sdata.
3886 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3888 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3890 (aarch64_crypto_aesd_fused): Likewise.
3892 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3894 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3895 (movsi_aarch64): Likewise.
3896 (load_pairsi): Likewise.
3897 (load_pairdi): Likewise.
3898 (store_pairsi): Likewise.
3899 (store_pairdi): Likewise.
3900 (load_pairsf): Likewise.
3901 (load_pairdf): Likewise.
3902 (store_pairsf): Likewise.
3903 (store_pairdf): Likewise.
3904 (zero_extend): Likewise.
3905 (trunc): Swap alternatives.
3906 (fcvt_target): Add '?' to prefer w over r.
3908 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3911 * config/i386/i386.md: Disallow non-commutative arithmetics in
3912 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3913 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3914 in the peephole2 before it.
3916 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3918 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3919 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3920 (ix86_handle_option): Handle -mcldemote.
3921 * config.gcc: New header.
3922 * config/i386/cldemoteintrin.h: New file.
3923 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3924 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3926 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3927 OPTION_MASK_ISA_CLDEMOTE.
3928 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3929 (ix86_valid_target_attribute_inner_p): Ditto.
3930 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3931 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3932 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3933 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3934 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3936 * config/i386/i386.opt: Add -mcldemote.
3937 * config/i386/x86intrin.h: New header.
3938 * doc/invoke.texi: Add -mcldemote.
3940 2018-05-14 Richard Biener <rguenther@suse.de>
3942 * doc/match-and-simplify.texi: Adjust :s documentation.
3944 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3946 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3947 intended memcpy size.
3948 (REORDER_45): Likewise.
3950 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3952 * sort.cc: New file.
3953 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3954 * vec.c (qsort_chk): Use gcc_qsort.
3955 * Makefile.in (OBJS-libcommon): Add sort.o.
3956 (build/sort.o): New target. Use it...
3957 (BUILD_RTL): ... here, and...
3958 (build/gencfn-macros): ... here, and...
3959 (build/genmatch): ... here.
3961 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3962 Chung-Ju Wu <jasonwucj@gmail.com>
3964 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3965 * config/nds32/nds32-graywolf.md: New file.
3966 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3967 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3969 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3970 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3971 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3972 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3973 * config/nds32/pipelines.md: Include n15 settings.
3975 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3976 Chung-Ju Wu <jasonwucj@gmail.com>
3978 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3979 * config/nds32/nds32-n13.md: New file.
3980 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3981 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3983 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3984 * config/nds32/nds32.md (pipeline_model): Add n13.
3985 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3986 * config/nds32/pipelines.md: Include n13 settings.
3988 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3989 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
3992 * config/nds32/nds32-n10.md: New file.
3993 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
3994 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
3996 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
3997 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
3998 * config/nds32/nds32.md (pipeline_model): Add n10.
3999 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4000 * config/nds32/pipelines.md: Include n10 settings.
4002 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4003 Kito Cheng <kito.cheng@gmail.com>
4004 Chung-Ju Wu <jasonwucj@gmail.com>
4006 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4007 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4008 Add enum values for DSP extension instructions.
4009 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4011 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4012 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4014 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4015 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4016 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4017 * config/nds32/nds32-intrinsic.md: Likewise.
4018 * config/nds32/nds32_intrinsic.h: Likewise.
4019 * config/nds32/nds32-md-auxiliary.c: Likewise.
4020 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4021 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4022 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4023 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4024 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4025 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4026 TYPE_DMAC in switch statement.
4027 * config/nds32/nds32.c: New checking and implementation for DSP
4028 extension instructions.
4029 * config/nds32/nds32.h: Likewise.
4030 * config/nds32/nds32.md: Likewise.
4031 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4032 * config/nds32/predicates.md: Implement new predicates for DSP
4035 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4037 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4038 Reformat alternatives and attributes so it is easier to identify
4039 which constraints/attributes go with which instruction.
4040 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4041 (mov<mode>_softfloat32, FMOVE64): Likewise.
4042 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4043 (mov<mode>_softfloat64, FMOVE64): Likewise.
4045 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4047 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4049 (Basic PowerPC Built-in Functions): The new name of the
4050 subsection previously known as "PowerPC Built-in Functions".
4051 (Basic PowerPC Built-in Functions Available on all Configurations):
4053 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4054 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4055 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4056 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4058 2018-05-11 Martin Jambor <mjambor@suse.cz>
4061 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4064 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4067 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4069 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4071 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4072 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4073 (ix86_handle_option): Handle -mwaitpkg.
4074 * config.gcc: New header.
4075 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4076 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4077 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4079 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4080 OPTION_MASK_ISA_WAITPKG.
4081 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4082 (ix86_option_override_internal): Add PTA_WAITPKG.
4083 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4084 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4085 IX86_BUILTIN_TPAUSE.
4086 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4087 __builtin_ia32_umwait and __builtin_ia32_tpause.
4088 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4089 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4090 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4091 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4092 UNSPECV_TPAUSE): New.
4093 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4094 * config/i386/i386.opt: Add -mwaitpkg.
4095 * config/i386/waitpkgintrin.h: New file.
4096 * config/i386/x86intrin.h: New header.
4097 * doc/invoke.texi: Add -mwaitpkg.
4099 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4102 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4104 (cortex-m0): Use armv6s-m isa.
4105 (cortex-m0plus): Likewise.
4106 (cortex-m1): Likewise.
4107 (cortex-m0.small-multiply): Likewise.
4108 (cortex-m0plus.small-multiply): Likewise.
4109 (cortex-m1.small-multiply): Likewise.
4111 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4112 Jakub Jelinek <jakub@redhat.com>
4114 PR tree-optimization/85692
4115 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4116 source permute as well.
4118 2018-05-11 Martin Liska <mliska@suse.cz>
4121 * doc/extend.texi: Document LLVM style format for no_sanitize
4124 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4126 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4127 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4128 (mode_supports_vsx_dform_quad): Likewise.
4129 (mode_supports_vmx_dform): Move these functions to be next to the
4130 other mode_supports functions.
4131 (mode_supports_dq_form): Likewise.
4132 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4133 mode_supports_dq_form.
4134 (reg_offset_addressing_ok_p): Likewise.
4135 (offsettable_ok_by_alignment): Likewise.
4136 (rs6000_legitimate_offset_address_p): Likewise.
4137 (legitimate_lo_sum_address_p): Likewise.
4138 (rs6000_legitimize_address): Likewise.
4139 (rs6000_legitimize_reload_address): Likewise.
4140 (rs6000_secondary_reload_inner): Likewise.
4141 (rs6000_preferred_reload_class): Likewise.
4142 (rs6000_output_move_128bit): Likewise.
4144 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4146 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4147 Generate SImode target register for null target.
4148 <case IX86_BUILTIN_XGETBV>: Ditto.
4149 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4150 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4152 2018-05-10 Carl Love <cel@us.ibm.com>
4154 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4155 dcbtt and dcbtstt if operands[2] is 0.
4157 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4160 * config/i386/sse.md (usadv64qi): New expander.
4162 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4164 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4165 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4166 -maltivec=be support.
4167 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4168 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4169 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4170 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4171 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4172 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4173 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4174 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4175 altivec_vsumsws): Adjust.
4176 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4177 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4178 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4180 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4181 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4182 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4183 (altivec_lve<VI_char>x): Delete expand.
4184 (*altivec_lve<VI_char>x_internal): Rename to...
4185 (altivec_lve<VI_char>x): ... this.
4186 (altivec_lvxl_<mode>): Delete expand.
4187 (*altivec_lvxl_<mode>_internal): Rename to ...
4188 (altivec_lvxl_<mode>): ... this.
4189 (altivec_stvxl_<mode>): Delete expand.
4190 (*altivec_stvxl_<mode>_internal): Rename to ...
4191 (altivec_stvxl_<mode>): ... this.
4192 (altivec_stve<VI_char>x): Delete expand.
4193 (*altivec_stve<VI_char>x_internal): Rename to ...
4194 (altivec_stve<VI_char>x): ... this.
4195 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4196 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4197 reduc_plus_scal_<mode>): Adjust.
4198 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4200 (rs6000_cpu_cpp_builtins): Adjust.
4201 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4202 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4203 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4204 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4205 -maltivec=be support.
4206 (rs6000_split_vec_extract_var): Adjust.
4207 (rs6000_split_v4si_init): Adjust.
4208 (swap_selector_for_mode): Delete.
4209 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4210 altivec_expand_stvex_be): Delete.
4211 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4212 -maltivec=be support.
4213 (rs6000_gimple_fold_builtin): Ditto.
4214 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4216 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4217 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4218 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4219 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4220 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4221 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4222 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4223 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4224 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4225 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4226 anonymous split): Adjust.
4227 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4228 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4230 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4232 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4233 when --with-gxx-include-dir is also specified.
4234 * configure: Regenerate.
4236 2018-05-09 Jim Wilson <jimw@sifive.com>
4239 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4240 * config/riscv/t-withmultilib: New.
4241 * config/riscv/withmultilib.h: New.
4242 * doc/install.texi: Document RISC-V --with-multilib-list support.
4244 2018-05-09 Richard Biener <rguenther@suse.de>
4246 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4248 (vect_bb_vectorization_profitable_p): Adjust. Compute
4249 actual scalar cost using the cost vector and the add_stmt_cost
4252 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4254 PR rtl-optimization/85645
4255 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4256 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4258 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4260 PR rtl-optimization/85645
4261 * shrink-wrap.c (spread_components): Return a boolean saying if
4262 anything was changed.
4263 (try_shrink_wrapping_separate): Iterate spread_components until
4264 nothing changes anymore.
4266 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4268 PR rtl-optimization/85645
4269 * regrename.c (build_def_use): Also kill the chains that include the
4270 destination of a REG_CFA_REGISTER note.
4272 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4274 PR rtl-optimization/85645
4275 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4276 insn that has a REG_CFA_REGISTER note.
4278 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4280 * cfgexpand.c (expand_clobber): New function.
4281 (expand_gimple_stmt_1): Use it.
4282 * tree-vect-stmts.c (vect_clobber_variable): New function,
4284 (vectorizable_simd_clone_call): ...here.
4285 (vectorizable_store): Emit a clobber either side of an
4286 IFN_STORE_LANES sequence.
4287 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4289 2018-05-09 Tom de Vries <tom@codesourcery.com>
4292 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4293 (define_insn "trap_if_false"): Add exit after trap.
4295 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4297 PR rtl-optimization/85638
4298 * bb-reorder.c: Include common/common-target.h.
4299 (create_forwarder_block): New function extracted from...
4300 (fix_up_crossing_landing_pad): ...here. Rename into...
4301 (dw2_fix_up_crossing_landing_pad): ...this.
4302 (sjlj_fix_up_crossing_landing_pad): New function.
4303 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4304 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4305 from both partitions and exit the loop after one iteration.
4307 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4310 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4312 (Basic PowerPC Built-in Functions): The new name of the
4313 subsection previously known as "PowerPC Built-in Functions".
4314 (Basic PowerPC Built-in Functions Available on all Configurations):
4316 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4318 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4320 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4322 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4325 2018-05-08 Jim Wilson <jimw@sifive.com>
4327 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4328 (LD_EMUL_SUFFIX): New.
4329 (LINK_SPEC): Use it.
4331 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4333 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4335 (Basic PowerPC Built-in Functions): The new name of the
4336 subsection previously known as "PowerPC Built-in Functions".
4337 (Basic PowerPC Built-in Functions Available on all Configurations):
4339 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4341 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4343 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4345 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4348 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4351 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4352 after cmpelim optimization.
4354 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4356 * config.gcc: Support "goldmont".
4357 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4358 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4360 * config/i386/i386.c (m_GOLDMONT): Define.
4361 (processor_target_table): Add "goldmont".
4362 (PTA_GOLDMONT): Define.
4363 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4364 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4365 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4366 (fold_builtin_cpu): Add "goldmont".
4367 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4368 (ix86_option_override_internal): Add "goldmont".
4369 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4370 (processor_type): Add PROCESSOR_GOLDMONT.
4371 * config/i386/i386.md: Add CPU "glm".
4372 * config/i386/glm.md: New file.
4373 * config/i386/x86-tune.def: Add m_GOLDMONT.
4374 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4376 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4379 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4381 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4382 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4383 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4386 * config/i386/i386.c (ix86_fold_builtin): Handle
4387 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4390 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4391 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4393 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4396 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4397 (check_arch): Likewise.
4398 (check_fpu): Return the result rather than printing it.
4399 (end arch): Fix operator precedence.
4400 (end cpu): Likewise.
4401 (END): Print the result from check_fpu.
4403 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4404 Alan Hayward <alan.hayward@arm.com>
4405 David Sherwood <david.sherwood@arm.com>
4407 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4408 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4409 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4410 (*fcmuo<mode>_and): New patterns.
4412 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4414 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4415 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4416 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4417 (cmp_op, sve_imm_con): New code attributes.
4418 (SVE_COND_INT_CMP, imm_con): Delete.
4419 (cmp_op): Remove above unspecs from int attribute.
4420 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4422 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4423 comparison-specific unspecs.
4424 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4425 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4426 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4427 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4428 (*vec_fcm<cmp_op><mode>): Rename to...
4429 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4430 (*vec_fcmuo<mode>): Rename to...
4431 (*fcmuo<mode>): ...this and adjust likewise.
4432 (*pred_fcm<cmp_op><mode>): New pattern.
4433 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4434 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4436 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4438 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4439 (aarch64_emit_sve_predicated_cond): New function.
4440 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4441 (aarch64_emit_unspec_cond_or): Replace with...
4442 (aarch64_emit_sve_or_conds): ...this new function. Use
4443 aarch64_emit_sve_ptrue_op for the individual comparisons and
4444 aarch64_emit_binop to OR them together.
4445 (aarch64_emit_inverted_unspec_cond): Replace with...
4446 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4447 aarch64_emit_sve_ptrue_op for the comparison and
4448 aarch64_emit_unop to invert the result.
4449 (aarch64_expand_sve_vec_cmp_float): Update after the above
4450 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4452 2018-05-07 Nathan Sidwell <nathan@acm.org>
4454 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4455 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4456 (Backwards Compatibility): Likewise.
4458 2018-05-07 Luis Machado <luis.machado@linaro.org>
4462 2018-05-07 Luis Machado <luis.machado@linaro.org>
4464 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4465 <prefetch_dynamic_strides>: New const bool field.
4466 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4467 prefetch_dynamic_strides.
4468 (exynosm1_prefetch_tune): Likewise.
4469 (thunderxt88_prefetch_tune): Likewise.
4470 (thunderx_prefetch_tune): Likewise.
4471 (thunderx2t99_prefetch_tune): Likewise.
4472 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4474 (aarch64_override_options_internal): Update to set
4475 PARAM_PREFETCH_DYNAMIC_STRIDES.
4476 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4477 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4478 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4479 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4480 prefetch-dynamic-strides setting.
4482 2018-05-07 Luis Machado <luis.machado@linaro.org>
4484 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4485 <minimum_stride>: New const int field.
4486 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4487 minimum_stride field.
4488 (exynosm1_prefetch_tune): Likewise.
4489 (thunderxt88_prefetch_tune): Likewise.
4490 (thunderx_prefetch_tune): Likewise.
4491 (thunderx2t99_prefetch_tune): Likewise.
4492 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4493 (aarch64_override_options_internal): Update to set
4494 PARAM_PREFETCH_MINIMUM_STRIDE.
4495 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4496 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4497 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4498 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4499 stride is constant and is below the minimum stride threshold.
4501 2018-05-07 Luis Machado <luis.machado@linaro.org>
4503 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4506 2018-05-07 Luis Machado <luis.machado@linaro.org>
4508 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4509 <prefetch_dynamic_strides>: New const bool field.
4510 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4511 prefetch_dynamic_strides.
4512 (exynosm1_prefetch_tune): Likewise.
4513 (thunderxt88_prefetch_tune): Likewise.
4514 (thunderx_prefetch_tune): Likewise.
4515 (thunderx2t99_prefetch_tune): Likewise.
4516 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4518 (aarch64_override_options_internal): Update to set
4519 PARAM_PREFETCH_DYNAMIC_STRIDES.
4520 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4521 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4522 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4523 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4524 prefetch-dynamic-strides setting.
4526 2018-05-07 Luis Machado <luis.machado@linaro.org>
4528 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4529 <minimum_stride>: New const int field.
4530 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4531 minimum_stride field.
4532 (exynosm1_prefetch_tune): Likewise.
4533 (thunderxt88_prefetch_tune): Likewise.
4534 (thunderx_prefetch_tune): Likewise.
4535 (thunderx2t99_prefetch_tune): Likewise.
4536 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4537 (aarch64_override_options_internal): Update to set
4538 PARAM_PREFETCH_MINIMUM_STRIDE.
4539 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4540 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4541 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4542 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4543 stride is constant and is below the minimum stride threshold.
4545 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4548 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4549 the type is addressable. Don't force op into register if it has
4552 2018-05-05 Roland McGrath <mcgrathr@google.com>
4555 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4556 any section for which we don't know a specific type it should have,
4557 regardless of name. Previously this was done only for the exact
4558 names ".init_array", ".fini_array", and ".preinit_array".
4559 (default_elf_asm_named_section): Add comment about
4560 relationship with default_section_type_flags and SECTION_NOTYPE.
4561 (get_section): Don't consider it a type conflict if one side has
4562 SECTION_NOTYPE and the other doesn't, as long as neither has the
4563 SECTION_BSS et al used in the default_section_type_flags logic.
4565 2018-05-05 Tom de Vries <tom@codesourcery.com>
4568 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4569 (workaround_barsyncs): New function.
4570 (nvptx_reorg): Use workaround_barsyncs.
4571 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4572 (define_expand "nvptx_membar_cta"): New define_expand.
4573 (define_insn "*nvptx_membar_cta"): New insn.
4575 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4577 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4578 To improve optimization opportunities.
4579 * builtin-types.def: The new needed builtin types for the above.
4581 2018-05-04 Richard Biener <rguenther@suse.de>
4583 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4584 * gimple-ssa-store-merging.c
4585 (imm_store_chain_info::output_merged_store): Remove redundant create,
4586 release split_store vector contents on failure.
4587 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4588 scalar stmt vector on cache hit.
4590 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4592 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4594 * config.gcc (powerpc-xilinx-eabi*): Remove.
4595 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4597 (fusion_addis_mem_combo_load): Ditto.
4598 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4600 (rs6000_cpu_cpp_builtins): Ditto.
4601 * config/rs6000/rs6000-linux.c
4602 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4603 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4604 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4606 (rs6000_setup_reg_addr_masks): Ditto.
4607 (rs6000_init_hard_regno_mode_ok): Ditto.
4608 (rs6000_option_override_internal): Ditto.
4609 (legitimate_lo_sum_address_p): Ditto.
4610 (rs6000_legitimize_address): Ditto.
4611 (rs6000_legitimize_reload_address): Ditto.
4612 (rs6000_legitimate_address_p): Ditto.
4613 (abi_v4_pass_in_fpr): Ditto.
4614 (setup_incoming_varargs): Ditto.
4615 (rs6000_gimplify_va_arg): Ditto.
4616 (rs6000_split_multireg_move): Ditto.
4617 (rs6000_savres_strategy): Ditto.
4618 (rs6000_emit_prologue_components): Ditto.
4619 (rs6000_emit_epilogue_components): Ditto.
4620 (rs6000_emit_prologue): Ditto.
4621 (rs6000_emit_epilogue): Ditto.
4622 (rs6000_elf_file_end): Ditto.
4623 (rs6000_function_value): Ditto.
4624 (rs6000_libcall_value): Ditto.
4625 * config/rs6000/rs6000.h: Ditto.
4626 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4627 (TARGET_MINMAX): ... this. New.
4628 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4629 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4630 (*movsi_internal1_single): Delete.
4631 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4632 mfpu=, mxilinx-fpu): Delete.
4633 * config/rs6000/singlefp.h: Delete.
4634 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4635 * config/rs6000/t-rs6000: Ditto.
4636 * config/rs6000/t-xilinx: Delete.
4637 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4638 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4639 (VStype_simple): Delete.
4640 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4641 * config/rs6000/xfpu.h: Delete.
4642 * config/rs6000/xfpu.md: Delete.
4643 * config/rs6000/xilinx.h: Delete.
4644 * config/rs6000/xilinx.opt: Delete.
4645 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4646 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4648 2018-05-04 Tom de Vries <tom@codesourcery.com>
4651 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4654 2018-05-04 Richard Biener <rguenther@suse.de>
4657 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4658 (expand_complex_div_wide): Likewise.
4659 (expand_complex_operations_1): Likewise.
4660 (expand_complex_libcall): Preserve EH info of the original stmt.
4661 (tree_lower_complex): Handle removed blocks.
4662 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4663 on complex multiplication and division libcall builtins.
4665 2018-05-04 Richard Biener <rguenther@suse.de>
4668 * fold-const.c (negate_expr_p): Restrict negation of operand
4669 zero of a division to when we know that can happen without
4671 (fold_negate_expr_1): Likewise.
4673 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4676 * real.h (real_nextafter): Declare.
4677 * real.c (real_nextafter): New function.
4678 * fold-const-call.c (fold_const_nextafter): New function.
4679 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4680 CASE_CFN_NEXTTOWARD.
4681 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4682 even when arg1_mode is different from arg0_mode.
4684 2018-05-03 Nathan Sidwell <nathan@acm.org>
4686 * doc/extend.texi (Deprecated Features): Remove
4688 (Backwards Compatibility): Likewise.
4689 * doc/invoke.texi (C++ Language Options): Likewise.
4690 (C++ Dialect Options): Likewise.
4692 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4695 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4696 _mm512_mask_mullox_epi64): New intrinsics.
4698 2018-05-03 Tom de Vries <tom@codesourcery.com>
4701 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4702 dump files): Add offload-tree.
4704 2018-05-03 Richard Biener <rguenther@suse.de>
4706 PR tree-optimization/85615
4707 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4708 to loops not nested in BBs loop father to avoid creating multi-entry
4711 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4713 PR tree-optimization/70291
4714 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4715 arguments. Change return type to tree. Emit libcall as a new
4716 statement rather than replacing existing one when inplace_p is true.
4717 (expand_complex_multiplication_components): New function.
4718 (expand_complex_multiplication): Expand floating-point complex
4719 multiplication using the above.
4720 (expand_complex_division): Rename inner_type parameter to type.
4721 Update expand_complex_libcall call-site.
4722 (expand_complex_operations_1): Update expand_complex_multiplication
4723 and expand_complex_division call-sites.
4725 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4728 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4729 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4730 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4731 the highest significant bit of the shift count mask is clear. In
4732 check whether and[sq]i3 is needed verify that all significant bits
4733 of the shift count other than the highest are set.
4735 2018-05-02 Tom de Vries <tom@codesourcery.com>
4738 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4739 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4740 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4741 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4742 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4743 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4744 __builtin_goacc_parlevel_size.
4746 2018-05-02 Richard Biener <rguenther@suse.de>
4748 PR tree-optimization/85597
4749 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4750 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4753 2018-05-02 Tom de Vries <tom@codesourcery.com>
4756 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4757 dump files): Add ltrans-tree.
4759 2018-05-02 Tom de Vries <tom@codesourcery.com>
4762 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4763 dump files): Add wpa-ipa.
4765 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4767 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4768 powerpc*-*-linux*paired* target.
4769 * config/rs6000/750cl.h: Delete.
4770 * config/rs6000/paired.h: Delete.
4771 * config/rs6000/paired.md: Delete.
4772 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4774 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4775 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4776 comment. Remove paired float support.
4777 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4778 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4780 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4781 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4783 * config/rs6000/rs6000.c: Remove paired float support.
4784 (paired_expand_vector_init, paired_expand_vector_move,
4785 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4786 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4787 paired_expand_builtin, paired_expand_predicate_builtin,
4788 paired_init_builtins): Delete.
4789 * config/rs6000/rs6000.h: Remove paired float support.
4790 * config/rs6000/rs6000.md: Remove paired float support.
4791 (move_from_CR_ov_bit): Delete.
4792 * config/rs6000/rs6000.opt (mpaired): Delete.
4793 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4794 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4796 2018-05-02 Richard Biener <rguenther@suse.de>
4799 * gimplify.c (gimplify_save_expr): When in SSA form allow
4800 SAVE_EXPRs to compute to SSA vars.
4802 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4805 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4806 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4807 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4808 clobber operands[2], instead use a new pseudo. Formatting fixes.
4810 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4812 PR tree-optimization/85586
4813 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4814 exit early for statements in the same group if the accesses are
4817 2018-05-02 Tom de Vries <tom@codesourcery.com>
4820 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4823 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4825 PR tree-optimization/85143
4826 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4828 2018-05-01 Tom de Vries <tom@codesourcery.com>
4831 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4832 not found" error message.
4834 2018-05-01 Tom de Vries <tom@codesourcery.com>
4837 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4838 * vec.c (test_ordered_remove_if): New function.
4839 (vec_c_tests): Call test_ordered_remove_if.
4840 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4841 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4842 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4843 VEC_ORDERED_REMOVE_IF.
4845 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4847 PR tree-optimization/82665
4848 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4849 pointer subtraction where arguments come from a memchr call.
4851 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4853 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4854 --push-state --as-needed and --pop-state instead of --as-needed and
4855 --no-as-needed if ld supports it.
4856 * configure: Regenerated.
4859 * doc/install.texi2html: Replace _002d with - and _002a with * in
4860 generated html files using sed.
4862 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4865 * gcc-rich-location.c (blank_line_before_p): New function.
4866 (use_new_line): New function.
4867 (gcc_rich_location::add_fixit_insert_formatted): New function.
4868 * gcc-rich-location.h
4869 (gcc_rich_location::add_fixit_insert_formatted): New function.
4871 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4873 * selftest.c (assert_streq): Rename "expected" and "actual" to
4874 "val1" and "val2". Extend NULL-handling to cover both inputs
4875 symmetrically, while still requiring both to be non-NULL for a pass.
4876 * selftest.h (assert_streq): Rename "expected" and "actual" to
4878 (ASSERT_EQ): Likewise.
4879 (ASSERT_EQ_AT): Likewise.
4880 (ASSERT_KNOWN_EQ): Likewise.
4881 (ASSERT_KNOWN_EQ_AT): Likewise.
4882 (ASSERT_NE): Likewise.
4883 (ASSERT_MAYBE_NE): Likewise.
4884 (ASSERT_MAYBE_NE_AT): Likewise.
4885 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4886 the assertion to pass.
4887 (ASSERT_STREQ_AT): Likewise.
4889 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4891 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4892 interaction with -pie.
4894 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4896 * selftest.h: Fix alphabetization of per-source-file selftest
4899 2018-04-30 Jason Merrill <jason@redhat.com>
4901 PR c++/61982 - dead stores to destroyed objects.
4902 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4905 2018-04-30 Jason Merrill <jason@redhat.com>
4907 * tree.c (build_clobber): New.
4908 * tree.h: Declare it.
4909 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4911 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4913 * diagnostic-show-locus.c (layout::layout): Update for
4914 location_get_source_line returning a char_span.
4915 (struct char_span): Move to input.h.
4916 (struct correction): Update for fields in char_span becoming
4918 (struct source_line): Update for location_get_source_line
4919 returning a char_span.
4920 (layout::print_line): Likewise.
4921 * edit-context.c (edited_file::print_content): Likewise.
4922 (edited_file::print_diff_hunk): Likewise.
4923 (edited_file::print_run_of_changed_lines): Likewise.
4924 (edited_file::get_num_lines): Likewise.
4925 (edited_line::edited_line): Likewise.
4926 * final.c (asm_show_source): Likewise.
4927 * input.c (location_get_source_line): Convert return type
4928 from const char * to char_span, losing the final "line_len"
4930 (dump_location_info): Update for the above.
4931 (get_substring_ranges_for_loc): Likewise. Use a char_span
4932 when handling the literal within the line.
4933 (test_reading_source_line): Update for location_get_source_line
4934 returning a char_span.
4935 * input.h (class char_span): Move here from
4936 diagnostic-show-locus.c, converting from a struct to a class.
4937 Make data members private.
4938 (char_span::operator bool): New.
4939 (char_span::length): New.
4940 (char_span::get_buffer): New.
4941 (char_span::operator[]): New.
4942 (char_span::subspan): Make const.
4943 (char_span::xstrdup): New.
4944 (location_get_source_line): Convert return type from const char *
4945 to char_span, losing the final "line_size" param.
4947 2018-04-30 Jan Hubicka <jh@suse.cz>
4949 * lto-wrapper.c (ltrans_priorities): New static var.
4950 (cmp_priority): New.
4951 (run_gcc): Read priorities and if doing parallel build order
4952 the Makefile by them.
4954 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4956 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4958 2018-04-30 Richard Biener <rguenther@suse.de>
4960 * tree-cfg.c (verify_address): Remove base argument, add
4961 flag whether to check TREE_ADDRESSABLE and do that.
4962 (verify_expr): Remove.
4963 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4964 (verify_gimple_assign_single): Likewise.
4965 (verify_gimple_switch): Likewise.
4966 (verify_expr_location_1): Dereference tp once. Add (disabled)
4967 piece from verify_expr.
4968 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4970 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4972 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4974 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4976 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4977 (small_data_pattern): Likewise.
4978 (arc_rewrite_small_data): Likewise.
4979 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4980 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4981 (get_symbol_alignment): New function.
4982 (legitimate_small_data_address_p): Likewise.
4983 (legitimate_scaled_address): Update, call
4984 legitimate_small_data_address_p.
4985 (output_sdata): New static variable.
4986 (arc_print_operand): Update how we handle small data operands.
4987 (arc_print_operand_address): Likewise.
4988 (arc_legitimate_address_p): Update, use
4989 legitimate_small_data_address_p.
4990 (arc_rewrite_small_data_p): Remove.
4991 (arc_rewrite_small_data_1): Likewise.
4992 (arc_rewrite_small_data): Likewise.
4993 (small_data_pattern): Likewise.
4994 (compact_sda_memory_operand): Update to use
4995 legitimate_small_data_address_p and get_symbol_alignment.
4996 (prepare_move_operands): Don't rewite sdata pattern.
4997 (prepare_extend_operands): Remove.
4998 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5000 (zero_extendqisi2): Likewise.
5001 (zero_extendhisi2): Likewise.
5002 (extendqihi2): Likewise.
5003 (extendqisi2): Likewise.
5004 (extendhisi2): Likewise.
5008 * config/arc/constraints.md (Usd): Change it to memory constraint.
5010 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5012 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5013 as source of std instructions.
5014 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5015 allow 6-bit constants as source for store instructions.
5016 (movdi_insn): Update instruction pattern to allow 6-bit constants
5017 as source for store instructions.
5019 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5021 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5023 2018-04-30 Nathan Sidwell <nathan@acm.org>
5024 Sandra Loosemore <sandra@codesourcery.com>
5026 * dumpfile.c (dump_open): Allow '-' for stdout.
5027 * doc/invoke.texi (Developer Options): Document dump filename
5028 determination early. Document stdin/stdout selection.
5030 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5032 Microblaze Target: PIC data text relative
5034 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5035 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5037 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5038 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5039 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5040 New addressing mode for data-text relative position indepenedent code.
5041 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5042 'ADDRESS_SYMBOLIC_TXT_REL'.
5043 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5044 (microblaze_legitimate_pic_operand): Exclude function calls from
5045 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5046 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5048 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5049 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5050 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5051 for 'address + offset'.
5052 (microblaze_expand_prologue): Add new function prologue call for
5054 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5055 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5056 table in case of TARGET_PIC_DATA_TEXT_REL.
5057 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5058 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5059 Add new macros 'UNSPEC_TEXT',
5060 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5061 + exclude function calls from 'UNSPEC_PLT' in case of data text
5063 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5064 new target hook for generating address diff vector tables in case of
5066 * doc/tm.texi : Regenerate.
5067 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5068 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5069 of addr diff vector generation.
5070 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5071 target hook definition.
5072 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5073 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5074 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5075 option for data text relative.
5077 2018-04-30 Richard Biener <rguenther@suse.de>
5079 * tree-chrec.h (evolution_function_is_constant_p): Remove
5081 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5083 2018-04-30 Richard Biener <rguenther@suse.de>
5086 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5088 2018-04-30 Richard Biener <rguenther@suse.de>
5090 PR tree-optimization/28364
5091 PR tree-optimization/85275
5092 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5093 copying first exit test.
5095 2018-04-28 Mark Wielaard <mark@klomp.org>
5097 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5099 (dwarf_AT): Handle DW_AT_addr_base.
5100 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5102 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5105 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5106 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5107 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5108 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5110 2018-04-28 Richard Biener <rguenther@suse.de>
5112 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5113 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5114 to reflect use. Only add interesting stmts.
5116 2018-04-27 Martin Jambor <mjambor@suse.cz>
5119 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5120 the jump function allows for passing through aggregate values.
5122 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5124 * input.h (in_system_header_at): Convert from macro to inline
5126 (from_macro_expansion_at): Likewise.
5127 (from_macro_definition_at): Likewise.
5129 2018-04-27 Jeff Law <law@redhat.com>
5131 * config.gcc: Mark tile* targets as deprecated/obsolete.
5133 2018-04-27 Richard Biener <rguenther@suse.de>
5135 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5138 2018-04-27 Richard Biener <rguenther@suse.de>
5140 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5142 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5144 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5145 with Yd constraint. Set "preferred_for_speed" attribute from
5146 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5148 (*movdi_internal): Ditto.
5149 (movti_interunit splitters): Remove
5150 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5151 (movdi_interunit splitters): Ditto.
5152 * config/i386/constraints.md (Ye): Remove.
5153 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5155 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5158 * config/aarch64/constraints.md (Usg): Limit to 31.
5161 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5163 PR tree-optimization/85529
5164 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5165 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5166 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5167 zero extension or masking of the MSB bit.
5168 (optimize_range_tests): Add FIRST_BB argument, pass it through
5169 to optimize_range_tests_var_bound.
5170 (maybe_optimize_range_tests, reassociate_bb): Adjust
5171 optimize_range_tests callers.
5173 2018-04-26 Richard Biener <rguenther@suse.de>
5174 Jakub Jelinek <jakub@redhat.com>
5176 * cgraph.h (symbol_table): Just declare debug method here.
5177 * symtab.c (symbol_table::debug): Define.
5179 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5181 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5183 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5185 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5186 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5187 (*movdi_internal): Substitute Yi and Yj constraint with x
5188 and Ym and Yn constraint with y constraint. Update "isa"
5189 attribute and set "preferred_for_speed" attribute from
5190 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5191 (*movsi_internal): Ditto.
5192 (*movdf_internal): Ditto.
5193 (*movsf_internal): Ditto.
5194 (*zero_extendsidi2): Ditto.
5195 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5196 (sse2_loadld): Ditto.
5197 (*vec_extract<ssevecmodelower>_0): Ditto.
5198 (*vec_extractv4si_0_zext_sse4): Ditto.
5199 (vec_concatv2di): Ditto.
5200 (*vec_dup<mode>): Ditto.
5201 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5202 * config/i386/constraints.md (Yi): Remove.
5207 2018-04-26 Nathan Sidwell <nathan@acm.org>
5209 * dumpfile.c (dump_open): New.
5210 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5211 (dump_finish): Detect stdio/stderr by value not name.
5213 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5215 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5217 2018-04-26 Tom de Vries <tom@codesourcery.com>
5220 * config/nvptx/nvptx.c (verify_neutering_jumps)
5221 (verify_neutering_labels): New function
5222 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5224 2018-04-26 Tom de Vries <tom@codesourcery.com>
5227 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5228 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5231 2018-04-26 Richard Biener <rguenther@suse.de>
5232 Tom de Vries <tom@codesourcery.com>
5235 * lto-streamer-out.c (output_function): Fixup loops if required to match
5236 discovery done in the reader.
5238 2018-04-26 Richard Biener <rguenther@suse.de>
5240 PR tree-optimization/85116
5241 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5242 have a loop exit from the single latch predecessor. Remove
5243 case of header with just condition.
5244 (ch_base::copy_headers): Exclude infinite loops from any
5246 (pass_ch::execute): Record exits.
5248 2018-04-26 Richard Biener <rguenther@suse.de>
5250 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5251 prologue cost vector and pass it to vect_get_load_cost.
5252 (vect_get_peeling_costs_all_drs): Likewise.
5253 (vect_peeling_hash_get_lowest_cost): Likewise.
5254 (vect_enhance_data_refs_alignment): Likewise.
5256 2018-04-26 Richard Biener <rguenther@suse.de>
5259 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5260 checking of integer<->pointer conversions.
5261 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5262 sign-/zero-extending pointer types.
5263 (expand_omp_for_static_chunk): Likewise.
5265 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5266 Jean Lee <xiaoyur347@gmail.com>
5268 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5269 (TARGET_ASAN_SHADOW_OFFSET): Define.
5270 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5271 true for -fsanitize=address.
5273 2018-04-25 Mark Wielaard <mark@klomp.org>
5275 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5278 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5280 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5281 than "alu", remove explicit "memory" and "imm_disp" attributes.
5282 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5285 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5286 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5287 gen_lowpart_no_emit.
5289 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5292 * config/i386/i386.c (ix86_expand_builtin): Change memory
5293 operand to XI, extend p0 to Pmode.
5294 * config/i386/i386.md: Change unspec volatile and operand
5295 1 mode to XI, change operand 0 mode to P.
5297 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5299 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5300 GET_MODE_MASK before any checking.
5301 (nds32_can_use_bset_p): Likewise.
5302 (nds32_can_use_btgl_p): Likewise.
5304 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5306 * config/nds32/nds32-doubleword.md: New define_split pattern for
5307 illegal register number.
5309 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5311 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5313 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5315 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5317 2018-04-25 Richard Biener <rguenther@suse.de>
5319 * lto-streamer.h (LTO_major_version): Bump to 8.
5321 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5323 * BASE-VER: Set to 9.0.0.
5325 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5327 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5328 in __abskf2 and __powikf2.
5330 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5333 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5334 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5335 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5336 Use the above on operand 2. Reindent.
5337 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5339 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5342 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5344 * config/i386/i386.opt (mcet): Removed.
5345 * doc/install.texi: Remove -mcet documentation.
5346 * doc/invoke.texi: Likewise.
5348 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5351 * doc/install.texi: Remove -mcet from bootstrap-cet.
5353 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5356 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5357 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5361 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5362 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5363 containing a CONST_VECTOR.
5365 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5367 * doc/install.texi: Update newlib dependency for nvptx.
5369 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5372 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5373 instead of INTVAL when shifting x left.
5375 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5377 PR tree-optimization/85478
5378 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5379 vect_grouped_store_supported for single element vectors.
5381 2018-04-24 Richard Biener <rguenther@suse.de>
5384 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5385 load cost increase to the case of non-constant step.
5387 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5390 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5391 destination if any_malformed_asm.
5393 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5396 * expr.c (store_field): In the bitfield case, if the value comes from
5397 a function call and is returned in registers by means of a PARALLEL,
5398 do not change the mode of the temporary unless BLKmode and VOIDmode.
5400 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5402 PR rtl-optimization/85423
5403 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5404 dependencies to debug insns when the previous insn is non-debug.
5406 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5408 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5409 enums into a single definition.
5410 (fls): Fix predicates and printing.
5413 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5415 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5416 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5417 and short u6 immediate.
5418 (check_if_valid_sleep_operand): Remove.
5419 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5421 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5423 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5424 flag_always_save_lp condition.
5425 * config/nds32/nds32.opt (malways-save-lp): New option.
5427 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5429 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5430 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5431 * config/nds32/nds32.h
5432 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5433 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5435 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5437 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5438 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5440 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5441 Chung-Ju Wu <jasonwucj@gmail.com>
5443 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5445 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5446 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5448 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5450 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5452 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5454 * config/nds32/nds32-protos.h (nds32_data_alignment,
5455 nds32_local_alignment): Declare.
5456 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5457 nds32_local_alignment): New functions.
5458 (TARGET_CONSTANT_ALIGNMENT): Define.
5459 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5461 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5463 * config/nds32/nds32.c
5464 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5465 (TARGET_MODES_TIEABLE_P): Likewise.
5467 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5469 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5472 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5473 Chung-Ju Wu <jasonwucj@gmail.com>
5475 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5476 for unaligned access.
5477 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5479 * config/nds32/nds32-intrinsic.md: Likewise.
5480 * config/nds32/nds32_intrinsic.h: Likewise.
5481 * config/nds32/nds32.h (nds32_builtins): Likewise.
5482 * config/nds32/nds32.opt (munaligned-access): New option.
5483 * config/nds32/nds32.c (nds32_asm_file_start): Display
5484 flag_unaligned_access status.
5486 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5488 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5489 -mno-relax is present.
5490 * config/riscv/linux.h (LINK_SPEC): Ditto.
5492 2018-04-20 Martin Sebor <msebor@redhat.com>
5495 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5497 (gimple_fold_builtin_stxcpy_chk): Same.
5498 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5500 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5503 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5504 __powikf2 when long double is IEEE 128-bit.
5506 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5508 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5509 step to make sure stack always aligned.
5511 2018-04-20 Carl Love <cel@us.ibm.com>
5514 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5515 size check for arg0.
5517 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5518 Tom de Vries <tom@codesourcery.com>
5521 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5522 Emit insns for calls too.
5523 (nvptx_find_par): Always look for worker-level predecessor insn.
5524 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5526 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5527 (nvptx_process_pars): Propagate frames for calls.
5529 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5532 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5534 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5535 (ix86_handle_option): Don't handle OPT_mibt.
5536 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5538 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5540 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5541 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5542 (ix86_target_macros): Define __CET__ with flag_cf_protection
5543 for -fcf-protection.
5544 * config/i386/i386.c (isa2_opts): Remove -mibt.
5545 * config/i386/i386.h (TARGET_IBT): Removed.
5546 (TARGET_IBT_P): Likewise.
5547 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5548 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5549 * config/i386/i386.opt (mcet): Update help message.
5552 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5553 -mcet as an alias for -mshstk.
5555 2018-04-20 Richard Biener <rguenther@suse.de>
5558 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5559 complexity by forcing a single use of the multiply operand.
5561 2018-04-20 Martin Jambor <mjambor@suse.cz>
5564 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5565 recursion dependency to only apply to non-clones.
5567 2018-04-20 Martin Jambor <mjambor@suse.cz>
5570 * ipa-cp.c (create_specialized_node): Check that clones of
5571 self-recursive edges exist during IPA-CP.
5573 2018-04-19 Toon Moene <toon@moene.org>
5575 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5578 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5580 PR tree-optimization/85467
5581 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5582 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5583 VECTOR_CST element to type.
5585 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5588 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5589 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5590 (builtin_longjmp): Likewise.
5591 (save_stack_nonlocal): New pattern.
5592 (restore_stack_nonlocal): Likewise.
5594 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5597 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5598 Replace ASM_OUTPUT_LABEL with fprintf.
5600 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5603 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5604 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5605 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5606 define __IBT__ and __SHSTK__ for -fcf-protection.
5607 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5609 (ix86_trampoline_init): Likewise.
5610 (x86_output_mi_thunk): Likewise.
5611 (ix86_notrack_prefixed_insn_p): Likewise.
5612 (ix86_option_override_internal): Don't disallow -fcf-protection.
5613 * config/i386/i386.md (rdssp<mode>): Also enable for
5615 (incssp<mode>): Likewise.
5616 (nop_endbr): Likewise.
5617 * config/i386/i386.opt (mcet): Change help message to built-in
5621 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5622 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5623 enable CET built-in functions.
5625 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5627 * common/config/i386/i386-common.c
5628 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5629 OPTION_MASK_ISA_MOVDIRI_UNSET,
5630 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5631 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5632 * config.gcc (movdirintrin.h): New header.
5633 * config/i386/cpuid.h (bit_MOVDIRI,
5634 bit_MOVDIR64B): New bits.
5635 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5637 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5638 (VOID, PVOID, PCVOID)): New function types.
5639 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5640 __builtin_ia32_directstoreu_u64,
5641 __builtin_ia32_movdir64b): New builtins.
5642 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5643 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5645 (ix86_valid_target_attribute_inner_p): Ditto.
5646 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5647 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5648 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5649 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5650 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5651 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5652 (movdiri<mode>, movdir64b_<mode>): New.
5653 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5654 * config/i386/immintrin.h: Include movdirintrin.h.
5655 * config/i386/movdirintrin.h: New file.
5656 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5658 2018-04-19 Richard Biener <rguenther@suse.de>
5661 * cfg.c (clear_bb_flags): When loop state says we have
5662 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5664 2018-04-19 Richard Biener <rguenther@suse.de>
5666 PR tree-optimization/84737
5667 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5668 copying restrict info.
5669 (vect_setup_realignment): Use it.
5670 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5671 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5672 the first DR to all generated stores.
5673 (vectorizable_load): Likewise for loads.
5675 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5677 PR tree-optimization/85446
5678 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5679 the integral and pointer types to have the same precision.
5681 * doc/install.texi: Document --disable-cet being the default and
5684 2018-04-18 Martin Liska <mliska@suse.cz>
5686 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5689 2018-04-18 Martin Liska <mliska@suse.cz>
5692 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5695 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5696 arguments if they are comparable.
5698 2018-04-18 Martin Liska <mliska@suse.cz>
5701 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5704 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5707 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5710 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5711 ENDBR after calling __morestack.
5713 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5716 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5717 by using gcc_base_ver to generate a gcc_driver_version, and use
5718 it when generating GCC_DRIVER_NAME.
5719 * configure: Regenerate.
5721 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5724 * config.gcc: Obsolete powerpc*-*-*spe*.
5726 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5729 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5730 (stabstr_D): Change type of unum from unsigned int to
5731 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5734 2018-04-17 Jim Wilson <jimw@sifive.com>
5737 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5738 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5739 Set arg_pointer_offset after using pretend_args_size.
5741 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5743 PR rtl-optimization/85431
5744 * dse.c (record_store): Ignore zero width stores.
5747 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5748 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5749 __builtin_stack_restore rather than after it.
5750 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5751 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5752 argument instead of virtual_dynamic_stack_rtx.
5754 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5756 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5758 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5759 Add note to error message to explain internal mapping of overloaded
5760 built-in function name to non-overloaded built-in function name.
5761 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5764 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5767 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5768 where the inputs overlap with the output.
5770 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5772 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5773 (=v, v) alternative and explicit "memory" attribute.
5774 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5775 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5777 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5778 "sselog1" type instead of "sselog".
5779 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5780 "sselog". Remove explicit "memory" attribute.
5781 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5782 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5784 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5785 "sselog1" type instead of "sselog". Remove explicit "memory"
5787 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5788 use "sselog1" type instead of "sselog". Remove explicit "memory"
5790 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5791 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5793 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5794 "sselog1" type instead of "sselog". Remove explicit "memory"
5796 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5797 use "sselog1" type instead of "sselog". Remove explicit "memory"
5801 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5804 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5807 2018-04-17 Martin Jambor <mjambor@suse.cz>
5810 * ipa-cp.c (create_specialized_node): Call
5811 expand_all_artificial_thunks if necessary.
5813 2018-04-17 Martin Liska <mliska@suse.cz>
5816 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5817 in message, remote space in between '_G' and '('.
5819 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5822 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5823 avx512f_vmcmp<mode>3<round_saeonly_name>,
5824 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5825 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5826 avx512f_rndscale<mode><round_saeonly_name>,
5827 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5828 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5829 Use %<iptr>2 instead of %2 for -masm=intel.
5830 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5831 avx512f_vcvttss2usi<round_saeonly_name>,
5832 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5834 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5835 avx512f_vcvttsd2usi<round_saeonly_name>,
5836 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5837 Use %q1 instead of %1 for -masm=intel.
5838 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5839 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5840 of %3 for -masm=intel.
5841 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5843 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5845 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5847 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5848 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5850 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5852 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5853 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5854 %g1 and one with %0 and %1.
5855 (avx512er_vmrcp28<mode><round_saeonly_name>,
5856 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5858 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5859 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5860 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5861 of %0 and %{%4%} for -masm=intel.
5862 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5863 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5864 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5865 order of %0 and %{%5%}%{z%} for -masm=intel.
5867 2018-04-17 Jan Hubicka <jh@suse.cz>
5870 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5872 2018-04-17 Martin Liska <mliska@suse.cz>
5875 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5876 for target_clone error message. Make default implementation
5877 clone to be a local declaration.
5878 (separate_attrs): Add new argument and check for an empty
5880 (expand_target_clones): Handle it.
5881 (ipa_target_clone): Make redirection just for target_clones
5884 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5885 Tom de Vries <tom@codesourcery.com>
5888 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5889 tiled basic blocks without omp continue statements.
5891 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5894 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5895 vec_extract expression as having side effects to make sure it gets
5898 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5901 * config/i386/i386.c (get_builtin_code_for_version): Check
5904 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5907 * gcc/config.gcc: Support "skylake".
5908 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5910 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5911 (processor_target_table): Add "skylake".
5912 (ix86_option_override_internal): Add "skylake".
5913 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5914 PROCESSOR_CANNONLAKE.
5915 (get_builtin_code_for_version): Fix priority for
5916 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5917 PROCESSOR_SKYLAKE-AVX512.
5918 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5919 (processor_type): Add PROCESSOR_SKYLAKE.
5921 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5922 Jason Merrill <jason@redhat.com>
5925 * convert.c (convert_to_integer_1): Use direct recursion for
5926 enumeral types and types with a precision less than the number
5927 of bits in their mode.
5929 2018-04-16 Julia Koval <julia.koval@intel.com>
5932 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5933 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5935 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5938 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5939 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5940 and -mno-direct-move.
5942 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5945 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5946 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5947 Ensure negative shifts result in {0}.
5949 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5951 PR rtl-optimization/79916
5952 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5953 regs (if any) to define how to gnerate SD moves when LRA is in
5956 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5958 PR rtl-optimization/85393
5959 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5960 * except.c (expand_dw2_landing_pad_for_region): Make static.
5961 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5962 a label and unconditional jump to old_bb, rather than
5963 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5966 PR rtl-optimization/85376
5967 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5968 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5969 instead of a specific value.
5971 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5972 Bin Cheng <bin.cheng@arm.com>
5974 PR tree-optimization/82965
5975 PR tree-optimization/83991
5976 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5977 by_profile_only parameter.
5978 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5979 information if the loop was predicted to iterate too many times.
5980 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5982 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5985 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5988 2018-04-13 Martin Liska <mliska@suse.cz>
5989 Jakub Jelinek <jakub@redhat.com>
5992 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
5993 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
5994 * builtins.c (expand_builtin_memory_copy_args): Use
5995 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
5996 handle dest_addr == pc_rtx.
5998 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6001 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6002 asked to not generate direct moves.
6003 (fix_trunc<mode>si2_stfiwx): Similar.
6004 (fix_trunc<mode>si2_internal): Similar.
6006 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6009 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6010 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6011 lookup if dest in some wider mode is known to be const0_rtx and
6012 if so, record permanent equivalence for it to be ZERO_EXTEND of
6013 the narrower mode destination.
6015 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6017 * lto-streamer-out.c (output_function): Revert 259346.
6018 * omp-expand.c (expand_oacc_for): Likewise.
6020 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6022 PR rtl-optimization/85354
6023 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6024 * sel-sched.c (sel_global_init): ... here.
6026 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6029 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6030 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6031 mode for PE-COFF targets.
6032 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6033 (i386_pe_asm_lto_end): Likewise.
6034 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6035 (TARGET_ASM_LTO_END): Likewise.
6036 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6037 (i386_pe_asm_lto_start): New function.
6038 (i386_pe_asm_lto_end): Likewise.
6040 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6041 Richard Biener <rguenther@suse.de>
6044 * lto-streamer-out.c (output_function): Fix CFG loop state before
6046 * omp-expand.c (expand_oacc_for): Handle calls to internal
6047 functions like regular functions.
6049 2018-04-12 Richard Biener <rguenther@suse.de>
6052 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6053 for the early LTO debug to properly generate references to it
6054 during DIE emission. Do not re-use that for the skeleton for
6056 (dwarf2out_early_finish): Likewise.
6058 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6061 * config/i386/sse.md
6062 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6063 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6064 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6065 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6066 and output is a reg, avoid creating invalid lowpart subreg, but
6067 instead split into a 512-bit move. Don't split if not AVX512VL,
6068 input is xmm16+ reg and output is a mem.
6069 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6070 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6071 xmm16+ reg and output is a mem.
6073 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6075 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6076 also for flag_dwarf2_cfi_asm.
6078 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6080 PR rtl-optimization/85342
6081 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6082 a bool scalar var inside of the loop instead. Don't try to update
6083 recog_data.operand after failed apply_change_group.
6085 2018-04-12 Tom de Vries <tom@codesourcery.com>
6088 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6089 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6090 array with flexible array member as array without given dimension.
6091 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6092 argument for undefined param to true.
6094 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6097 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6098 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6099 from PowerPC section.
6100 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6101 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6102 help text that is too long.
6103 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6104 help text that is too long.
6105 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6106 help text that is too long.
6108 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6110 * config/alpha/alpha.md (stack_probe_internal): Rename
6111 from "probe_stack". Update all callers.
6113 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6115 PR rtl-optimization/84566
6116 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6117 sched_macro_fuse_insns.
6119 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6122 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6123 (compute_block_dependences): ... from here.
6125 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6127 PR tree-optimization/85331
6128 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6129 from int to HOST_WIDE_INT.
6131 2018-04-11 Martin Jambor <mjambor@suse.cz>
6134 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6135 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6136 not the same as the source val.
6137 (cgraph_edge_brings_value_p): New parameter.
6138 (gather_edges_for_value): Pass destination value to
6139 cgraph_edge_brings_value_p.
6140 (perhaps_add_new_callers): Likewise.
6141 (get_info_about_necessary_edges): Likewise and exclude values brought
6142 only by self-recursive edges.
6143 (create_specialized_node): Redirect only clones of self-calling edges.
6144 (+self_recursive_pass_through_p): New function.
6145 (find_more_scalar_values_for_callers_subset): Use it.
6146 (find_aggregate_values_for_callers_subset): Likewise.
6147 (known_aggs_to_agg_replacement_list): Removed.
6148 (decide_whether_version_node): Re-calculate known constants for all
6149 remaining context clones.
6151 2018-04-11 Richard Biener <rguenther@suse.de>
6154 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6155 from early DWARF output.
6156 (dwarf2out_early_finish): Output line info unconditionally into
6157 early DWARF and add reference to it.
6159 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6162 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6163 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6164 other than V2DFmode using iptr mode attribute.
6165 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6167 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6169 PR rtl-optimization/84659
6170 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6172 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6175 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6177 (output_loc_list): Pass address of a dummy size variable even in the
6178 locview handling loop.
6179 (index_location_lists): Add comment on why skip_loc_list_entry can't
6182 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6185 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6188 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6191 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6192 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6193 and -mstring-compare-inline-limit.
6195 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6198 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6199 for stack clash protection in a register whenever we need it to be in
6202 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6204 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6205 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6207 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6210 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6212 (mlong-double-): Ditto.
6213 * config/rs6000/sysv4.opt (msdata=): Ditto.
6214 (mtls-size=): Ditto.
6216 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6218 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6219 erroneous entries for
6220 "vector int vec_ldl (int, long int *)", and
6221 "vector unsigned int vec_ldl (int, unsigned long int *)".
6222 Add comments and entries for
6223 "vector bool char vec_ldl (int, bool char *)",
6224 "vector bool short vec_ldl (int, bool short *)",
6225 "vector bool int vec_ldl (int, bool int *)",
6226 "vector bool long long vec_ldl (int, bool long long *)",
6227 "vector pixel vec_ldl (int, pixel *)",
6228 "vector long long vec_ldl (int, long long *)",
6229 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6230 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6231 type tree bool_long_long_type_node and correct definition of
6232 bool_V2DI_type_node to make reference to this new type tree.
6233 (rs6000_mangle_type): Replace erroneous reference to
6234 bool_long_type_node with bool_long_long_type_node.
6235 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6236 comments to emphasize sign distinctions for char and int types and
6237 replace RS6000_BTI_bool_long constant with
6238 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6239 use of RS6000_BTI_pixel.
6240 (bool_long_type_node): Remove this macro definition.
6241 (bool_long_long_type_node): New macro definition
6243 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6245 PR rtl-optimization/85300
6246 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6247 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6248 simplify_unary_operation fails.
6250 2018-04-10 Martin Liska <mliska@suse.cz>
6252 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6253 cgraph_edge and ipa_ref.
6255 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6259 * config/i386/sse.md
6260 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6261 computation of the VEC_MERGE selector from mask.
6262 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6263 Fix decoding of the VEC_MERGE selector into mask.
6265 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6267 PR tree-optimization/85286
6268 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6270 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6272 * final.c (final_1): Set insn_last_address as well as
6273 insn_current_address.
6275 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6278 * explow.c (emit_stack_probe): Call validize_mem on memory location
6279 before passing it to gen_probe_stack. Create address operand and
6280 legitimize it for the probe_stack_address case.
6282 2018-04-09 Jan Hubicka <jh@suse.cz>
6285 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6286 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6287 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6289 (free_lang_data): Rebuild type inheritance graph.
6291 2018-04-09 Martin Sebor <msebor@redhat.com>
6293 * invoke.texi (-finline-small-functions): Mention other optimization
6295 (-findirect-inlining, -fpartial-inlining): Same.
6296 (-finline-functions-called-once): Same.
6297 (-freorder-blocks-and-partition): Same.
6299 2018-04-09 Jan Hubicka <jh@suse.cz>
6302 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6303 jumps; choose last target that matches the criteria (i.e.
6304 no partition changes for non-crossing jumps).
6305 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6306 support for redirecting crossing jumps to non-crossing.
6308 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6310 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6311 also for naked functions.
6313 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6315 * config/arc/arc.md (add_shift): New pattern.
6316 (add_shift2): Likewise.
6317 (sub_shift): Likewise.
6318 (sub_shift_cmp0_noout): Likewise.
6319 (compare_si_ashiftsi): Likewise.
6320 (xbfu_cmp0_noout): New combine pattern.
6321 (xbfu_cmp0"): Likewise.
6322 (movsi_set_cc_insn): Place the predicable variant first.
6323 (commutative_binary_cmp0_noout): Remove clobber.
6324 (commutative_binary_cmp0): New pattern.
6325 (noncommutative_binary_cmp0): Likewise.
6326 (noncommutative_binary_cmp0_noout): Likewise.
6327 (noncommutative_binary_comparison_result_used): Removed.
6328 (rsub_cmp0): New pattern.
6329 (rsub_cmp0_noout): Likewise.
6330 (extzvsi): Changed, keep only meaningful variants.
6331 (SQH, SEZ): New iterators.
6332 (SQH_postfix): New mode attribute.
6333 (SEZ_prefix): New code attribute.
6334 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6335 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6336 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6338 (noncommutative_operator): Check the availability of barrel
6341 2018-04-09 Richard Biener <rguenther@suse.de>
6343 PR tree-optimization/85284
6344 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6345 Only use the niter constraining form of simple_iv when the exit
6348 2018-04-09 Tom de Vries <tom@codesourcery.com>
6351 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6352 (define_expand "*memory_barrier"): New define_expand.
6353 (define_insn "memory_barrier"): New insn.
6355 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6357 PR rtl-optimization/80463
6358 PR rtl-optimization/83972
6359 PR rtl-optimization/83480
6361 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6362 correct producer for the insn.
6363 (tidy_control_flow): Fixup seqnos in case of debug insns.
6365 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6367 PR rtl-optimization/83913
6369 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6370 different sched-times when merging exprs.
6372 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6374 PR rtl-optimization/83962
6376 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6377 tidy_fallthru_edge and tidy_control_flow.
6379 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6381 PR rtl-optimization/83530
6383 * sel-sched.c (force_next_insn): New global variable.
6384 (remove_insn_for_debug): When force_next_insn is true, also leave only
6385 next insn in the ready list.
6386 (sel_sched_region): When the region wasn't scheduled, make another pass
6387 over it with force_next_insn set to 1.
6389 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6391 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6393 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6394 for interrupt control.
6395 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6396 functions for interrupt control.
6397 * config/nds32/nds32-intrinsic.md: Likewise.
6398 * config/nds32/nds32_intrinsic.h: Likewise.
6399 * config/nds32/nds32.h (nds32_builtins): Likewise.
6401 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6403 * config/nds32/nds32.c (nds32_init_machine_status,
6404 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6405 strict_aligned_p field.
6406 (nds32_expand_to_rtl_hook): New function.
6407 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6408 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6410 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6411 Chung-Ju Wu <jasonwucj@gmail.com>
6413 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6414 * config/nds32/nds32-n7.md: New file.
6415 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6416 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6418 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6419 * config/nds32/nds32.md (pipeline_model): Add n7.
6420 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6421 * config/nds32/pipelines.md: Include n7 settings.
6423 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6424 Chung-Ju Wu <jasonwucj@gmail.com>
6426 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6427 * config/nds32/nds32-e8.md: New file.
6428 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6429 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6431 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6432 * config/nds32/nds32.md (pipeline_model): Add e8.
6433 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6434 * config/nds32/pipelines.md: Include e8 settings.
6436 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6437 Chung-Ju Wu <jasonwucj@gmail.com>
6439 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6440 * config/nds32/nds32-n8.md: New file.
6441 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6442 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6444 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6445 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6446 * config/nds32/nds32.md (pipeline_model): Add n8.
6447 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6448 * config/nds32/pipelines.md: Include n8 settings.
6450 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6451 Chung-Ju Wu <jasonwucj@gmail.com>
6453 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6454 * config/nds32/nds32-n9-2r1w.md: New file.
6455 * config/nds32/nds32-n9-3r2w.md: New file.
6456 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6457 nds32_register_ports): New or modify for cpu n9.
6458 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6460 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6461 * config/nds32/nds32-utils.c: New file.
6462 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6463 TARGET_MUL_SLOW): Define.
6464 * config/nds32/nds32.md (pipeline_model): New attribute.
6465 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6466 New options that support cpu n9.
6467 * config/nds32/pipelines.md: Include n9 settings.
6468 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6470 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6472 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6473 information if necessary.
6474 (output_cond_branch_compare_zero): Likewise.
6475 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6476 (nds32_target_alignment): Refine for alignment.
6477 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6478 (FUNCTION_BOUNDARY): Modify.
6479 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6481 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6483 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6485 * config/nds32/constants.md (unspec_volatile_element): Add values for
6486 TLB operation and data prefetch.
6487 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6488 functions for TLB operation and data prefetch.
6489 * config/nds32/nds32-intrinsic.md: Likewise.
6490 * config/nds32/nds32_intrinsic.h: Likewise.
6491 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6492 (nds32_print_operand): Likewise.
6493 * config/nds32/nds32.h (nds32_builtins): Likewise.
6495 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6496 Andrew Pinski <pinsika@gcc.gnu.org>
6499 * match.pd: Use constant_boolean_node of correct type instead of
6500 boolean_true_node or boolean_false_node for simplifying
6501 pointer comparisons to zero.
6503 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6505 PR tree-optimization/80021
6506 * tree.c (verify_type_variant): Make error call in verify_variant_match
6507 translatable and remove final full stop.
6509 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6511 * config/nds32/constants.md (unspec_volatile_element): Add
6512 UNSPEC_VOLATILE_EH_RETURN.
6513 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6514 nds32_output_stack_pop): Support dwarf exception handling process.
6515 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6516 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6517 exception handling process.
6518 (nds32_compute_stack_frame): Likewise.
6519 (nds32_return_addr_rtx): Likewise.
6520 (nds32_initial_elimination_offset): Likewise.
6521 (nds32_expand_prologue): Likewise.
6522 (nds32_expand_epilogue): Likewise.
6523 (nds32_dynamic_chain_address): New function.
6524 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6526 (DYNAMIC_CHAIN_ADDRESS): Define.
6527 (EH_RETURN_DATA_REGNO): Define.
6528 (EH_RETURN_STACKADJ_RTX): Define.
6529 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6530 patterns for dwarf exception handling.
6532 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6534 * config/nds32/nds32.h: Clean up obsolete macros.
6536 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6538 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6539 Add enum values for particular instructions.
6540 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6541 particular intrinsic functions.
6542 * config/nds32/nds32-intrinsic.md: Likewise.
6543 * config/nds32/nds32_intrinsic.h: Likewise.
6544 * config/nds32/nds32.h (nds32_builtins): Likewise.
6545 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6546 (btst, ave): New patterns for particular instructions.
6548 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6550 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6551 Add enum values for atomic load/store and memory sync.
6552 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6554 * config/nds32/nds32-intrinsic.md: Likewise.
6555 * config/nds32/nds32_intrinsic.h: Likewise.
6556 * config/nds32/nds32.h (nds32_builtins): Likewise.
6558 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6560 PR tree-optimization/85257
6561 * fold-const.c (native_encode_vector): If not all elts could fit
6562 and off is -1, return 0 rather than offset.
6563 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6564 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6565 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6566 adjust buffer in native_interpret_expr call.
6568 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6570 * config/nds32/constants.md (unspec_volatile_element): Add cache
6571 control enum values.
6572 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6573 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6574 * config/nds32/nds32.c (nds32_cctl_names): New.
6575 (nds32_print_operand): Handle cache control register names.
6576 * config/nds32/nds32.h (nds32_builtins): New enum values.
6577 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6579 * config/nds32/nds32.md (type): Add mmu.
6580 * config/nds32/pipelines.md (simple_insn): Add mmu.
6582 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6584 * config/nds32/nds32.md (type): Remove call.
6585 * config/nds32/pipelines.md (simple_insn): Likewise.
6587 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6589 * config/nds32/constants.md (unspec_volatile_element): Add
6590 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6591 UNSPEC_VOLATILE_FMFCFG.
6592 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6593 description for fmfcfg and fmfcsr.
6594 (bdesc_1arg): Add fmtcsr.
6595 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6596 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6597 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6598 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6599 unspec_fmfcfg): New patterns.
6600 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6601 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6602 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6603 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6604 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6605 __nds32__fmfcfg): Define.
6607 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6609 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6610 intrinsic register names.
6611 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6612 intrinsic register enum values and macros.
6614 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6616 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6617 for load/store addressing form.
6618 (nds32_print_operand_address): Likewise.
6620 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6623 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6624 based on LABEL_REF. Remove useless assertion.
6625 (pic_address_needs_scratch): Fix formatting.
6626 (sparc_legitimize_pic_address): Minor tweaks.
6627 (sparc_delegitimize_address): Adjust assertion accordingly.
6628 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6629 into symbolic_operand.
6630 (movsi_high_pic_label_ref): Likewise.
6631 (movsi_lo_sum_pic_label_ref): Likewise.
6632 (movdi_pic_label_ref): Likewise.
6633 (movdi_high_pic_label_ref): Likewise.
6634 (movdi_lo_sum_pic_label_ref): Likewise.
6636 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6638 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6639 custom LIB_SPEC setup.
6641 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6642 Kito Cheng <kito.cheng@gmail.com>
6644 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6645 * config/riscv/freebsd.h: New.
6647 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6649 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6650 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6653 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6654 Kito Cheng <kito.cheng@gmail.com>
6656 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6657 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6658 * config/nds32/nds32-protos.h (nds32_output_call,
6659 nds32_output_return): Declare.
6660 * config/nds32/nds32.md: Refine all the call and return patterns.
6662 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6665 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6666 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6668 PR rtl-optimization/84872
6669 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6670 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6673 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6675 * expr.c (copy_blkmode_to_reg): Revert 254862.
6676 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6678 2018-04-06 Richard Biener <rguenther@suse.de>
6681 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6682 after seeing a component reference with an adjacent field. Treat
6683 refs to arrays at struct end of external decls similar to
6684 refs to unconstrained commons.
6686 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6689 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6690 look through SAVE_EXPRs with non-side-effects argument. Adjust
6692 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6695 2018-04-06 Richard Biener <rguenther@suse.de>
6698 * alias.c (find_base_term): New wrapper around find_base_term
6699 unwinding CSELIB_VAL_PTR changes.
6700 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6703 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6705 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6707 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6708 constant definitions.
6709 ("nop"): lr 0,0 -> nopr r0
6710 ("nop_lr0", "nop_lr1"): New insn definitions.
6712 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6714 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6715 NDS32_V3PUSH_AVAILABLE_P macro.
6717 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6718 Chung-Ju Wu <jasonwucj@gmail.com>
6720 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6721 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6722 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6723 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6724 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6725 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6726 * config/nds32/constraints.md: New constraints and checking for hard
6727 float configuration.
6728 * config/nds32/iterators.md: New mode iterator and attribute for hard
6729 float configuration.
6730 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6732 * config/nds32/nds32-fpu.md: New file.
6733 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6734 deal with hard float code generation.
6735 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6737 (abi_type, float_reg_number): New enum type.
6738 * config/nds32/nds32-predicates.c: New predicates for hard float.
6739 * config/nds32/nds32-protos.h: Declare functions for hard float.
6740 * config/nds32/nds32.c: Implementation for hard float configuration.
6741 * config/nds32/nds32.h: Definitions for hard float configuration.
6742 * config/nds32/nds32.md: Include hard float machine description and
6743 modify patterns for hard float configuration.
6744 * config/nds32/nds32.opt: New options for hard float configuration.
6745 * config/nds32/predicates.md: New predicates for hard float
6748 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6750 * common/config/nds32/nds32-common.c
6751 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6753 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6756 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6757 CONSTRUCTOR_ELT (ctor, ...)->value.
6759 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6762 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6764 2018-04-05 Tom de Vries <tom@codesourcery.com>
6767 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6770 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6771 Kito Cheng <kito.cheng@gmail.com>
6773 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6774 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6775 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6776 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6778 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6779 Kito Cheng <kito.cheng@gmail.com>
6781 * config/nds32/constraints.md (Ufe): New memory constraint.
6782 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6783 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6784 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6786 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6787 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6789 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6791 * config/nds32/nds32.md: Use optimize_size in the condition for
6792 alu-shift instructions.
6794 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6796 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6798 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6800 * config/nds32/nds32.md (negsi2): Refine pattern.
6802 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6803 Chung-Ju Wu <jasonwucj@gmail.com>
6805 * config/nds32/iterators.md (shift_rotate): New code iterator.
6806 (shift): New code attribute.
6807 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6808 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6809 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6810 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6811 bit-wise operations.
6812 (andsi3, *andsi3): Ditto.
6813 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6814 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6815 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6816 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6817 nds32_ior_operand, nds32_xor_operand): New predicates.
6819 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6821 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6822 (addsi3, subsi3): ... this.
6824 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6826 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6828 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6830 * config/nds32/nds32.md: Adjust indention.
6832 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6834 * config/nds32/nds32.md (feature): New attribute.
6836 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6838 * config/nds32/nds32.md (subtype): New attribute.
6840 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6843 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6844 expansion to perform a bitwise AND of the argument followed by a
6845 boolean negation of the result.
6847 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6849 PR rtl-optimization/84878
6850 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6851 the basic block. Assert the use reference is not artificial and that
6852 it has an associated insn.
6854 2018-04-04 Michael Matz <matz@suse.de>
6856 * builtins.c (compute_objsize): Pass correct operand
6857 to array_at_struct_end_p.
6859 2018-04-04 Richard Biener <rguenther@suse.de>
6862 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6863 from contexts for DINFO_LEVEL_TERSE and below.
6865 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6867 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6868 resiter_operand condition.
6869 * config/nds32/nds32.md (*move<mode>): Ditto.
6871 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6872 Monk Chiang <sh.chiang04@gmail.com>
6874 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6876 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6878 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6880 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6881 Kito Cheng <kito.cheng@gmail.com>
6883 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6884 nds32_cond_code_str, output_cond_branch,
6885 output_cond_branch_compare_zero, nds32_expand_cbranch,
6886 nds32_expand_cstore, nds32_expand_movcc,
6887 nds32_output_cbranchsi4_equality_zero,
6888 nds32_output_cbranchsi4_equality_reg,
6889 nds32_output_cbranchsi4_equality_reg_or_const_int,
6890 nds32_output_cbranchsi4_greater_less_zero: New functions.
6891 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6892 nds32_expand_cstore, nds32_expand_movcc,
6893 nds32_output_cbranchsi4_equality_zero,
6894 nds32_output_cbranchsi4_equality_reg,
6895 nds32_output_cbranchsi4_equality_reg_or_const_int,
6896 nds32_output_cbranchsi4_greater_less_zero): Declare.
6897 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6898 nds32_rimm11s_operand): New predicates.
6899 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6900 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6903 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6905 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6906 * config/nds32/nds32.md: Ditto.
6907 * config/nds32/pipelines.md: Ditto.
6909 2018-04-04 Richard Biener <rguenther@suse.de>
6911 PR tree-optimization/85168
6912 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6913 propagating abnormals.
6915 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6917 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6919 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6920 Kito Cheng <kito.cheng@gmail.com>
6922 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6923 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6924 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6925 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6926 * config/nds32/nds32.md (sibcall_internal): New.
6927 (sibcall_register): Remove.
6928 (sibcall_immediate): Remove.
6929 (sibcall_value_internal): New.
6930 (sibcall_value_register): Remove.
6931 (sibcall_value_immediate): Remove.
6932 * config/nds32/predicates.md (nds32_general_register_operand): New.
6933 (nds32_call_address_operand): New.
6935 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6937 PR rtl-optimization/85167
6938 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6939 bb_defs if *split_p, instead preinitialize it to NULL.
6941 PR tree-optimization/85156
6942 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6943 evaluating the argument multiple times.
6945 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6947 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6949 (_mm_cvtpd_ps): Likewise.
6950 (_mm_cvttpd_epi32): Likewise.
6951 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6952 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6953 vector, pixel, and bool following altivec.h include.
6955 2018-04-03 Martin Sebor <msebor@redhat.com>
6957 * doc/extend.texi (Common Function Attributes): Clarify.
6958 (const attribute): Likewise.
6959 (pure attribute): Likewise.
6961 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6964 * config/i386/i386.c (ix86_expand_vector_set): Use
6965 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6967 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6969 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6970 instructions when changing rounding bits to preserve precision bits
6971 in the x87 control word.
6973 2018-04-03 Martin Liska <mliska@suse.cz>
6975 PR tree-optimization/82491
6976 * rtl.h (strip_offset_and_add): Replace += suboffset with
6977 poly_uint64 () + suboffset.
6979 2018-03-29 Martin Liska <mliska@suse.cz>
6980 Martin Jambor <mjambor@suse.cz>
6983 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6984 param_type is not an integral or pointer type.
6986 2018-04-03 Richard Biener <rguenther@suse.de>
6988 * sese.h (recompute_all_dominators): Remove.
6990 2018-04-02 Martin Sebor <msebor@redhat.com>
6992 * doc/invoke.texi (-Wrestrict): Fix typos.
6994 2018-04-02 Jim Wilson <jimw@sifive.com>
6996 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
6997 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
6998 (<optab>di3, <optab>si3_extend): Likewise.
6999 (<optab>si3_mask, <optab>si3_mask_1): New.
7000 (<optab>di3_mask, <optab>di3_mask_1): New.
7001 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7002 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7003 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7005 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7007 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7010 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7012 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7013 (nds32_canonicalize_comparison): New function.
7015 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7016 Kito Cheng <kito.cheng@gmail.com>
7017 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7019 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7020 * config/nds32/constants.md (unspec_volatile_element): Add
7021 UNSPEC_VOLATILE_RELAX_GROUP.
7022 * config/nds32/nds32-relax-opt.c: New file.
7023 * config/nds32/nds32-predicates.c
7024 (nds32_symbol_load_store_p): New function.
7025 * config/nds32/nds32-protos.h
7026 (nds32_symbol_load_store_p): Declare function.
7027 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7028 * config/nds32/nds32.c
7029 (nds32_register_pass): New function to register pass.
7030 (nds32_register_passes): New function to register passes.
7031 * config/nds32/nds32.md (relax_group): New pattern.
7032 * config/nds32/nds32.opt (mrelax-hint): New option.
7033 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7035 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7037 * config/nds32/t-nds32: Modify files dependency.
7039 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7041 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7042 (PROFILE_HOOK): Define its implementation.
7044 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7046 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7047 type and 32-bit size.
7049 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7052 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7053 (V_128_256): New mode iterator.
7054 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7055 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7056 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7058 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7059 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7061 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7064 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7065 NaN inputs correctly.
7067 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7070 * config/rs6000/vsx.md (??r): New mode attribute.
7071 (*vsx_mov<mode>_64bit): Use it.
7072 (*vsx_mov<mode>_32bit): Likewise.
7074 2018-03-30 Martin Sebor <msebor@redhat.com>
7076 PR tree-optimization/84818
7077 * builtins.c (check_access): Use warning_n.
7079 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7082 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7084 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7087 2018-03-30 Julia Koval <julia.koval@intel.com>
7090 * x86-tune.def (movx, partial_reg_dependency): Enable for
7093 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7096 * lra-constraints.c (process_alt_operands): Move setting
7097 this_alternative_matches below.
7099 2018-03-29 Martin Liska <mliska@suse.cz>
7102 * doc/invoke.texi: Document how LTO works with debug info.
7103 Describe auto-load support of binutils. Mention 'x86-64'
7104 as valid option value of -march option.
7106 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7108 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7111 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7112 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7113 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7116 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7119 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7120 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7121 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7122 for RS6000_BTM_POWERPC64.
7123 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7124 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7125 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7130 2018-03-28 Carl Love <cel@us.ibm.com>
7133 2017-09-27 Carl Love <cel@us.ibm.com>
7135 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7136 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7137 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7140 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7142 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7143 instead of __vector bool.
7144 (_mm_max_pu8): Likewise.
7145 (_mm_min_pi16): Likewise.
7147 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7150 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7151 (DIVWEUO): Likewise.
7153 (DIVDEUO): Likewise.
7154 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7155 DIVWEUO and DIVDEUO.
7156 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7157 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7158 (div_extend): Likewise.
7159 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7161 (__builtin_divweuo): Likewise.
7162 (__builtin_divdeo): Likewise.
7163 (__builtin_divdeuo): Likewise.
7165 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7168 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7169 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7171 PR tree-optimization/82004
7172 * gimple-match-head.c (optimize_pow_to_exp): New function.
7173 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7174 Don't fold to exp if optimize_pow_to_exp is false.
7176 2018-03-28 Martin Liska <mliska@suse.cz>
7179 * calls.c (initialize_argument_information): Fix trailing space.
7180 * common.opt: Fix typo and provide better explanation for
7181 -fsanitize-coverage option.
7182 * config/i386/i386.opt: Fix typo.
7184 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7185 Martin Liska <mliska@suse.cz>
7188 * gimplify.c (asan_poison_variable): Don't do the check for
7189 gimplify_omp_ctxp here.
7190 (gimplify_decl_expr): Do it here.
7191 (gimplify_target_expr): Likewise.
7193 2018-03-28 Martin Liska <mliska@suse.cz>
7196 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7197 chkp_type_bounds_count if MPX is not enabled.
7199 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7201 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7203 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7206 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7207 function to create the function decl for complex long double
7208 multiply and divide for -mabi=ieeelongdouble.
7209 (init_float128_ieee): Call it.
7211 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7214 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7215 -fcf-protection=branch -mibt.
7216 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7218 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7221 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7222 arm_word_relocations.
7224 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7227 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7228 extern array declarations.
7230 2018-03-27 Richard Biener <rguenther@suse.de>
7233 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7234 explicit single_use checks.
7236 2018-03-27 Richard Biener <rguenther@suse.de>
7238 PR tree-optimization/85082
7239 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7242 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7244 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7245 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7246 Turn on fasynchronous-unwind-tables and funwind-tables.
7248 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7251 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7252 (*bmi_blsr_<mode>_ccz): Ditto.
7254 2018-03-26 Tom de Vries <tom@codesourcery.com>
7256 PR tree-optimization/85063
7257 * omp-general.c (offloading_function_p): New function. Factor out
7259 * omp-offload.c (pass_omp_target_link::gate): ... here.
7260 * omp-general.h (offloading_function_p): Declare.
7261 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7262 with attribute omp declare target for offloading functions.
7264 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7266 PR tree-optimization/84005
7267 * tree-data-ref.h (get_base_for_alignment): Declare.
7268 * tree-data-ref.c (get_base_for_alignment_1): New function.
7269 (get_base_for_alignment): Likewise.
7270 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7271 get_base_for_alignment to find a suitable base object, instead
7272 of always using drb->base_address.
7274 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7277 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7278 known size by default.
7280 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7283 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7284 and non BLKmode operands.
7286 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7289 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7290 Clean up attributes.
7292 2018-03-23 Richard Biener <rguenther@suse.de>
7295 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7296 we are going to emit early debug for LTO.
7298 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7301 * function.c (match_asm_constraints_1): Don't optimize if input
7302 doesn't satisfy general_operand predicate for output's mode.
7305 * alias.c (write_dependence_p): Don't require for x_canonicalized
7306 non-VOIDmode if x has VOIDmode.
7309 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7310 just don't try to optimize it rather than assert it never happens.
7312 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7314 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7315 macro expansions for definition of ST_INTERNAL_<mode> and
7316 LD_INTERNAL_<mode> builtins.
7317 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7319 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7321 (altivec_expand_st_builtin): Likewise.
7322 (altivec_expand_builtin): Remove calls to deleted functions.
7323 (rs6000_address_for_altivec): Delete this function.
7324 * config/rs6000/vector.md: Remove expands for
7325 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7327 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7330 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7331 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7332 re-computing once computed.
7333 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7334 (arm_init_machine_status): Initialize
7335 machine->static_chain_stack_bytes.
7337 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7340 * doc/extend.texi: Add four new prototypes for vec_ld.
7341 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7342 definitions for more logical presentation.
7343 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7344 entries for V1TI variants of __builtin_altivec_ld builtin.
7345 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7346 handling of V1TI variant of LVX icode pattern.
7347 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7348 (rs6000_gimple_fold_builtin): Likewise.
7349 (altivec_init_builtins): Add code to define
7350 __builtin_altivec_lvx_v1ti function.
7352 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7355 * function.c (match_asm_constraints_1): Don't do the optimization
7356 if input isn't a REG, SUBREG, MEM or constant.
7358 2018-03-22 Tom de Vries <tom@codesourcery.com>
7360 PR tree-optimization/84956
7361 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7362 bb_has_abnormal_pred.
7364 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7367 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7368 DECL_INITIAL (decl) to decl at the end.
7369 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7372 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7374 * doc/extend.texi (__builtin_tgmath): Document when complex
7375 integer types are treated as _Complex _Float64.
7377 2018-03-21 Tom de Vries <tom@codesourcery.com>
7379 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7381 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7383 PR tree-optimization/84960
7384 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7385 if it is ENTRY block, move them into single succ of ENTRY in that case.
7387 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7389 PR tree-optimization/84811
7390 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7391 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7392 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7393 (known_subrange_p): Update accordingly. Cast each value involved
7394 in the size comparison, rather than casting the result of the
7397 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7399 PR tree-optimization/84982
7400 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7401 by flipping the least significant bit rather than all bits from
7402 bitpos to bitpos + bitsize - 1.
7404 2018-03-21 Nathan Sidwell <nathan@acm.org>
7406 * doc/extend.texi (Deprecated Features): Remove mention of
7407 long-deleted deprecations.
7409 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7412 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7413 * configure: Regenerate.
7415 2018-03-21 Tom de Vries <tom@codesourcery.com>
7417 PR tree-optimization/83126
7418 * tree-parloops.c (num_phis): New function.
7419 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7421 2018-03-21 Nathan Sidwell <nathan@acm.org>
7423 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7424 mention anon-struct/union members and trailing attributes.
7426 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7428 PR tree-optimization/84969
7429 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7430 builtin memset partitions if they set different rhs values.
7432 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7434 PR rtl-optimization/84989
7435 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7436 VEC_DUPLICATE with scalar result mode.
7438 2018-03-21 Martin Liska <mliska@suse.cz>
7441 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7442 not intended return statement.
7444 2018-03-21 Martin Liska <mliska@suse.cz>
7447 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7448 (chkp_find_bound_slots_1): Limit number of iterations.
7450 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7453 * Minor grammar fixes for x86 options.
7455 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7458 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7459 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7461 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7464 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7465 (altivec_lvx_<mode>_1op): Likewise.
7466 (altivec_stvx_<mode>_2op): Likewise.
7467 (altivec_stvx_<mode>_1op): Likewise.
7468 (altivec_lvx_<VM2:mode>): New define_expand.
7469 (altivec_stvx_<VM2:mode>): Likewise.
7470 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7471 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7472 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7473 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7474 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7475 (rs6000_gen_lvx): Likewise.
7476 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7477 (altivec_expand_stv_builtin): Likewise.
7478 (altivec_expand_builtin): Likewise.
7479 * config/rs6000/vector.md: Likewise.
7481 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7484 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7487 2018-03-20 Richard Biener <rguenther@suse.de>
7490 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7491 sign-conversions as zero, fall back to standard scalar_stmt
7494 2018-03-20 Martin Liska <mliska@suse.cz>
7497 * predict.c (rebuild_frequencies): Handle case when we have
7498 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7500 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7503 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7504 flag_section_anchors.
7505 * varasm.c (use_blocks_for_decl_p): Remove hack for
7506 dw2_force_const_mem.
7509 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7511 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7512 be created, use lowpart_subreg of operands[0] rather than operands[0]
7514 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7515 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7516 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7517 and n constraint instead of aarch64_shift_imm_di and Usd.
7518 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7519 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7521 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7524 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7525 to favor GPR over NEON registers.
7526 (<shift>di3_neon): Likewise.
7528 2018-03-20 Tom de Vries <tom@codesourcery.com>
7531 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7532 (nvptx_process_pars): Emit bar.sync asap and alap.
7534 2018-03-20 Tom de Vries <tom@codesourcery.com>
7537 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7538 seen_label if seen_label is already set.
7540 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7543 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7544 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7545 Use 1U instead of 1. Formatting fixes.
7548 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7549 instead of TREE_TYPE (s1) for the return value.
7551 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7553 PR tree-optimization/84946
7554 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7555 bitsize + bitsize in poly_uint64 rather than poly_int64.
7558 * dwarf2asm.c: Include fold-const.c.
7559 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7560 of decl rather than decl itself.
7562 PR rtl-optimization/84643
7563 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7565 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7568 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7569 calling assemble_variable.
7571 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7574 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7575 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7577 2018-03-19 Jim Wilson <jimw@sifive.com>
7580 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7581 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7582 (riscv_first_stack_step): Likewise.
7583 (riscv_option_override): Use STACK_BOUNDARY instead of
7585 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7587 (BIGGEST_ALIGNMENT): Set to 128.
7588 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7589 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7592 2018-03-19 Richard Biener <rguenther@suse.de>
7594 PR tree-optimization/84933
7595 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7596 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7598 2018-03-19 Richard Biener <rguenther@suse.de>
7600 PR tree-optimization/84859
7601 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7602 (cond_if_else_store_replacement): Perform sinking operation on
7603 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7604 Generalize what a BB with a single eligible store is.
7606 2018-03-19 Richard Biener <rguenther@suse.de>
7608 PR tree-optimization/84929
7609 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7610 chrec_is_positive against non-chrec arg.
7612 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7615 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7617 2018-03-18 Martin Liska <mliska@suse.cz>
7619 PR rtl-optimization/84635
7620 * regrename.c (build_def_use): Use matches_mode only when
7623 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7625 PR tree-optimization/84913
7626 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7627 vectorize chains of COND_EXPRs.
7629 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7631 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7633 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7635 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7637 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7639 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7641 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7642 Kito Cheng <kito.cheng@gmail.com>
7644 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7645 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7646 (nds32_adjust_reg_alloc_order): New function.
7647 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7649 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7651 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7652 nds32_print_operand, nds32_print_operand_address): Use
7653 HOST_WIDE_INT_PRINT_DEC instead.
7655 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7657 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7659 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7662 * config/i386/i386.c (initial_ix86_tune_features,
7663 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7665 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7666 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7667 rather than 1u << ix86_tune. Formatting fix.
7668 (ix86_option_override_internal): Change ix86_arch_mask from
7669 unsigned int to unsigned HOST_WIDE_INT, initialize to
7670 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7671 (ix86_function_specific_restore): Likewise.
7673 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7676 * postreload.c (reload_combine_recognize_pattern): Perform
7677 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7678 truncate_int_for_mode the result for the destination's mode.
7681 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7682 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7684 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7687 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7690 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7691 regno_allocno_class_array and sorted_pseudos.
7692 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7693 insns where regno is used.
7695 2018-03-16 Martin Liska <mliska@suse.cz>
7698 * multiple_target.c (create_dispatcher_calls): Redirect
7699 reference in the symbol table.
7701 2018-03-16 Martin Liska <mliska@suse.cz>
7704 * multiple_target.c (create_dispatcher_calls): Redirect also
7707 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7711 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7712 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7713 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7715 2018-03-16 Julia Koval <julia.koval@intel.com>
7717 * doc/invoke.texi (Skylake Server): Add CLWB.
7718 Cannonlake): Remove CLWB.
7720 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7722 PR tree-optimization/84841
7723 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7725 (FLOAT_ONE_CONST_TYPE): Define.
7726 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7727 (sort_by_operand_rank): Put entries with higher constant_type last
7728 rather than first to match comments.
7730 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7732 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7735 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7738 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7739 check and use address of target always.
7741 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7744 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7745 (indirect_thunk_bnd_needed): Likewise.
7746 (indirect_thunks_used): Likewise.
7747 (indirect_thunks_bnd_used): Likewise.
7748 (indirect_return_needed): New.
7749 (indirect_return_bnd_needed): Likewise.
7750 (output_indirect_thunk_function): Add a bool argument for
7752 (output_indirect_thunk_function): Don't generate alias for
7753 function return thunk.
7754 (ix86_code_end): Call output_indirect_thunk_function to generate
7755 function return thunks.
7756 (ix86_output_function_return): Set indirect_return_bnd_needed
7757 and indirect_return_needed instead of indirect_thunk_bnd_needed
7758 and indirect_thunk_needed.
7760 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7762 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7763 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7764 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7766 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7767 Paul Hua <paul.hua.gm@gmail.com>
7770 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7772 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7774 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7775 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7778 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7781 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7782 instead of GET_MODE_SIZE when comparing Units.
7784 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7787 * varasm.c (hash_section): Return an unchangeble hash value
7788 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7789 Return !aarch64_can_use_per_function_literal_pools_p ().
7791 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7794 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7795 rather than address of cmode as last argument to prepare_cmp_insn.
7797 2018-03-15 Julia Koval <julia.koval@intel.com>
7799 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7800 F_AVX512VNNI, F_AVX512BITALG): New.
7802 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7805 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7806 insn for floating-point loads and stores.
7808 2018-03-14 Carl Love <cel@us.ibm.com>
7810 * config/rs6000/rs6000-c.c: Add macro definitions for
7811 ALTIVEC_BUILTIN_VEC_PERMXOR.
7812 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7813 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7814 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7815 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7817 * config/doc/extend.texi: Add prototypes for vec_permxor.
7819 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7822 * diagnostic-show-locus.c (class layout_point): Convert m_line
7823 from int to linenum_type.
7824 (line_span::comparator): Use linenum "compare" function when
7825 comparing line numbers.
7826 (test_line_span): New function.
7827 (layout_range::contains_point): Convert param "row" from int to
7829 (layout_range::intersects_line_p): Likewise.
7830 (layout::will_show_line_p): Likewise.
7831 (layout::print_source_line): Likewise.
7832 (layout::should_print_annotation_line_p): Likewise.
7833 (layout::print_annotation_line): Likewise.
7834 (layout::print_leading_fixits): Likewise.
7835 (layout::annotation_line_showed_range_p): Likewise.
7836 (struct line_corrections): Likewise for field m_row.
7837 (line_corrections::line_corrections): Likewise for param "row".
7838 (layout::print_trailing_fixits): Likewise.
7839 (layout::get_state_at_point): Likewise.
7840 (layout::get_x_bound_for_row): Likewise.
7841 (layout::print_line): Likewise.
7842 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7843 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7844 * input.c (selftest::test_linenum_comparisons): New function.
7845 (selftest::input_c_tests): Call it.
7846 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7847 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7848 * selftest.h (ASSERT_GT): New macro.
7849 (ASSERT_GT_AT): New macro.
7850 (ASSERT_LT): New macro.
7851 (ASSERT_LT_AT): New macro.
7853 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7855 PR rtl-optimization/84780
7856 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7858 2018-03-14 Martin Liska <mliska@suse.cz>
7860 * tree.c (record_node_allocation_statistics): Use
7861 get_stats_node_kind.
7862 (get_stats_node_kind): New function extracted from
7863 record_node_allocation_statistics.
7864 (free_node): Use get_stats_node_kind.
7866 2018-03-14 Richard Biener <rguenther@suse.de>
7868 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7869 that the value-set of ANTIC_IN doesn't grow.
7872 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7874 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7875 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7877 2018-03-14 Julia Koval <julia.koval@intel.com>
7879 * config.gcc (icelake-client, icelake-server): New.
7881 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7882 (initial_ix86_arch_features): Ditto.
7883 (PTA_SKYLAKE): Add SGX.
7884 (PTA_ICELAKE): Remove.
7885 (PTA_ICELAKE_CLIENT): New.
7886 (PTA_ICELAKE_SERVER): New.
7887 (ix86_option_override_internal): Split up icelake on icelake client and
7889 (get_builtin_code_for_version): Ditto.
7890 (fold_builtin_cpu): Ditto.
7891 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7892 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7893 * config/i386/i386.h (processor_type): Ditto.
7894 * doc/invoke.texi: Ditto.
7896 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7899 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7900 INTEGER_CST offset, add it together with bitpos / 8 and
7901 sign extend based on POINTER_SIZE.
7905 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7908 * config/i386/constraints.md (Yc): New register constraint.
7909 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7910 Use Yc constraint for alternative 2 of operand 0. Remove
7911 preferred_for_speed attribute.
7913 2018-03-14 Richard Biener <rguenther@suse.de>
7915 PR tree-optimization/84830
7916 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7917 with the old one to avoid oscillations.
7919 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7922 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7924 (assign_by_spills): Return a flag of reload assignment failure.
7925 Do not process the reload assignment failures. Do not spill other
7926 reload pseudos if they has the same reg class. Update n if
7928 (lra_assign): Add a return arg. Set up from the result of
7929 assign_by_spills call.
7930 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7931 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7932 usage_insns if it is not NULL.
7933 (spill_hard_reg_in_range): New function.
7934 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7935 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7936 function prototypes.
7937 (lra_assign): Change prototype.
7938 * lra.c (lra): Add code to deal with fails by splitting hard reg
7941 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7943 * config/riscv/riscv.opt (mrelax): New option.
7944 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7945 "norelax" when riscv_mrelax is disabled.
7946 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7948 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7951 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7952 reassociation for int modes.
7954 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7956 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7957 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7959 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7960 * config/aarch64/aarch64-sve.md
7961 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7962 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7963 (*extend<mode><Vwide>2): Rename to...
7964 (aarch64_sve_extend<mode><Vwide>2): ...this.
7965 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7966 renaming the old pattern to...
7967 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7969 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7970 define_expand, renaming the old pattern to...
7971 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7972 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7973 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7974 account when deciding which SVE instruction the optab should use.
7975 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7977 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7979 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7980 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7981 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7982 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7983 (tlsdesc_small_<mode>): Turn a define_expand and use
7984 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7985 (tlsdesc_small_advsimd_<mode>): ...this.
7986 (tlsdesc_small_sve_<mode>): New pattern.
7988 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7990 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7991 (UNSPEC_UMUL_HIGHPART): New constants.
7992 (MUL_HIGHPART): New int iteraor.
7993 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
7994 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
7996 (*<su>mul<mode>3_highpart): New define_insn.
7998 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8001 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8004 2018-03-13 Martin Liska <mliska@suse.cz>
8007 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8009 (sem_item_optimizer::~sem_item_optimizer): Release it.
8010 (sem_item_optimizer::merge_classes): Register variable aliases.
8011 (sem_item_optimizer::fixup_pt_set): New function.
8012 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8013 * ipa-icf.h: Declare new variables and functions.
8015 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8018 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8019 integer_pow2p@2 and test integer_pow2p in condition.
8020 (A < 0 ? C : 0): Similarly for @1.
8023 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8024 characters starting at p contain '\0' character, don't look beyond
8028 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8029 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8032 * reg-stack.c (change_stack): Change update_end var from int to
8033 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8034 also call set_block_for_insn on the newly added insns and rescan.
8037 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8038 on the last operand.
8041 * tree.c (stabilize_reference_1): Return save_expr (e) for
8042 STATEMENT_LIST even if it doesn't have side-effects.
8044 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8046 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8048 2018-03-12 Renlin Li <renlin.li@arm.com>
8050 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8051 aarch64_output_scalar_simd_mov_immediate.
8053 2018-03-12 Martin Sebor <msebor@redhat.com>
8055 PR tree-optimization/83456
8056 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8057 for perfectly overlapping calls to memcpy.
8058 (gimple_fold_builtin_memory_chk): Same.
8059 (gimple_fold_builtin_strcpy): Handle no-warning.
8060 (gimple_fold_builtin_stxcpy_chk): Same.
8061 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8063 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8065 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8066 parameter. Use it for SFmode.
8067 (rs6000_function_arg_advance_1): Adjust.
8068 (rs6000_function_arg): Adjust.
8069 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8071 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8073 PR rtl-optimization/84169
8074 PR rtl-optimization/84780
8075 * combine.c (can_combine_p): Check for a 2-insn combination whether
8076 the destination register is used between the two insns, too.
8078 2018-03-12 Richard Biener <rguenther@suse.de>
8080 PR tree-optimization/84803
8081 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8082 for refs DR analysis didn't process.
8084 2018-03-12 Richard Biener <rguenther@suse.de>
8086 PR tree-optimization/84777
8087 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8088 force-vectorize loops ignore whether we are optimizing for size.
8090 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8092 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8093 (TARGET_MD_ASM_ADJUST): Define.
8095 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8096 Kito Cheng <kito.cheng@gmail.com>
8097 Chung-Ju Wu <jasonwucj@gmail.com>
8099 * config/nds32/nds32.c (nds32_compute_stack_frame,
8100 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8101 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8102 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8103 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8104 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8105 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8106 * config/nds32/nds32.md (prologue, epilogue): Use macro
8107 NDS32_V3PUSH_AVAILABLE_P to do checking.
8109 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8112 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8113 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8114 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8115 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8116 addition of most attributes on !orig_type_die or the attribute not
8117 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8119 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8120 Chung-Ju Wu <jasonwucj@gmail.com>
8122 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8124 * config/nds32/nds32.opt (mvh): New option.
8126 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8127 Chung-Ju Wu <jasonwucj@gmail.com>
8129 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8131 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8132 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8135 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8136 Chung-Ju Wu <jasonwucj@gmail.com>
8138 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8140 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8141 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8143 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8144 Kito Cheng <kito.cheng@gmail.com>
8145 Chung-Ju Wu <jasonwucj@gmail.com>
8147 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8148 UNSPEC_FFMISM and UNSPEC_FLMISM.
8149 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8150 for ffb, ffmism and flmism.
8151 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8152 (unspec_ffmism): Ditto.
8153 (unspec_flmism): Ditto.
8154 (nds32_expand_builtin_impl): Check if string extension is available.
8155 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8156 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8158 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8161 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8164 * lra-assigns.c (assign_by_spills): Return a flag of reload
8165 assignment failure. Do not process the reload assignment
8166 failures. Do not spill other reload pseudos if they has the same
8168 (lra_assign): Add a return arg. Set up from the result of
8169 assign_by_spills call.
8170 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8171 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8172 usage_insns if it is not NULL.
8173 (spill_hard_reg_in_range): New function.
8174 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8175 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8176 function prototypes.
8177 (lra_assign): Change prototype.
8178 * lra.c (lra): Add code to deal with fails by splitting hard reg
8181 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8184 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8186 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8189 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8190 (dw_val_node): Add val_symbolic_view.
8191 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8192 (symview_upper_bound): New.
8193 (new_line_info_table): Initialize symviews_since_reset.
8194 (dwarf2out_source_line): Count symviews_since_reset and set
8195 symview_upper_bound.
8196 (dw_val_equal_p): Handle symview.
8197 (add_AT_symview): New.
8198 (print_dw_val): Handle symview.
8199 (attr_checksum, attr_checksum_ordered): Likewise.
8200 (same_dw_val_p, size_of_die): Likewise.
8201 (value_format, output_die): Likewise.
8202 (add_high_low_attributes): Use add_AT_symview for entry_view.
8203 (dwarf2out_finish): Reset symview_upper_bound, clear
8206 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8209 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8210 Add strict argument and use it.
8211 (rs6000_split_multireg_move): Update for new strict argument.
8212 (mem_operand_gpr): Disallow all non-offsettable addresses.
8213 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8215 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8218 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8219 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8220 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8223 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8224 decl, use remap_type if we want to use the type.
8226 2018-03-09 Martin Sebor <msebor@redhat.com>
8228 PR tree-optimization/84526
8229 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8231 (builtin_access::generic_overlap): Be prepared to handle non-array
8234 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8236 PR rtl-optimization/84682
8237 * lra-constraints.c (process_address_1): Check is_address flag
8238 for address constraints.
8239 (process_alt_operands): Likewise.
8240 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8241 preprocess_constraints.
8242 * recog.h (preprocess_constraints): Add oploc parameter.
8244 * recog.c (preprocess_constraints): Test address_operand for
8245 CT_ADDRESS constraints.
8247 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8250 * lra-assigns.c (assign_by_spills): Return a flag of reload
8251 assignment failure. Do not process the reload assignment
8252 failures. Do not spill other reload pseudos if they has the same
8254 (lra_assign): Add a return arg. Set up from the result of
8255 assign_by_spills call.
8256 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8257 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8258 usage_insns if it is not NULL.
8259 (spill_hard_reg_in_range): New function.
8260 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8261 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8262 function prototypes.
8263 (lra_assign): Change prototype.
8264 * lra.c (lra): Add code to deal with fails by splitting hard reg
8267 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8270 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8271 Accept complain bool parameter. Only emit errors if it is true.
8272 (arm_parse_cpu_option_name): Likewise.
8273 (arm_target_thumb_only): Adjust callers of the above.
8274 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8275 prototype to take a default true bool parameter.
8276 (arm_parse_arch_option_name): Likewise.
8278 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8279 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8283 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8284 * configure: Regenerate.
8285 * configure.ac ("linker --version-script option"): New.
8286 ("linker soname option"): New.
8288 2018-03-09 Richard Biener <rguenther@suse.de>
8290 PR tree-optimization/84775
8291 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8292 immediate uses of predicate stmts and mark them modified.
8295 PR tree-optimization/84178
8296 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8298 (version_loop_for_if_conversion): Delay update_ssa call.
8299 (tree_if_conversion): Delay update_ssa until after predicate
8302 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8305 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8306 when the function accesses prior frames.
8308 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8311 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8312 gen_llsym, otherwise call maybe_gen_llsym.
8315 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8316 has ',' character inside of it.
8318 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8321 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8322 as clobbering CC_REGNUM.
8324 2018-03-08 Richard Biener <rguenther@suse.de>
8327 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8328 (follow_copies_to_constant): Do not follow SSA names registered
8331 2018-03-08 Richard Biener <rguenther@suse.de>
8333 PR tree-optimization/84178
8334 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8336 (version_loop_for_if_conversion): Delay update_ssa call.
8337 (tree_if_conversion): Delay update_ssa until after predicate
8340 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8342 PR tree-optimization/84178
8343 * tree-if-conv.c (release_bb_predicate): Remove the
8344 the assertion that the stmts have NULL use_ops.
8345 Discard the statements, asserting that they haven't
8346 yet been added to a BB.
8348 2018-03-08 Richard Biener <rguenther@suse.de>
8350 PR tree-optimization/84746
8351 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8352 (phi_translate): Pass in destination ANTIC_OUT set.
8353 (phi_translate_1): Likewise. For a simplified result lookup
8354 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8355 (phi_translate_set): Adjust.
8356 (do_pre_regular_insertion): Likewise.
8357 (do_pre_partial_partial_insertion): Likewise.
8359 2018-03-08 Martin Liska <mliska@suse.cz>
8361 PR gcov-profile/84735
8362 * doc/gcov.texi: Document usage of profile files.
8363 * gcov-io.h: Document changes in the format.
8365 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8369 * dwarf2out.c (struct dw_line_info_table): Update comments for
8371 (FORCE_RESET_NEXT_VIEW): New.
8372 (FORCE_RESETTING_VIEW_P): New.
8373 (RESETTING_VIEW_P): Check for -1 too.
8374 (ZERO_VIEW_P): Likewise.
8375 (new_line_info_table): Force-reset next view.
8376 (dwarf2out_begin_function): Likewise.
8377 (dwarf2out_source_line): Simplify zero_view_p initialization.
8378 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8379 view directly. Omit view when omitting .loc at line 0.
8381 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8383 PR tree-optimization/84740
8384 * tree-switch-conversion.c (process_switch): Call build_constructors
8385 only if info.phi_count is non-zero.
8387 PR tree-optimization/84739
8388 * tree-tailcall.c (find_tail_calls): Check call arguments against
8389 DECL_ARGUMENTS (current_function_decl) rather than
8390 DECL_ARGUMENTS (func) when checking for tail recursion.
8392 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8394 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8395 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8396 Volker Reichelt's entry and add entries for people that perform
8397 GCC fuzzy testing and report numerous bugs.
8399 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8402 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8403 readonly data in sdata, if that is disabled.
8404 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8405 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8406 -mreadonly-in-sdata option.
8408 2018-03-07 Martin Sebor <msebor@redhat.com>
8410 PR tree-optimization/84468
8411 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8412 basic block when looking for nul assignment.
8414 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8417 * except.h (output_function_exception_table): Adjust prototype.
8418 * except.c (output_function_exception_table): Remove FNNAME parameter
8419 and add SECTION parameter. Ouput one part of the table at a time.
8420 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8421 the first part of the exception table and emit unwind directives.
8422 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8423 (i386_pe_seh_cold_init): Likewise.
8424 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8425 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8426 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8427 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8428 * config/i386/winnt.c: Include except.h.
8429 (struct seh_frame_state): Add reg_offset, after_prologue and
8430 in_cold_section fields.
8431 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8432 (i386_pe_seh_cold_init): New function.
8433 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8434 to seh->in_cold_section.
8435 (seh_emit_push): Record the offset of the push.
8436 (seh_emit_save): Record the offet of the save.
8437 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8438 Test seh->after_prologue to disregard the epilogue.
8439 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8440 (i386_pe_end_cold_function): New function.
8442 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8445 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8446 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8449 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8451 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8452 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8455 * multiple_target.c: Include tree-inline.h and intl.h.
8456 (expand_target_clones): Diagnose and fail if node->definition and
8457 !tree_versionable_function_p (node->decl).
8459 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8461 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8463 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8464 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8465 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8467 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8470 * combine.c (try_combine): Use reg_or_subregno instead of handling
8471 just paradoxical SUBREGs and REGs.
8473 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8475 * config/arc/arc.c (arc_finalize_pic): Remove function.
8476 (arc_must_save_register): We use single base PIC register, remove
8477 checks to save/restore the PIC register.
8478 (arc_expand_prologue): Likewise.
8479 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8481 (arc_verify_short): Likewise.
8482 (arc_attr_type): Likewise.
8483 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8484 (walk_stores): Likewise.
8485 (arc_address_cost): Make it static.
8486 (arc_verify_short): Likewise.
8487 (branch_dest): Likewise.
8488 (arc_attr_type): Likewise.
8489 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8490 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8491 (arc_final_prescan_insn): Remove inserting the nops due to
8492 hardware hazards. It is done in reorg step.
8493 (insn_length_variant_t): Remove.
8494 (insn_length_parameters_t): Likewise.
8495 (arc_insn_length_parameters): Likewise.
8496 (arc_get_insn_variants): Likewise.
8497 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8499 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8502 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8505 PR tree-optimization/84687
8506 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8508 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8510 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8512 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8513 Rename to ppc_speculation_barrier.
8514 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8515 __builtin_ppc_speculation_barrier.
8517 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8520 * combine.c (combine_simplify_rtx): Don't try to simplify if
8521 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8524 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8526 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8527 to 32 bytes when compiling for POWER9.
8529 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8532 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8533 regparm >= 3 with no arg reg available also for calls with
8534 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8537 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8539 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8541 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8544 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8546 2018-03-05 Richard Biener <rguenther@suse.de>
8548 PR tree-optimization/84486
8549 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8550 When inserting a __builtin_assume_aligned call set the LHS
8551 SSA name alignment info accordingly.
8553 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8555 PR tree-optimization/84114
8556 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8557 Avoid reassociation of FLOAT_MODE addition.
8559 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8561 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8562 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8563 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8564 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8565 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8566 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8567 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8569 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8570 __builtin_ia32_wbinvd): New builtins.
8571 (SPECIAL_ARGS2): New.
8572 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8573 (SPECIAL_ARGS2): New.
8574 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8575 (ix86_valid_target_attribute_inner_p): Ditto.
8576 (ix86_init_mmx_sse_builtins): Add special_args2.
8577 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8578 TARGET_WBNOINVD_P): New.
8579 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8580 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8581 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8582 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8583 * config/i386/pconfigintrin.h: New file.
8584 * config/i386/wbnoinvdintrin.h: Ditto.
8585 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8587 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8589 2018-03-05 Richard Biener <rguenther@suse.de>
8591 PR tree-optimization/84670
8592 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8594 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8595 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8596 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8597 doesn't grow if all successors (recursively) were visited at least
8600 2018-03-05 Richard Biener <rguenther@suse.de>
8602 PR tree-optimization/84650
8603 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8604 if executed in the loop pipeline.
8606 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8608 * doc/configfiles.texi (Configuration Files): Move info about
8609 conditionalizing $target-protos.h to...
8610 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8611 differs from $target-protos.h.
8613 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8614 Chung-Ju Wu <jasonwucj@gmail.com>
8616 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8617 * config/nds32/nds32-multiple.md (setmemsi): Define.
8618 * config/nds32/nds32-memory-manipulation.c
8619 (nds32_gen_dup_4_byte_to_word_value): New.
8620 (emit_setmem_word_loop): New.
8621 (emit_setmem_byte_loop): New.
8622 (nds32_expand_setmem_loop): New.
8623 (nds32_expand_setmem_loop_v3m): New.
8624 (nds32_expand_setmem_unroll): New.
8625 (nds32_expand_setmem): New.
8627 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8628 Chung-Ju Wu <jasonwucj@gmail.com>
8630 * config/nds32/nds32-memory-manipulation.c
8631 (nds32_emit_load_store): New.
8632 (nds32_emit_post_inc_load_store): New.
8633 (nds32_emit_mem_move): New.
8634 (nds32_emit_mem_move_block): New.
8635 (nds32_expand_movmemsi_loop_unknown_size): New.
8636 (nds32_expand_movmemsi_loop_known_size): New.
8637 (nds32_expand_movmemsi_loop): New.
8638 (nds32_expand_movmemsi_unroll): New.
8639 (nds32_expand_movmemqi): Rename ...
8640 (nds32_expand_movmemsi): ... to this.
8641 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8642 (movmemsi): ... to this.
8643 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8644 (nds32_expand_movmemsi): ... to this.
8646 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8647 Monk Chiang <sh.chiang04@gmail.com>
8648 Chung-Ju Wu <jasonwucj@gmail.com>
8650 * config/nds32/nds32-protos.h
8651 (nds32_expand_load_multiple): New arguments.
8652 (nds32_expand_store_multiple): Ditto.
8653 (nds32_valid_multiple_load_store): Rename ...
8654 (nds32_valid_multiple_load_store_p): ... to this.
8655 * config/nds32/nds32-memory-manipulation.c
8656 (nds32_expand_load_multiple): Refine implementation.
8657 (nds32_expand_store_multiple): Ditto.
8658 * config/nds32/nds32-multiple.md
8659 (load_multiple): Update nds32_expand_load_multiple interface.
8660 (store_multiple): Update nds32_expand_store_multiple interface.
8661 * config/nds32/nds32-predicates.c
8662 (nds32_valid_multiple_load_store): Rename ...
8663 (nds32_valid_multiple_load_store_p): ... to this and refine
8665 * config/nds32/predicates.md
8666 (nds32_load_multiple_and_update_address_operation): New predicate.
8667 (nds32_store_multiple_and_update_address_operation): New predicate.
8669 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8670 Chung-Ju Wu <jasonwucj@gmail.com>
8672 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8673 (combo): New attribute.
8674 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8676 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8678 * config/nds32/nds32.opt: Change -mcmodel= default value.
8680 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8681 Monk Chiang <sh.chiang04@gmail.com>
8682 Chung-Ju Wu <jasonwucj@gmail.com>
8684 * config/nds32/constants.md (unspec_element): New enum.
8685 * config/nds32/constraints.md (Umw): New constraint.
8686 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8687 * config/nds32/nds32-intrinsic.md: Likewise.
8688 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8689 (nds32_valid_smw_lwm_base_p): New.
8690 (nds32_output_smw_single_word): New.
8691 (nds32_output_lmw_single_word): New.
8692 (nds32_expand_unaligned_load): New.
8693 (nds32_expand_unaligned_store): New.
8694 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8695 (nds32_output_smw_single_word): Declare.
8696 (nds32_output_lmw_single_word): Declare.
8697 (nds32_expand_unaligned_load): Declare.
8698 (nds32_expand_unaligned_store): Declare.
8699 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8700 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8701 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8702 NDS32_BUILTIN_UASTORE_DW.
8703 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8706 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8707 Kito Cheng <kito.cheng@gmail.com>
8708 Chung-Ju Wu <jasonwucj@gmail.com>
8710 * config/nds32/nds32-intrinsic.c
8711 (nds32_expand_builtin_null_ftype_reg): Delete.
8712 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8713 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8714 (nds32_read_argument): New.
8715 (nds32_legitimize_target): Ditto.
8716 (nds32_legitimize_argument): Ditto.
8717 (nds32_check_constant_argument): Ditto.
8718 (nds32_expand_unop_builtin): Ditto.
8719 (nds32_expand_unopimm_builtin): Ditto.
8720 (nds32_expand_binop_builtin): Ditto.
8721 (nds32_builtin_decl_impl): Ditto.
8722 (builtin_description): Ditto.
8723 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8724 (nds32_init_builtins_impl): Ditto.
8725 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8726 (nds32_builtin_decl): New.
8727 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8728 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8730 2018-03-02 Jeff Law <law@redhat.com>
8732 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8733 (redundant_insn, fill_simple_delay_slots): Likewise.
8734 (fill_slots_from_thread): Likewise.
8735 * resource.c (mark_referenced_resources): Likewise.
8736 (mark_set_resources, find_dead_or_set_registers): Likewise.
8738 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8740 * substring-locations.h (format_warning_va): Formatting fix for
8742 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8744 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8745 * substring-locations.c: Include intl.h.
8746 (format_warning_va): Turned into small wrapper around
8747 format_warning_n_va, renamed to ...
8748 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8749 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8751 (format_warning_at_substring_n): New function.
8752 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8753 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8754 format_warning_at_substring with just a shorter name instead of
8755 const function pointer.
8756 (fmtwarn_n): New function.
8757 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8758 appropriate, get rid of all the fmtstr temporaries, move conditionals
8759 with G_() wrapped string literals directly into fmtwarn arguments,
8760 cast dir.len to (int), formatting fixes.
8762 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8764 * doc/invoke.texi: Remove "Cilk Plus" references.
8766 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8767 Richard Biener <rguenther@suse.de>
8770 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8771 for error or warning attributes if CALL_FROM_THUNK_P is set.
8774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8777 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8778 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8781 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8782 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8784 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8785 instead of -1U in last predictors element's probability member.
8787 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8790 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8791 arguments if they are comparable.
8793 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8795 PR tree-optimization/84634
8796 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8797 masks and masked_loop_p with a single loop_masks, making sure it's
8798 null for bb vectorization.
8800 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8802 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8803 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8804 loop->force_vectorize to check whether there is no alias.
8806 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8809 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8811 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8813 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8814 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8815 instead of a loop around prev_real_insn.
8816 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8820 * config/i386/i386.c (ix86_print_operand): Use conditional
8821 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8824 2018-03-02 Richard Biener <rguenther@suse.de>
8826 PR tree-optimization/84427
8827 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8828 (bitmap_set_subtract_values): Rewrite to handle multiple
8831 (prune_clobbered_mems): Likewise.
8832 (phi_translate): Take edge instead of pred/phiblock.
8833 (phi_translate_1): Likewise.
8834 (phi_translate_set): Likewise. Insert all translated
8835 exprs for a value into the set, keeping possibly multiple
8836 expressions per value.
8837 (compute_antic_aux): Adjust for phi_translate changes.
8838 When intersecting union the expressions and prune those
8839 not in the final value set, keeping possibly multiple
8840 expressions per value. Do not use value-insertion
8841 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8842 all expressions. Add verification that the value-sets
8843 only shrink during iteration.
8844 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8845 (do_pre_regular_insertion): Likewise.
8846 (do_pre_partial_partial_insertion): Likewise.
8848 2018-03-02 Richard Biener <rguenther@suse.de>
8851 * config/darwin.c (saved_debug_info_level): New static global.
8852 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8853 (darwin_asm_lto_end): Restore debug info generation settings.
8855 2018-03-01 Martin Liska <mliska@suse.cz>
8858 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8861 2018-03-01 Richard Biener <rguenther@suse.de>
8864 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8865 type annotation with LTO when debug was disabled at compile-time.
8867 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8869 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8871 (mips_final_postscan_insn): Likewise.
8873 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8875 PR rtl-optimization/84528
8876 * alias.c (init_alias_target): Add commentary.
8877 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8878 a unique base value if the frame pointer is not eliminated
8879 to the stack pointer.
8881 2018-03-01 Tom de Vries <tom@codesourcery.com>
8883 PR rtl-optimization/83327
8884 * lra-int.h (hard_regs_spilled_into): Declare.
8885 * lra.c (hard_regs_spilled_into): Define.
8886 (init_reg_info): Init hard_regs_spilled_into.
8887 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8888 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8889 (process_bb_lives): Handle hard_regs_spilled_into.
8890 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8891 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8893 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8895 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8896 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8897 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8898 * config/rs6000/aix72.h: New file.
8900 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8902 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8903 instead of warning_at with conditional singular and plural messages
8907 * stor-layout.c (update_alignment_for_field): For
8908 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8909 && !DECL_PACKED (field), do the alignment update, just use
8910 only desired_align instead of MAX (type_align, desired_align)
8912 (place_field): Don't do known_align < desired_align handling
8913 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8914 is non-NULL, instead do it after rli->prev_field handling and
8915 only if not within a bitfield word. For DECL_PACKED (field)
8916 use type_align of BITS_PER_UNIT.
8918 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8920 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8921 superfluous parentheses and trailing spaces.
8923 2018-02-28 Richard Biener <rguenther@suse.de>
8925 PR tree-optimization/84584
8926 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8927 SCoPs with fake exit edge.
8929 2018-02-28 Martin Liska <mliska@suse.cz>
8932 * timevar.c (timer::print): Fix format to properly print 100%
8935 2018-02-28 Richard Biener <rguenther@suse.de>
8938 * genmatch.c (capture_info::walk_match): Do not mark
8939 captured expressions without operands as expr_p given
8940 they act more like predicates and should be subject to
8941 "lost tail" side-effect preserving.
8943 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8945 PR rtl-optimization/81611
8946 * auto-inc-dec.c (attempt_change): Move dead note from
8947 mem_insn if it's the next use of regno
8948 (find_address): Take address use of reg holding
8949 non-incremented value. Add parm to limit search to the named
8951 (merge_in_block): Attempt to use a mem insn that is the next
8952 use of the original regno.
8954 2018-02-27 Martin Sebor <msebor@redhat.com>
8957 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8958 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8960 2018-02-27 Martin Sebor <msebor@redhat.com>
8962 PR translation/84207
8963 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8964 n argument to unsigned HOST_WIDE_INT.
8965 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8966 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8967 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8970 2018-02-27 Richard Biener <rguenther@suse.de>
8972 PR tree-optimization/84512
8973 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8974 Do not use the estimate returned from record_stmt_cost for
8975 the scalar iteration cost but sum properly using add_stmt_cost.
8977 2018-02-27 Richard Biener <rguenther@suse.de>
8979 PR tree-optimization/84466
8980 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8981 Adjust last change to less strictly validate use operands.
8983 2018-02-27 Martin Liska <mliska@suse.cz>
8985 PR gcov-profile/84548
8986 * gcov.c (process_file): Allow partial overlap and consider it
8987 also as group functions.
8988 (output_lines): Properly calculate range of lines for a group.
8990 2018-02-27 Martin Liska <mliska@suse.cz>
8992 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
8993 'ggc' suffixes. Change first column width.
8994 (timer::print): Fix formatting of the column.
8996 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
8998 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
8999 preserve inline entry blocks for the sake of debug inline
9000 entry point markers alone.
9001 (remove_unused_locals): Suggest in comments a better place to
9002 force the preservation of inline entry blocks that are
9003 otherwise unused, but do not preserve them.
9005 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9007 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9009 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9012 * config/i386/constraints.md (Bs): Replace
9013 ix86_indirect_branch_register with
9014 TARGET_INDIRECT_BRANCH_REGISTER.
9016 * config/i386/i386.md (indirect_jump): Likewise.
9017 (tablejump): Likewise.
9018 (*sibcall_memory): Likewise.
9019 (*sibcall_value_memory): Likewise.
9020 Peepholes of indirect call and jump via memory: Likewise.
9021 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9022 (*sibcall_value_GOT_32): Likewise.
9023 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9024 (GOT_memory_operand): Likewise.
9025 (call_insn_operand): Likewise.
9026 (sibcall_insn_operand): Likewise.
9027 (GOT32_symbol_operand): Likewise.
9028 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9030 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9032 PR rtl-optimization/83496
9033 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9034 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9035 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9036 redundant insn, if any.
9037 (relax_delay_slots): Likewise.
9038 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9040 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9042 PR tree-optimization/83965
9043 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9044 that grouped statements are part of a reduction chain. Return
9045 true if the statement is not marked as a reduction itself but
9047 (vect_recog_dot_prod_pattern): Don't check whether the statement
9048 is part of a group here.
9049 (vect_recog_sad_pattern): Likewise.
9050 (vect_recog_widen_sum_pattern): Likewise.
9052 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9055 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9057 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9060 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9062 (ix86_output_indirect_function_return): New prototype.
9063 (ix86_split_simple_return_pop_internal): Likewise.
9064 * config/i386/i386.c (indirect_return_via_cx): New.
9065 (indirect_return_via_cx_bnd): Likewise.
9066 (indirect_thunk_name): Handle return va CX_REG.
9067 (output_indirect_thunk_function): Create alias for
9068 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9069 (ix86_output_indirect_jmp): Remove the bool argument.
9070 (ix86_output_indirect_function_return): New function.
9071 (ix86_split_simple_return_pop_internal): Likewise.
9072 * config/i386/i386.md (*indirect_jump): Don't pass false
9073 to ix86_output_indirect_jmp.
9074 (*tablejump_1): Likewise.
9075 (simple_return_pop_internal): Change it to define_insn_and_split.
9076 Call ix86_split_simple_return_pop_internal to split it for
9078 (simple_return_indirect_internal): Call
9079 ix86_output_indirect_function_return instead of
9080 ix86_output_indirect_jmp.
9082 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9085 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9086 memset and value initialization afterwards.
9088 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9090 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9092 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9095 * common/config/aarch64/aarch64-common.c
9096 (aarch_option_optimization_table[]): Switch
9097 off fomit-frame-pointer
9099 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9100 Chung-Ju Wu <jasonwucj@gmail.com>
9102 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9104 (store_multiple): Ditto.
9106 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9108 * config.gcc: Add --with-cpu support for nds32 target.
9109 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9110 * config/nds32/nds32.opt: Add -mcpu= option.
9112 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9114 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9115 isel=yes): Warn for these deprecated options.
9117 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9119 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9120 ISA_2_5_MASKS_EMBEDDED.
9122 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9124 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9125 p->max as pointers rather than using iterative_hash_expr.
9127 2018-02-23 Carl Love <cel@us.ibm.com>
9129 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9130 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9132 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9133 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9134 P8V_BUILTIN_VEC_VUNSIGNED2.
9136 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9139 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9140 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9142 * lra-constraints.c (curr_insn_transform): Set up
9143 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9145 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9146 * lra-eliminations.c (spill_pseudos): Ditto.
9147 (process_insn_for_elimination): Ditto.
9148 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9149 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9150 LRA_NON_CLOBBERED_ALT.
9152 2018-02-22 Martin Sebor <msebor@redhat.com>
9154 PR tree-optimization/84480
9155 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9156 to maybe_diag_stxncpy_trunc. Call it.
9157 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9158 from gimple_fold_builtin_strcpy. Print inlining stack.
9159 (handle_builtin_stxncpy): Print inlining stack.
9160 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9162 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9165 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9166 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9167 and -fcheck-pointer-bounds are used together.
9168 (indirect_thunk_prefix): New enum.
9169 (indirect_thunk_need_prefix): New function.
9170 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9171 "_nt" instead of "_bnd" for NOTRACK prefix.
9172 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9173 (output_indirect_thunk_function): Likewise.
9175 (ix86_code_end): Update output_indirect_thunk_function calls.
9176 (ix86_output_indirect_branch_via_reg): Replace
9177 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9178 (ix86_output_indirect_branch_via_push): Likewise.
9179 (ix86_output_function_return): Likewise.
9180 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9181 incompatible with -fcf-protection=branch and
9182 -fcheck-pointer-bounds.
9184 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9187 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9188 Change gcc_assert call to output_operand_lossage.
9190 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9192 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9194 2018-02-22 DJ Delorie <dj@redhat.com>
9195 Sebastian Perta <sebastian.perta@renesas.com>
9196 Oleg Endo <olegendo@gcc.gnu.org>
9198 * config/rx/rx.c (rx_rtx_costs): New function.
9199 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9201 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9203 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9205 2018-02-22 Martin Liska <mliska@suse.cz>
9208 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9209 Add "native" as a possible value.
9211 2018-02-22 Martin Liska <mliska@suse.cz>
9214 * config/i386/i386.c (ix86_option_override_internal):
9215 Add "native" as a possible value for -march and -mtune.
9217 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9220 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9221 to all type variants.
9223 PR tree-optimization/84503
9224 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9225 width as info->bitpos + info->bitsize - start.
9226 (merged_store_group::merge_overlapping): Simplify width computation.
9227 (check_no_overlap): New function.
9228 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9229 start + width and last_order of the group, fail if check_no_overlap
9231 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9232 to group if check_no_overlap fails.
9234 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9236 * config/rs6000/altivec.md: Delete contraint arguments to
9237 define_expand, define_split, and define_peephole2, and in
9238 define_insn_and_split if always unused.
9239 * config/rs6000/darwin.md: Ditto.
9240 * config/rs6000/dfp.md: Ditto.
9241 * config/rs6000/rs6000.md: Ditto.
9242 * config/rs6000/sync.md: Ditto.
9243 * config/rs6000/vector.md: Ditto.
9244 * config/rs6000/vsx.md: Ditto.
9246 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9248 * config/rs6000/altivec.md: Write output control strings as braced
9249 blocks instead of double-quoted strings.
9250 * config/rs6000/darwin.md: Ditto.
9251 * config/rs6000/rs6000.md: Ditto.
9252 * config/rs6000/vector.md: Ditto.
9253 * config/rs6000/vsx.md: Ditto.
9255 2018-02-21 Jason Merrill <jason@redhat.com>
9257 PR c++/84314 - ICE with templates and fastcall attribute.
9258 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9260 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9262 * ipa-cp.c (determine_versionability): Fix comment typos.
9264 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9267 * ipa-cp.c (determine_versionability): Do not version functions caling
9270 2018-02-21 Martin Liska <mliska@suse.cz>
9273 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9274 Add "native" as a possible value.
9275 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9276 the macro when native cpu detection is available.
9278 2018-02-21 Martin Liska <mliska@suse.cz>
9281 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9282 Add "native" as a possible value.
9283 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9284 when native cpu detection is available.
9286 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9287 Martin Sebor <msebor@redhat.com>
9289 PR tree-optimization/84478
9290 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9292 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9293 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9294 support which is conservatively correct, for 2 only stay conservative
9295 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9296 argument to the 2 argument get_range_strlen, adjust 6 arg
9297 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9299 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9300 (gimple_fold_builtin_strlen): Pass true as last argument to
9303 2018-02-20 Martin Sebor <msebor@redhat.com>
9306 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9307 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9308 (builtin_memref::builtin_memref): Factor out parts into
9309 set_base_and_offset and call it.
9311 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9314 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9315 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9316 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9317 search at the associated MODE_INT.
9319 2018-02-20 Jeff Law <law@redhat.com>
9322 PR tree-optimization/81592
9324 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9325 for range data rather than using global data.
9326 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9327 range data rather than using global data.
9328 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9329 pass it to children as needed.
9330 (struct directive::fmtresult): Similarly.
9331 (struct directive::set_width): Similarly.
9332 (struct directive::set_precision): Similarly.
9333 (format_integer, format_directive, parse_directive): Similarly.
9334 (format_none): Accept unnamed vr_values parameter.
9335 (format_percent, format_floating, format_character): Similarly.
9336 (format_string, format_plain): Similarly.
9337 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9338 the EVRP range analyzer for range data rather than using global data.
9339 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9340 gimple-ssa-evrp-analyze.h
9341 (class sprintf_dom_walker): Add after_dom_children member function.
9342 Add evrp_range_analyzer member.
9343 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9344 range analyzer as needed.
9345 (sprintf_dom_walker::after_dom_children): New member function.
9346 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9348 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9349 (evrp_range_analyzer::pop_to_marker): Likewise.
9351 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9353 PR tree-optimization/84419
9354 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9355 with the required type if its current type is compatible but
9358 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9361 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9362 after vectorization.
9364 2018-02-20 Martin Liska <mliska@suse.cz>
9367 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9368 possible values if we don't have a hint.
9370 2018-02-20 Martin Liska <mliska@suse.cz>
9374 * final.c (shorten_branches): Build align_tab array with one
9376 * opts.c (finish_options): Add alignment option limit check.
9377 (MAX_CODE_ALIGN): Likewise.
9378 (MAX_CODE_ALIGN_VALUE): Likewise.
9379 * doc/invoke.texi: Document maximum allowed option value for
9380 all -falign-* options.
9382 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9385 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9386 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9387 * var-tracking.c (emit_note_insn_var_location): Remove all references
9388 to NOTE_INSN_CALL_ARG_LOCATION.
9389 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9390 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9391 Use copy_rtx_if_shared.
9392 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9393 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9394 (dwarf2out_var_location): Remove handling of
9395 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9397 * final.c (final_scan_insn): Remove all references to
9398 NOTE_INSN_CALL_ARG_LOCATION.
9399 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9400 before dumping final insns.
9401 * except.c (emit_note_eh_region_end): Remove all references to
9402 NOTE_INSN_CALL_ARG_LOCATION.
9403 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9404 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9405 * config/arc/arc.c (hwloop_optimize): Likewise.
9406 * config/arm/arm.c (create_fix_barrier): Likewise.
9407 * config/s390/s390.c (s390_chunkify_start): Likewise.
9408 * config/sh/sh.c (find_barrier): Likewise.
9409 * config/i386/i386.c (rest_of_insert_endbranch,
9410 ix86_seh_fixup_eh_fallthru): Likewise.
9411 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9412 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9413 * config/frv/frv.c (frv_function_prologue): Likewise.
9414 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9416 (note_outside_basic_block_p): Remove all references to
9417 NOTE_INSN_CALL_ARG_LOCATION.
9418 * gengtype.c (adjust_field_rtx_def): Likewise.
9419 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9421 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9422 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9425 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9428 PR tree-optimization/84452
9429 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9430 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9433 2018-02-19 Martin Liska <mliska@suse.cz>
9436 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9438 2018-02-19 Martin Liska <mliska@suse.cz>
9439 Richard Sandiford <richard.sandiford@linaro.org>
9441 PR tree-optimization/82491
9442 * gimple-fold.c (get_base_constructor): Make earlier bail out
9445 2018-02-19 Carl Love <cel@us.ibm.com>
9447 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9448 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9450 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9451 P8V_BUILTIN_VEC_NEG.
9453 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9455 * config/rl78/rl78.md (movdf): New define expand.
9457 2018-02-19 Martin Liska <mliska@suse.cz>
9460 * doc/invoke.texi: Fix typo.
9461 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9463 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9465 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9466 handle rs6000_single_float and rs6000_double_float specially for
9469 2018-02-16 Jeff Law <law@redhat.com>
9471 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9474 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9477 * ipa-inline.c (inline_small_functions): Fix a typo.
9479 2018-02-16 Nathan Sidwell <nathan@acm.org>
9481 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9483 2018-02-16 Carl Love <cel@us.ibm.com>
9485 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9486 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9487 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9488 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9489 expansion to P8V_BUILTIN_VEC_FLOAT2.
9491 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9493 PR rtl-optimization/70023
9494 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9495 src_regno into account.
9497 2018-02-16 Carl Love <cel@us.ibm.com>
9499 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9500 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9501 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9502 * config/rs6000/rs6000.c: Remove case statements for
9503 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9504 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9505 and P9V_BUILTIN_VEC_VINSERT4B.
9506 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9507 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9508 * config/rs6000/vsx.md:
9509 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9512 2018-02-16 Carl Love <cel@us.ibm.com>
9514 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9516 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9518 * config/rs6000/rs6000-c.c: Add the definitions for
9519 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9520 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9521 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9522 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9523 definition for insert4b and define insn *insert3b_internal.
9524 * doc/extend.texi: Add documentation for vec_extract4b.
9526 2018-02-16 Nathan Sidwell <nathan@acm.org>
9528 * doc/extend.texi (Backwards Compatibility): Mention friend
9529 injection. Note for-scope is deprecated.
9530 * doc/invoke.texi (-ffriend-injection): Deprecate.
9532 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9534 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9535 that moved to I2, also allow destinations that are a paradoxical
9536 subreg (instead of a normal reg).
9538 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9541 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9544 2018-02-16 Richard Biener <rguenther@suse.de>
9546 PR tree-optimization/84037
9547 PR tree-optimization/84016
9549 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9550 Adjust vec_construct for the fact we need additional higher latency
9551 128bit inserts for AVX256 and AVX512 vector builds.
9552 (ix86_add_stmt_cost): Scale vector construction cost for
9555 2018-02-16 Richard Biener <rguenther@suse.de>
9557 PR tree-optimization/84417
9558 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9559 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9560 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9562 2018-02-16 Martin Liska <mliska@suse.cz>
9565 * internal-fn.def (ASAN_CHECK): Set proper flags.
9566 (ASAN_MARK): Likewise.
9568 2018-02-16 Julia Koval <julia.koval@intel.com>
9570 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9571 from PTA_CANNONLAKE.
9573 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9576 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9577 Use ++iter rather than iter++ for std::list iterators.
9578 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9579 defer deleting them until all nodes in the forest are processed. Do
9580 free even leaf nodes. Change to_process into auto_vec.
9583 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9584 * vec.h (vec_default_construct): Use memset instead of placement new
9585 if BROKEN_VALUE_INITIALIZATION is defined.
9586 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9587 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9590 PR rtl-optimization/83723
9591 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9592 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9593 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9595 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9597 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9599 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9601 PR rtl-optimization/81443
9602 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9603 from inner REGs to paradoxical SUBREGs.
9605 2018-02-16 Richard Biener <rguenther@suse.de>
9607 PR tree-optimization/84399
9608 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9609 For operands we can analyze at their definition make sure we can
9610 analyze them at each use as well.
9612 2018-02-16 Richard Biener <rguenther@suse.de>
9614 PR tree-optimization/84190
9615 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9616 volatile accesses if the decl isn't volatile.
9618 2018-02-15 Jason Merrill <jason@redhat.com>
9620 PR c++/84314 - ICE with templates and fastcall attribute.
9621 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9622 TYPE_CANONICAL on an existing type.
9624 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9626 PR tree-optimization/84383
9627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9628 dstoff nor call operand_equal_p if dstbase is NULL.
9630 PR tree-optimization/84334
9631 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9632 also a CONSTANT_CLASS_P, punt.
9634 2018-02-14 Jim Wilson <jimw@sifive.com>
9636 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9637 first SMALL_OPERAND check. New local min_second_step. Move assert
9638 to where locals are set. Add TARGET_RVC support.
9639 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9641 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9643 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9645 2018-02-14 Martin Sebor <msebor@redhat.com>
9647 PR tree-optimization/83698
9648 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9649 arrays constrain the offset range to their bounds.
9650 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9651 (builtin_access::overlap): Avoid setting the size of overlap if it's
9653 (maybe_diag_overlap): Also consider arrays when deciding what values
9654 of offsets to include in diagnostics.
9656 2018-02-14 Martin Sebor <msebor@redhat.com>
9659 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9660 that correspond to the kind of a declaration.
9662 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9665 * config/pa/pa.md: Load address of PIC label using the linkage table
9666 if the label is nonlocal.
9668 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9670 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9671 warning message if user requests -maltivec=be.
9672 * doc/invoke.texi: Document deprecation of -maltivec=be.
9674 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9677 * config/rs6000/rs6000-c.c: Update definitions for
9678 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9679 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9681 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9684 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9685 add _get_ssp intrinsics. Remove argument from
9686 __builtin_ia32_rdssp[d|q].
9687 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9688 * config/i386/i386-builtin.def: Remove argument from
9689 __builtin_ia32_rdssp[d|q].
9690 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9691 ix86_expand_special_args_builtin for _rdssp[d|q].
9692 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9693 Clear register before usage.
9694 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9695 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9697 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9699 PR tree-optimization/84357
9700 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9701 operand 1 of an ARRAY_REF too.
9703 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9706 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9707 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9709 (set_of_reg): New struct.
9710 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9711 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9712 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9714 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9715 Split into bitclr, bitset, bitinvert patterns if appropriate.
9716 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9717 use rx_fuse_in_memory_bitop.
9718 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9719 to named insn, correct maximum insn length.
9721 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9724 * machmode.def: Define a complex mode for PARTIAL_INT.
9725 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9727 * doc/rtl.texi: Document CSPImode.
9728 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9730 (msp430_hard_regno_nregs_with_padding): Likewise.
9732 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9735 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9737 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9739 PR rtl-optimization/84169
9740 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9741 we generated a parallel as new i3 and we split that to new i2 and i3
9742 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9743 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9744 those to i2, not i1. Partially rewrite this scan code.
9746 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9749 * stor-layout.c (place_field): For variable length fields, adjust
9750 offset_align afterwards not just based on the field's alignment,
9751 but also on the size.
9754 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9755 of exps and logs in the use_exp2 case.
9757 2018-02-13 Jeff Law <law@redhat.com>
9759 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9762 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9765 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9769 * common.opt (gas-loc-support, gas-locview-support): New.
9770 (ginline-points, ginternal-reset-location-views): New.
9771 * doc/invoke.texi: Document them. Use @itemx where intended.
9772 (gvariable-location-views): Adjust.
9773 * target.def (reset_location_view): New.
9774 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9775 (TARGET_RESET_LOCATION_VIEW): New.
9776 * doc/tm.texi: Rebuilt.
9777 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9778 (dwarf2out_default_as_locview_support): New.
9779 (output_asm_line_debug_info): Use option variables.
9780 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9781 (output_loc_list): Likewise.
9782 (add_high_low_attributes): Check option variables.
9783 Don't output entry view attribute in strict mode.
9784 (gen_inlined_subroutine_die): Check option variables.
9785 (dwarf2out_inline_entry): Likewise.
9786 (init_sections_and_labels): Likewise.
9787 (dwarf2out_early_finish): Likewise.
9788 (maybe_reset_location_view): New, from...
9789 (dwarf2out_var_location): ... here. Call it.
9790 * debug.h (dwarf2out_default_as_loc_support): Declare.
9791 (dwarf2out_default_as_locview_support): Declare.
9792 * hooks.c (hook_int_rtx_insn_0): New.
9793 * hooks.h (hook_int_rtx_insn_0): Declare.
9794 * toplev.c (process_options): Take -gas-loc-support and
9795 -gas-locview-support from dwarf2out. Enable
9796 -gvariable-location-views by default only with locview
9797 assembler support. Enable -ginternal-reset-location-views by
9798 default only if the target defines the corresponding hook.
9799 Enable -ginline-points by default if location views are
9800 enabled; force it disabled if statement frontiers are
9802 * tree-inline.c (expand_call_inline): Check option variables.
9803 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9805 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9807 PR tree-optimization/84321
9808 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9809 handling. Also check whether the anti-range contains any values
9810 that satisfy the mask; switch to a VR_RANGE if not.
9812 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9815 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9817 2018-02-13 Martin Jambor <mjambor@suse.cz>
9820 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9821 of call statements, also set location of a load to a temporary.
9823 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9825 * config/rl78/rl78.c (add_vector_labels): New function.
9826 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9827 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9828 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9829 which checks that no arguments are passed.
9830 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9831 * doc/extend.texi: Documentation for the new attribute.
9833 2018-02-13 Andreas Schwab <schwab@suse.de>
9835 * config/riscv/linux.h (CPP_SPEC): Define.
9837 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9840 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9841 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9842 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9843 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9844 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9845 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9846 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9847 temporarily for AES and PCLMUL builtins.
9849 PR tree-optimization/84339
9850 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9851 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9855 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9856 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9857 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9859 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9861 * omp-simd-clone.h: New file.
9862 * omp-simd-clone.c: Include omp-simd-clone.h.
9863 (expand_simd_clones): No longer static.
9864 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9865 cgraph.h and omp-simd-clone.h.
9866 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9867 (vect_recog_widen_shift_pattern): Formatting fix.
9868 (vect_pattern_recog_1): Don't check optab for calls.
9871 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9872 operands[2] into a REG before using gen_lowpart on it.
9874 2018-02-12 Jeff Law <law@redhat.com>
9877 * config/sh/sh.c (find_barrier): Consider a sibling call
9880 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9881 successfully back substituting a reg.
9883 2018-02-12 Richard Biener <rguenther@suse.de>
9885 PR tree-optimization/84037
9886 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9887 parameter, move visited init to caller.
9888 (vect_slp_analyze_operations): Separate cost from validity
9889 check, initialize visited once for all instances.
9890 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9892 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9894 (vect_model_promotion_demotion_cost): Likewise.
9895 (vectorizable_bswap): Guard cost modeling with !slp_node
9896 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9898 (vectorizable_call): Likewise.
9899 (vectorizable_conversion): Likewise.
9900 (vectorizable_assignment): Likewise.
9901 (vectorizable_shift): Likewise.
9902 (vectorizable_operation): Likewise.
9903 (vectorizable_store): Likewise.
9904 (vectorizable_load): Likewise.
9905 (vectorizable_condition): Likewise.
9906 (vectorizable_comparison): Likewise.
9908 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9911 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9912 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9913 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9915 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9918 * params.def (inline-min-speedup): Increase from 8 to 15.
9919 (max-inline-insns-auto): Decrease from 40 to 30.
9920 * ipa-split.c (consider_split): Add some buffer for function to
9921 be considered inlining candidate.
9922 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9925 2018-02-12 Richard Biener <rguenther@suse.de>
9927 PR tree-optimization/84037
9928 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9929 matched stmts if we cannot swap the non-matched ones.
9931 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9933 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9934 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9935 _mm_maskz_scalef_round_ss): New intrinsics.
9936 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9937 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9938 __builtin_ia32_scalefss_round): Remove.
9939 (__builtin_ia32_scalefsd_mask_round,
9940 __builtin_ia32_scalefss_mask_round): New intrinsics.
9941 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9942 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9943 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9944 "<round_constraint>")): Changed to ...
9945 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9946 "<round_scalar_constraint>")): ... this.
9947 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9948 %0, %1, %2<round_op3>}"): Changed to ...
9949 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9950 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9951 %2<round_scalar_mask_op3>}"): ... this.
9952 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9954 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9956 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9957 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9958 (_mm_maskz_sqrt_round_ss): New intrinsics.
9959 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9960 (__builtin_ia32_sqrtsd_mask_round)
9961 (__builtin_ia32_sqrtss_mask_round): New builtins.
9962 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9963 (__builtin_ia32_sqrtss_round): Remove.
9964 (__builtin_ia32_sqrtsd_mask_round)
9965 (__builtin_ia32_sqrtss_mask_round): New builtins.
9966 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9967 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9968 ((match_operand:VF_128 1 "vector_operand"
9969 "xBm,<round_constraint>")): Changed to ...
9970 ((match_operand:VF_128 1 "vector_operand"
9971 "xBm,<round_scalar_constraint>")): ... this.
9972 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9973 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9974 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9975 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9976 %<iptr>1<round_scalar_mask_op3>}): ... this.
9977 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9978 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9980 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9983 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9984 Cast vec_cmpeq result to correct type.
9985 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9986 Cast vec_cmpgt result to correct type.
9988 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9990 * final.c (final_scan_insn_1): Renamed from...
9991 (final_scan_insn): ... this. New wrapper, to recover
9992 seen from the outermost call in recursive ones.
9993 * config/sparc/sparc.c (output_return): Drop seen from call.
9994 (output_sibcall): Likewise.
9995 * config/visium/visium.c (output_branch): Likewise.
9997 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
9999 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10002 2018-02-10 Alan Modra <amodra@gmail.com>
10005 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10006 Specify LR as an input.
10008 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10011 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10012 remove_member_access_dummy_vars): New functions.
10013 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10014 lower_omp_1, execute_lower_omp): Use them.
10016 PR rtl-optimization/84308
10017 * shrink-wrap.c (spread_components): Release todo vector.
10019 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10021 PR rtl-optimization/57193
10022 * ira-color.c (struct allocno_color_data): Add member
10023 conflict_allocno_hard_prefs.
10024 (update_conflict_allocno_hard_prefs): New.
10025 (bucket_allocno_compare_func): Add a preference based on
10026 conflict_allocno_hard_prefs.
10027 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10028 (color_allocnos): Remove a dead code. Initiate
10029 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10031 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10034 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10035 constraint from =wa to wa. Avoid a subreg on the output operand,
10036 instead use a pseudo and subreg it in a move.
10037 (p9_xxbrd_<mode>): Changed to ...
10038 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10039 (p9_xxbrd_v2df): New expander.
10040 (p9_xxbrw_<mode>): Changed to ...
10041 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10042 (p9_xxbrw_v4sf): New expander.
10044 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10046 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10048 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10051 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10052 multiply in 32-bit mode.
10053 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10054 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10057 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10059 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10060 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10061 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10062 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10064 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10067 * dwarf2out.c (is_trivial_indirect_ref): New function.
10068 (dwarf2out_late_global_decl): Do not generate a location
10069 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10070 and that are not defined in the current unit.
10072 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10074 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10075 instead of a libcall for UNORDERED.
10077 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10080 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10081 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10083 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10086 * config/s390/s390.c (s390_set_current_function): Invoke
10087 s390_indirect_branch_settings also if fndecl didn't change.
10089 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10091 * config/rs6000/rs6000.md (blockage): Set length to zero.
10093 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10095 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10097 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10100 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10101 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10105 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10106 PARALLEL incoming that failed vt_get_decl_and_offset check.
10108 PR middle-end/84237
10109 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10110 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10112 (get_variable_section): For decls in named .bss* sections pass true as
10113 second argument to bss_initializer_p.
10115 2018-02-09 Marek Polacek <polacek@redhat.com>
10116 Jakub Jelinek <jakub@redhat.com>
10119 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10120 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10121 Sync some changes from cxx_fold_indirect_ref.
10123 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10125 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10127 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10128 (BLOCK_INLINE_ENTRY_LABEL): New.
10129 (dwarf2out_var_location): Disregard inline entry markers.
10130 (inline_entry_data): New struct.
10131 (inline_entry_data_hasher): New hashtable type.
10132 (inline_entry_data_hasher::hash): New.
10133 (inline_entry_data_hasher::equal): New.
10134 (inline_entry_data_table): New variable.
10135 (add_high_low_attributes): Add DW_AT_entry_pc and
10136 DW_AT_GNU_entry_view attributes if a pending entry is found
10137 in inline_entry_data_table. Add old entry_pc attribute only
10138 if debug nonbinding markers are disabled.
10139 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10140 markers are enabled.
10141 (block_within_block_p, dwarf2out_inline_entry): New.
10142 (dwarf2out_finish): Check that no entries remained in
10143 inline_entry_data_table.
10144 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10145 (final_scan_insn, notice_source_line): Likewise.
10146 (rest_of_clean_state): Skip inline entry markers.
10147 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10149 * gimple.c (gimple_build_debug_inline_entry): New.
10150 * gimple.h (enum gimple_debug_subcode): Add
10151 GIMPLE_DEBUG_INLINE_ENTRY.
10152 (gimple_build_debug_inline_entry): Declare.
10153 (gimple_debug_inline_entry_p): New.
10154 (gimple_debug_nonbind_marker_p): Adjust.
10155 * insn-notes.def (INLINE_ENTRY): New.
10156 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10157 inline entry marker notes.
10158 (print_insn): Likewise.
10159 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10160 (INSN_DEBUG_MARKER_KIND): Likewise.
10161 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10162 * tree-inline.c (expand_call_inline): Build and insert
10163 debug_inline_entry stmt.
10164 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10165 inline entry blocks early, if nonbind markers are enabled.
10166 (dump_scope_block): Dump fragment info.
10167 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10168 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10169 (gimple_build_debug_inline_entry): New.
10170 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10171 Enable/disable inline entry points too.
10172 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10173 (DEBUG_INSN): Describe inline entry markers.
10175 * common.opt (gvariable-location-views): New.
10176 (gvariable-location-views=incompat5): New.
10177 * config.in: Rebuilt.
10178 * configure: Rebuilt.
10179 * configure.ac: Test assembler for view support.
10180 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10181 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10182 * dwarf2out.c (var_loc_view): New typedef.
10183 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10184 (dwarf2out_locviews_in_attribute): New.
10185 (dwarf2out_locviews_in_loclist): New.
10186 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10187 (enum dw_line_info_opcode): Add LI_adv_address.
10188 (struct dw_line_info_table): Add view.
10189 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10190 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10191 (zero_view_p): New variable.
10192 (ZERO_VIEW_P): New macro.
10193 (output_asm_line_debug_info): New.
10194 (struct var_loc_node): Add view.
10195 (add_AT_view_list, AT_loc_list): New.
10196 (add_var_loc_to_decl): Add view param. Test it against last.
10197 (new_loc_list): Add view params. Record them.
10198 (AT_loc_list_ptr): Handle loc and view lists.
10199 (view_list_to_loc_list_val_node): New.
10200 (print_dw_val): Handle dw_val_class_view_list.
10201 (size_of_die): Likewise.
10202 (value_format): Likewise.
10203 (loc_list_has_views): New.
10204 (gen_llsym): Set vl_symbol too.
10205 (maybe_gen_llsym, skip_loc_list_entry): New.
10206 (dwarf2out_maybe_output_loclist_view_pair): New.
10207 (output_loc_list): Output view list or entries too.
10208 (output_view_list_offset): New.
10209 (output_die): Handle dw_val_class_view_list.
10210 (output_dwarf_version): New.
10211 (output_compilation_unit_header): Use it.
10212 (output_skeleton_debug_sections): Likewise.
10213 (output_rnglists, output_line_info): Likewise.
10214 (output_pubnames, output_aranges): Update version comments.
10215 (output_one_line_info_table): Output view numbers in asm comments.
10216 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10217 Call maybe_gen_llsym.
10218 (loc_list_from_tree_1): Adjust.
10219 (add_AT_location_description): Create view list attribute if
10220 needed, check it's absent otherwise.
10221 (convert_cfa_to_fb_loc_list): Adjust.
10222 (maybe_emit_file): Call output_asm_line_debug_info for test.
10223 (dwarf2out_var_location): Reset views as needed. Precompute
10224 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10225 attribute. Set view.
10226 (new_line_info_table): Reset next view.
10227 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10228 (dwarf2out_source_line): Likewise. Output view resets and labels to
10229 the assembler, or select appropriate line info opcodes.
10230 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10231 (optimize_string_length): Catch it. Adjust.
10232 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10233 dw_val_class_view_list, and remove it if no longer needed.
10234 (hash_loc_list): Hash view numbers.
10235 (loc_list_hasher::equal): Compare them.
10236 (optimize_location_lists): Check whether a view list symbol is
10237 needed, and whether the locview attribute is present, and
10238 whether they match. Remove the locview attribute if no longer
10240 (index_location_lists): Call skip_loc_list_entry for test.
10241 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10242 Use output_dwarf_version.
10243 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10244 (struct dw_val_node): Add val_view_list.
10245 * final.c (SEEN_NEXT_VIEW): New.
10246 (set_next_view_needed): New.
10247 (clear_next_view_needed): New.
10248 (maybe_output_next_view): New.
10249 (final_start_function): Rename to...
10250 (final_start_function_1): ... this. Take pointer to FIRST,
10251 add SEEN parameter. Emit param bindings in the initial view.
10252 (final_start_function): Reintroduce SEEN-less interface.
10253 (final): Rename to...
10254 (final_1): ... this. Take SEEN parameter. Output final pending
10255 next view at the end.
10256 (final): Reintroduce seen-less interface.
10257 (final_scan_insn): Output pending next view before switching
10258 sections or ending a block. Mark the next view as needed when
10259 outputting variable locations. Notify debug backend of section
10260 changes, and of location view changes.
10261 (rest_of_handle_final): Adjust.
10262 * toplev.c (process_options): Autodetect value for debug variable
10263 location views option. Warn on incompat5 without -gdwarf-5.
10264 * doc/invoke.texi (gvariable-location-views): New.
10265 (gvariable-location-views=incompat5): New.
10266 (gno-variable-location-views): New.
10268 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10270 PR tree-optimization/84136
10271 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10272 that the result of find_edge is non-NULL.
10274 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10277 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10278 storing integer register in SImode. Fix cost of 256 and 512
10279 byte aligned SSE register store.
10281 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10283 * config/i386/i386.c (ix86_multiplication_cost): Fix
10284 multiplication cost for TARGET_AVX512DQ.
10286 2018-02-08 Marek Polacek <polacek@redhat.com>
10288 PR tree-optimization/84238
10289 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10292 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10294 PR tree-optimization/84265
10295 * tree-vect-stmts.c (vectorizable_store): Don't treat
10296 VMAT_CONTIGUOUS accesses as grouped.
10297 (vectorizable_load): Likewise.
10299 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10301 PR tree-optimization/81635
10302 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10303 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10304 (test_round_for_mask): New functions.
10305 (wide_int_cc_tests): Call test_round_for_mask.
10306 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10307 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10308 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10309 range returned by get_range_info.
10311 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10314 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10315 * symtab.c: Include builtins.h
10316 (symtab_node::output_to_lto_symbol_table_p): Move here
10317 from lto-streamer-out.c:output_symbol_p.
10318 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10319 (output_symbol_p): Move all logic to symtab.c
10320 (produce_symtab): Update.
10322 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10324 * config/s390/s390-opts.h (enum indirect_branch): Define.
10325 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10326 (s390_indirect_branch_via_thunk)
10327 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10328 (enum s390_indirect_branch_type): Define.
10329 * config/s390/s390.c (struct s390_frame_layout, struct
10330 machine_function): Remove.
10331 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10332 (indirect_branch_table_label_no, indirect_branch_table_name):
10334 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10335 (enum s390_indirect_branch_option): Define.
10336 (s390_return_addr_from_memory): New function.
10337 (s390_handle_string_attribute): New function.
10338 (s390_attribute_table): Add new attribute handler.
10339 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10340 (s390_indirect_branch_via_thunk): New function.
10341 (s390_indirect_branch_via_inline_thunk): New function.
10342 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10343 sibling call optimization for non z10 compiles.
10344 (s390_emit_call): Force indirect branch target to be a single
10345 register. Add r1 clobber for non-z10 compiles.
10346 (s390_emit_epilogue): Emit return jump via return_use expander.
10347 (s390_reorg): Handle JUMP_INSNs as execute targets.
10348 (s390_option_override_internal): Perform validity checks for the
10349 new command line options.
10350 (s390_indirect_branch_attrvalue): New function.
10351 (s390_indirect_branch_settings): New function.
10352 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10353 (s390_output_indirect_thunk_function): New function.
10354 (s390_code_end): Implement target hook.
10355 (s390_case_values_threshold): Implement target hook.
10356 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10358 * config/s390/s390.h (struct s390_frame_layout)
10359 (struct machine_function): Move here from s390.c.
10360 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10361 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10362 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10363 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10364 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10365 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10366 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10367 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10368 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10369 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10370 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10371 (mnemonic attribute): Add values which aren't recognized
10373 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10374 pattern for branch conversion. Fix mnemonic attribute.
10375 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10376 indirect branch via thunk if requested.
10377 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10378 ("*indirect_jump"): Disable for branch conversion using out of
10380 ("indirect_jump_via_thunk<mode>_z10")
10381 ("indirect_jump_via_thunk<mode>")
10382 ("indirect_jump_via_inlinethunk<mode>_z10")
10383 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10384 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10385 ("casesi_jump_via_inlinethunk<mode>_z10")
10386 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10387 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10388 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10389 ("*indirect2_jump"): Disable for branch conversion.
10390 ("casesi_jump"): Turn into expander and expand patterns for branch
10392 ("return_use"): New expander.
10393 ("*return"): Emit return via thunk and rename it to ...
10394 ("*return<mode>"): ... this one.
10395 * config/s390/s390.opt: Add new options and and enum for the
10398 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10400 * lra-constraints.c (match_reload): Unconditionally use
10401 gen_lowpart_SUBREG, rather than selecting between that
10402 and equivalent gen_rtx_SUBREG code.
10404 2018-02-08 Richard Biener <rguenther@suse.de>
10406 PR tree-optimization/84233
10407 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10408 changed flag instead of boguously re-using phi_inserted.
10410 2018-02-08 Martin Jambor <mjambor@suse.cz>
10412 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10413 static local variables.
10415 2018-02-08 Richard Biener <rguenther@suse.de>
10417 PR tree-optimization/84278
10418 * tree-vect-stmts.c (vectorizable_store): When looking for
10419 smaller vector types to perform grouped strided loads/stores
10420 make sure the mode is supported by the target.
10421 (vectorizable_load): Likewise.
10423 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10425 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10426 Increase LDP/STP opportunities by adding adjacent callee-saves.
10428 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10430 PR rtl-optimization/84068
10431 PR rtl-optimization/83459
10432 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10434 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10436 PR tree-optimization/84224
10437 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10438 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10439 non-zero arguments.
10441 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10444 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10445 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10446 count, remove one USE.
10448 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10450 * doc/install.texi (Configuration): Document the
10451 --with-long-double-format={ibm,ieee} PowerPC configuration
10455 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10456 Convert from define_expand to be define_insn_and_split. Rework
10457 float/double/_Float128 conversions to QI/HI/SImode to work with
10458 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10459 conversions to QI/HImode types did a store and then a load to
10460 truncate the value. For conversions to VSX registers, don't split
10461 the insn, instead emit the code directly. Use the code iterator
10462 any_fix to combine signed and unsigned conversions.
10463 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10464 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10465 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10466 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10467 (fix_<mode>di2_hw): Likewise.
10468 (fixuns_<mode>di2_hw): Likewise.
10469 (fix_<mode>si2_hw): Likewise.
10470 (fixuns_<mode>si2_hw): Likewise.
10471 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10472 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10473 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10474 fix<uns>_trunc<SFDF:mode>si2_p8.
10475 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10476 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10477 (fix<uns>_<mode>_mem): Likewise.
10478 (fctiw<u>z_<mode>_mem): Likewise.
10479 (fix<uns>_<mode>_mem): Likewise.
10480 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10481 the register allocator from doing a direct move to the GPRs to do
10482 a store, and instead use the ISA 3.0 store byte/half-word from
10483 vector register instruction. For IEEE 128-bit floating point,
10484 also optimize stores of 32-bit ints.
10485 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10487 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10489 * genextract.c (push_pathstr_operand): New function to support
10491 (walk_rtx): Call push_pathstr_operand.
10492 (print_path): Support [a-zA-Z].
10494 2018-02-07 Richard Biener <rguenther@suse.de>
10496 PR tree-optimization/84037
10497 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10498 (cse_and_gimplify_to_preheader): Declare.
10499 (vect_get_place_in_interleaving_chain): Likewise.
10500 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10502 (_loop_vec_info::~_loop_vec_info): Delete it.
10503 (cse_and_gimplify_to_preheader): New function.
10504 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10505 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10506 (vectorizable_load): Likewise. For grouped stores always base
10507 the IV on the first element.
10508 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10509 condition before gimplifying.
10511 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10513 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10514 *DIV_EXPR and *MOD_EXPR.
10516 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10519 * config/i386/i386.c (ix86_option_override_internal): Mask out
10520 the CF_SET bit when checking -fcf-protection.
10522 2018-02-07 Tom de Vries <tom@codesourcery.com>
10525 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10528 2018-02-07 Richard Biener <rguenther@suse.de>
10530 PR tree-optimization/84204
10531 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10534 PR tree-optimization/84205
10535 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10536 special-case isl_ast_op_zdiv_r.
10538 PR tree-optimization/84223
10539 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10540 Only add conditions from within the region.
10541 (gather_bbs::after_dom_children): Adjust.
10543 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10546 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10547 * config/avr/avr.md: Only post-reload split REG-REG moves if
10548 either register is GENERAL_REG_P.
10550 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10552 PR tree-optimization/84235
10553 * tree-ssa-scopedtables.c
10554 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10555 if the subtraction is performed in floating point type where NaNs are
10556 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10557 build 1. Formatting fix.
10559 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10562 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10563 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10564 and skip it regardless of bb boundaries. Use CALL_P macro,
10565 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10566 unnecessarily, formatting fix.
10568 2018-02-06 Michael Collison <michael.collison@arm.com>
10570 * config/arm/thumb2.md:
10571 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10572 (*thumb_mov_notscc): Ditto.
10574 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10577 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10578 unsigned_fix, not "s".
10580 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10582 * configure.ac (gcc_fn_eh_frame_ro): New function.
10583 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10584 correct .eh_frame permissions.
10585 * configure: Regenerate.
10587 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10589 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10590 irrelevant options.
10592 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10594 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10595 Display warning message for -mno-speculate-indirect-jumps.
10597 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10599 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10601 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10603 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10605 PR tree-optimization/84225
10606 * tree-eh.c (find_trapping_overflow): Only call
10607 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10609 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10612 * config/i386/i386.c: Reimplement the check of possible options
10613 -mibt/-mshstk conbination. Change error messages.
10614 * doc/invoke.texi: Fix a typo: remove extra '='.
10616 2018-02-06 Marek Polacek <polacek@redhat.com>
10618 PR tree-optimization/84228
10619 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10621 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10624 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10625 emitted arch directives.
10626 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10627 __ARM_FEATURE_COPROC before changing architectures.
10629 2018-02-06 Richard Biener <rguenther@suse.de>
10631 * config/i386/i386.c (print_reg): Fix typo.
10632 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10634 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10636 * configure: Regenerate.
10638 2018-02-05 Martin Sebor <msebor@redhat.com>
10640 PR tree-optimization/83369
10641 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10644 2018-02-05 Martin Liska <mliska@suse.cz>
10646 * doc/invoke.texi: Cherry-pick upstream r323995.
10648 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10650 * ira.c (ira_init_register_move_cost): Adjust comment.
10652 2018-02-05 Martin Liska <mliska@suse.cz>
10654 PR gcov-profile/84137
10655 * doc/gcov.texi: Fix typo in documentation.
10657 2018-02-05 Martin Liska <mliska@suse.cz>
10659 PR gcov-profile/83879
10660 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10661 using dlopen functionality.
10663 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10665 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10666 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10667 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10668 _mm_maskz_range_round_ss): New intrinsics.
10669 (__builtin_ia32_rangesd128_round)
10670 (__builtin_ia32_rangess128_round): Remove.
10671 (__builtin_ia32_rangesd128_mask_round,
10672 __builtin_ia32_rangess128_mask_round): New builtins.
10673 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10674 __builtin_ia32_rangess128_round): Remove.
10675 (__builtin_ia32_rangesd128_mask_round,
10676 __builtin_ia32_rangess128_mask_round): New builtins.
10677 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10678 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10679 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10680 "<round_saeonly_constraint>")): Changed to ...
10681 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10682 "<round_saeonly_scalar_constraint>")): ... this.
10683 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10684 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10685 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10686 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10687 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10689 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10691 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10693 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10694 Remove all values except native, 8540 and 8548.
10696 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10698 * config/i386/i386.c (ix86_output_function_return): Pass
10699 INVALID_REGNUM, instead of -1, as invalid register number to
10700 indirect_thunk_name and output_indirect_thunk.
10702 2018-02-02 Julia Koval <julia.koval@intel.com>
10704 * config.gcc: Add -march=icelake.
10705 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10706 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10707 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10708 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10709 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10710 (processor_target_table): Add icelake.
10711 (ix86_option_override_internal): Handle new PTAs.
10712 (get_builtin_code_for_version): Handle icelake.
10713 (M_INTEL_COREI7_ICELAKE): New.
10714 (fold_builtin_cpu): Handle icelake.
10715 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10716 * doc/invoke.texi: Add -march=icelake.
10718 2018-02-02 Julia Koval <julia.koval@intel.com>
10720 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10721 to wide_int_bitmask.
10722 * wide-int-bitmask.h: New.
10724 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10727 * config/i386/i386.md: Replace Pmode with word_mode in
10728 builtin_setjmp_setup and builtin_longjmp to support x32.
10730 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10734 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10736 (rs6000_supported_cpu_names): New static variable.
10737 (linux_cpu_translation_table): Likewise.
10738 (elf_platform) <cpu>: Define new static variable and use it.
10739 Translate kernel AT_PLATFORM name to canonical name if needed.
10740 Error if platform name is unknown.
10742 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10745 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10747 2018-02-01 Jeff Law <law@redhat.com>
10750 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10751 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10752 the scratch if RELEASE_VIA_POP is false.
10753 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10754 If we have to save a temporary register, decrement SIZE appropriately.
10755 Pass new arguments to release_scratch_register_on_entry.
10756 (ix86_adjust_stack_and_probe): Likewise.
10757 (ix86_emit_probe_stack_range): Pass new arguments to
10758 release_scratch_register_on_entry.
10760 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10762 PR rtl-optimization/84157
10763 * combine.c (change_zero_ext): Use REG_P predicate in
10764 front of HARD_REGISTER_P predicate.
10766 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10768 * config/avr/avr.c (avr_option_override): Move disabling of
10769 -fdelete-null-pointer-checks to...
10770 * common/config/avr/avr-common.c (avr_option_optimization_table):
10773 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10775 PR tree-optimization/81635
10776 * tree-data-ref.c (split_constant_offset_1): For types that
10777 wrap on overflow, try to use range info to prove that wrapping
10780 2018-02-01 Renlin Li <renlin.li@arm.com>
10783 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10784 TAILCALL_ADDR_REGS.
10785 (aarch64_register_move_cost): Likewise.
10786 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10787 TAILCALL_ADDR_REGS.
10788 (REG_CLASS_NAMES): Likewise.
10789 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10790 TAILCALL_ADDR_REGS. Remove IP registers.
10791 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10793 2018-02-01 Richard Biener <rguenther@suse.de>
10795 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10796 for specifying RPO order and allow NULL for that.
10797 * domwalk.c (dom_walker::dom_walker): Likewise.
10798 (dom_walker::walk): Handle NULL RPO order.
10799 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10801 (rewrite_update_dom_walker): Likewise.
10802 (mark_def_dom_walker): Likewise.
10804 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10806 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10807 (aarch64_maybe_expand_sve_subreg_move): Declare.
10808 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10809 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10811 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10812 that are semantically a reverse operation.
10813 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10814 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10815 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10817 (aarch64_can_change_mode_class): For big-endian, forbid changes
10818 between two SVE modes if they have different element sizes.
10820 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10822 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10823 the TImode handling for big-endian targets.
10825 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10827 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10828 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10830 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10831 Remove BSWAP handing for big-endian targets and use the form of
10832 LD1RQ appropariate for the mode.
10834 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10836 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10837 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10838 duplicated element.
10840 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10843 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10844 check for operands that need to go through aarch64_sve_reload_be.
10846 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10848 PR tree-optimization/81661
10849 PR tree-optimization/84117
10850 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10851 * tree-eh.c: Include gimplify.h.
10852 (find_trapping_overflow, replace_trapping_overflow,
10853 rewrite_to_non_trapping_overflow): New functions.
10854 * tree-vect-loop.c: Include tree-eh.h.
10855 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10856 * tree-data-ref.c: Include tree-eh.h.
10857 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10859 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10861 PR rtl-optimization/84123
10862 * combine.c (change_zero_ext): Check if hard register satisfies
10863 can_change_dest_mode before calling gen_lowpart_SUBREG.
10865 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10868 * ira.c (ira_init_register_move_cost): Remove assert.
10870 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10872 PR rtl-optimization/84071
10873 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10874 * doc/tm.texi: Regenerate.
10876 2018-01-31 Richard Biener <rguenther@suse.de>
10878 PR tree-optimization/84132
10879 * tree-data-ref.c (analyze_miv_subscript): Properly
10880 check whether evolution_function_is_affine_multivariate_p
10881 before calling gcd_of_steps_may_divide_p.
10883 2018-01-31 Julia Koval <julia.koval@intel.com>
10886 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10887 * config/i386/i386.md (rdpid_rex64) New.
10888 (rdpid): Make 32bit only.
10890 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10893 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10894 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10896 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10899 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10901 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10903 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10905 PR rtl-optimization/84071
10906 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10907 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10909 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10911 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10912 (arc_attribute_table): Add 'aux' attribute.
10913 (arc_in_small_data_p): Consider aux like variables.
10914 (arc_is_aux_reg_p): New function.
10915 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10916 (arc_get_aux_arg): New function.
10917 (prepare_move_operands): Handle aux-register access.
10918 (arc_handle_aux_attribute): New function.
10919 * doc/extend.texi (ARC Variable attributes): Add subsection.
10921 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10923 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10924 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10925 (arc_attribute_table): Add 'uncached' attribute.
10926 (arc_print_operand): Print '.di' flag for uncached memory
10928 (arc_in_small_data_p): Do not consider for small data the uncached
10930 (arc_is_uncached_mem_p): New function.
10931 * config/arc/predicates.md (compact_store_memory_operand): Check
10932 for uncached memory accesses.
10933 (nonvol_nonimm_operand): Likewise.
10934 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10936 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10939 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10940 falign-loops=): Add Optimization flag.
10942 2018-01-30 Jeff Law <law@redhat.com>
10945 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10946 INT_REGISTERS_SAVED. Check it prior to calling
10947 get_scratch_register_on_entry.
10948 (ix86_adjust_stack_and_probe): Similarly.
10949 (ix86_emit_probe_stack_range): Similarly.
10950 (ix86_expand_prologue): Corresponding changes.
10952 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10955 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10956 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10958 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10961 * lra-constraints.c (curr_insn_transform): Process AND in the
10964 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10966 PR rtl-optimization/83986
10967 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10968 dependence against last_pending_memory_flush in addition to
10969 pending_jump_insns.
10971 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10973 PR tree-optimization/81611
10974 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10977 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10980 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10983 2018-01-30 Richard Biener <rguenther@suse.de>
10984 Jakub Jelinek <jakub@redhat.com>
10986 PR tree-optimization/84111
10987 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10988 inner loops added during recursion, as they don't have up-to-date
10991 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
10994 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
10995 (can_inline_edge_by_limits_p): ... here.
10996 (can_early_inline_edge_p, check_callers,
10997 update_caller_keys, update_callee_keys, recursive_inlining,
10998 add_new_edges_to_heap, speculation_useful_p,
10999 inline_small_functions,
11000 inline_small_functions, flatten_function,
11001 inline_to_all_callers_1): Update.
11003 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11005 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11008 2018-01-30 Richard Biener <rguenther@suse.de>
11010 PR tree-optimization/83008
11011 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11012 invariant and constant vector uses in stmts when they need
11013 more than one stmt.
11015 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11018 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11019 * configure: Regenerate.
11021 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11023 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11025 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11026 Use gen_rtx_REG rather than gen_lowpart.
11028 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11030 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11031 rather than 0 when creating partial subregs.
11033 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11035 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11038 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11041 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11042 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11043 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11044 flags. This restores the settings used before the 2017-07-24.
11045 Turning off pre increment/decrement/modify allows IVOPTS to
11046 optimize DF/SF loops where the index is an int.
11048 2018-01-29 Richard Biener <rguenther@suse.de>
11049 Kelvin Nilsen <kelvin@gcc.gnu.org>
11052 * tree-vect-stmts.c (vectorizable_call): Don't call
11053 targetm.vectorize_builtin_md_vectorized_function if callee is
11056 2018-01-22 Carl Love <cel@us.ibm.com>
11058 * doc/extend.tex: Fix typo in second arg in
11059 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11061 2018-01-29 Richard Biener <rguenther@suse.de>
11063 PR tree-optimization/84086
11064 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11065 (flush_ssaname_freelist): When SSA names were released reset
11066 the SCEV hash table.
11068 2018-01-29 Richard Biener <rguenther@suse.de>
11070 PR tree-optimization/84057
11071 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11072 removed paths when removing edges.
11074 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11076 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11077 -mfunction-return=@var{choice}.
11079 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11081 PR diagnostic/84034
11082 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11083 Handle CR like TAB.
11084 (layout::print_source_line): Likewise.
11085 (test_get_line_width_without_trailing_whitespace): Add test cases.
11087 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11089 PR middle-end/84040
11090 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11093 2018-01-26 Jim Wilson <jimw@sifive.com>
11095 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11097 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11100 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11102 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11103 and CMP + SUB-immediate -> SUBS.
11105 2018-01-26 Martin Sebor <msebor@redhat.com>
11107 PR tree-optimization/83896
11108 * tree-ssa-strlen.c (get_string_len): Rename...
11109 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11110 Avoid assuming length is constant.
11111 (handle_char_store): Use HOST_WIDE_INT for string length.
11113 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11116 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11117 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11119 2018-01-26 Richard Biener <rguenther@suse.de>
11121 PR rtl-optimization/84003
11122 * dse.c (record_store): Only record redundant stores when
11123 the earlier store aliases at least all accesses the later one does.
11125 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11127 PR rtl-optimization/83985
11128 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11129 REG_CFA_RESTORE insns.
11130 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11131 REG_CFA_RESTORE insns here.
11134 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11135 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11137 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11139 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11140 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11141 (arc_init): Likewise.
11142 (arc_override_options): Likewise.
11143 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11145 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11147 * config/arc/arc.h (TARGET_DBNZ): Define.
11148 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11149 properly set the tune attribute.
11150 (dbnz): Use TARGET_DBNZ guard.
11151 * config/arc/arc.opt (mtune): Add core3 option.
11153 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11155 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11156 recognize new pic like addresses.
11157 (arc_delegitimize_address): Clean up.
11159 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11161 * config/arc/arc-arches.def: Option mrf16 valid for all
11163 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11164 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11165 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11166 * config/arc/arc-tables.opt: Regenerate.
11167 * config/arc/arc.c (arc_conditional_register_usage): Handle
11168 reduced register file case.
11169 (arc_file_start): Set must have build attributes.
11170 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11171 mrf16 option value.
11172 * config/arc/arc.opt (mrf16): Add new option.
11173 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11174 * config/arc/genmultilib.awk: Handle new mrf16 option.
11175 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11176 * config/arc/t-multilib: Regenerate.
11177 * doc/invoke.texi (ARC Options): Document mrf16 option.
11179 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11181 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11182 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11183 (arc_attribute_table): Add 'secure_call' attribute.
11184 (arc_print_operand): Print secure call operand.
11185 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11187 (arc_is_secure_call_p): New function. * config/arc/arc.md
11188 (call_i): Add support for sjli instruction.
11189 (call_value_i): Likewise.
11190 * config/arc/constraints.md (Csc): New constraint.
11192 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11193 John Eric Martin <John.Martin@emmicro-us.com>
11195 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11196 * config/arc/arc.c (_arc_jli_section): New struct.
11197 (arc_jli_section): New type.
11198 (rc_jli_sections): New static variable.
11199 (arc_handle_jli_attribute): New function.
11200 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11201 (arc_file_end): New function.
11202 (TARGET_ASM_FILE_END): Define.
11203 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11204 (arc_add_jli_section): New function.
11205 (jli_call_scan): Likewise.
11206 (arc_reorg): Call jli_call_scan.
11207 (arc_output_addsi): Remove 'S' from printing asm operand.
11208 (arc_is_jli_call_p): New function.
11209 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11211 (movhi_insn): Likewise.
11212 (movsi_insn): Likewise.
11213 (movsi_set_cc_insn): Likewise.
11214 (loadqi_update): Likewise.
11215 (load_zeroextendqisi_update): Likewise.
11216 (load_signextendqisi_update): Likewise.
11217 (loadhi_update): Likewise.
11218 (load_zeroextendhisi_update): Likewise.
11219 (load_signextendhisi_update): Likewise.
11220 (loadsi_update): Likewise.
11221 (loadsf_update): Likewise.
11222 (movsicc_insn): Likewise.
11223 (bset_insn): Likewise.
11224 (bxor_insn): Likewise.
11225 (bclr_insn): Likewise.
11226 (bmsk_insn): Likewise.
11227 (bicsi3_insn): Likewise.
11228 (cmpsi_cc_c_insn): Likewise.
11229 (movsi_ne): Likewise.
11230 (movsi_cond_exec): Likewise.
11231 (clrsbsi2): Likewise.
11232 (norm_f): Likewise.
11242 (call_i): Remove 'S' asm letter, add jli instruction.
11243 (call_value_i): Likewise.
11244 * config/arc/arc.op (mjli-always): New option.
11245 * config/arc/constraints.md (Cji): New constraint.
11246 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11248 (subsf3_fpx): Likewise.
11249 (mulsf3_fpx): Likewise.
11250 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11252 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11253 function attrbutes.
11254 * doc/invoke.texi (ARC): Document mjli-always option.
11256 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11258 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11259 avoid addition with 0 and use incw and decw where possible.
11261 2018-01-26 Richard Biener <rguenther@suse.de>
11263 PR tree-optimization/81082
11264 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11265 association if it requires casting to unsigned.
11266 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11267 from fold_plusminus_mult_expr to catch important cases late when
11268 range info is available.
11270 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11272 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11273 * configure.ac (hidden_linkonce): New test.
11274 * configure: Regenerate.
11275 * config.in: Regenerate.
11277 2018-01-26 Julia Koval <julia.koval@intel.com>
11279 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11280 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11281 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11282 _mm_mask_bitshuffle_epi64_mask): Fix type.
11283 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11284 USI_FTYPE_V4DI_V4DI_USI): Remove.
11285 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11286 __builtin_ia32_vpshufbitqmb256_mask,
11287 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11288 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11289 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11291 2018-01-26 Alan Modra <amodra@gmail.com>
11294 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11295 UNSPEC_VBPERMQ. Sort other unspecs.
11297 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11299 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11301 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11303 PR middle-end/83055
11304 * predict.c (drop_profile): Do not push/pop cfun; update also
11306 (handle_missing_profiles): Fix logic looking for zero profiles.
11308 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11310 PR middle-end/83977
11311 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11312 on functions with #pragma omp declare simd or functions with simd
11314 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11315 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11316 Remove trailing \n from warning_at calls.
11318 2018-01-25 Tom de Vries <tom@codesourcery.com>
11321 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11322 for neutered workers.
11324 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11327 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11328 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11330 2018-01-24 Jeff Law <law@redhat.com>
11333 * i386.c (get_probe_interval): Move to earlier point.
11334 (ix86_compute_frame_layout): If -fstack-clash-protection and
11335 the frame is larger than the probe interval, then use pushes
11336 to save registers rather than reg->mem moves.
11337 (ix86_expand_prologue): Remove conditional for int_registers_saved
11340 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11343 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11344 min/max for never referenced object.
11346 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11348 PR middle-end/83977
11349 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11351 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11352 attributes from DECL_ATTRIBUTES (decl) without affecting
11353 DECL_ATTRIBUTES (current_function_decl).
11354 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11355 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11357 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11359 PR tree-optimization/83979
11360 * fold-const.c (fold_comparison): Use constant_boolean_node
11361 instead of boolean_{true,false}_node.
11363 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11365 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11368 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11370 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11371 Simplify the clause that sets the length attribute.
11372 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11373 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11374 clause that sets the length attribute.
11375 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11377 2018-01-24 Tom de Vries <tom@codesourcery.com>
11380 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11381 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11382 Add strict parameter.
11383 (prevent_branch_around_nothing): Insert dummy insn between branch to
11384 label and label with no ptx insn inbetween.
11385 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11387 2018-01-24 Tom de Vries <tom@codesourcery.com>
11390 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11391 for neutered threads in warp.
11392 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11394 2018-01-24 Richard Biener <rguenther@suse.de>
11396 PR tree-optimization/83176
11397 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11400 2018-01-24 Richard Biener <rguenther@suse.de>
11402 PR tree-optimization/82819
11403 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11404 code generating pluses that are no-ops in the target precision.
11406 2018-01-24 Richard Biener <rguenther@suse.de>
11408 PR middle-end/84000
11409 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11411 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11413 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11414 to merge probabilities.
11415 * predict.c (probably_never_executed): Also mark as cold functions
11416 with global 0 profile and guessed local profile.
11417 * profile-count.c (profile_probability::combine_with_count): New
11419 * profile-count.h (profile_probability::operator*,
11420 profile_probability::operator*=, profile_probability::operator/,
11421 profile_probability::operator/=): Reduce precision to adjusted
11422 and set value to guessed on contradictory divisions.
11423 (profile_probability::combine_with_freq): Remove.
11424 (profile_probability::combine_wiht_count): Declare.
11425 (profile_count::force_nonzero):: Set to adjusted.
11426 (profile_count::probability_in):: Set quality to adjusted.
11427 * tree-ssa-tail-merge.c (replace_block_by): Use
11428 combine_with_count.
11430 2018-01-23 Andrew Waterman <andrew@sifive.com>
11431 Jim Wilson <jimw@sifive.com>
11433 * config/riscv/riscv.c (riscv_stack_boundary): New.
11434 (riscv_option_override): Set riscv_stack_boundary. Handle
11435 riscv_preferred_stack_boundary_arg.
11436 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11437 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11438 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11439 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11440 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11441 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11443 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11446 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11447 of struct ix86_frame.
11448 (ix86_expand_epilogue): Likewise. Add a local variable for
11449 the reg_save_offset field in struct ix86_frame.
11451 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11453 PR tree-optimization/82604
11454 * tree-loop-distribution.c (enum partition_kind): New enum item
11455 PKIND_PARTIAL_MEMSET.
11456 (partition_builtin_p): Support above new enum item.
11457 (generate_code_for_partition): Ditto.
11458 (compute_access_range): Differentiate cases that equality can be
11459 proven at all loops, the innermost loops or no loops.
11460 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11461 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11462 (finalize_partitions, distribute_loop): Don't fuse partition of
11463 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11464 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11465 parloop is enabled.
11467 2018-01-23 Martin Liska <mliska@suse.cz>
11469 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11470 order to ignore the predictor.
11471 (PRED_POLYMORPHIC_CALL): Likewise.
11472 (PRED_RECURSIVE_CALL): Likewise.
11474 2018-01-23 Martin Liska <mliska@suse.cz>
11476 * tree-profile.c (tree_profiling): Print function header to
11477 aware reader which function we are working on.
11478 * value-prof.c (gimple_find_values_to_profile): Do not print
11479 not interesting value histograms.
11481 2018-01-23 Martin Liska <mliska@suse.cz>
11483 * profile-count.h (enum profile_quality): Add
11484 profile_uninitialized as the first value. Do not number values
11485 as they are zero based.
11486 (profile_count::verify): Update sanity check.
11487 (profile_probability::verify): Likewise.
11489 2018-01-23 Nathan Sidwell <nathan@acm.org>
11491 * doc/invoke.texi (ffor-scope): Deprecate.
11493 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11495 PR tree-optimization/83510
11496 * domwalk.c (set_all_edges_as_executable): New function.
11497 (dom_walker::dom_walker): Convert bool param
11498 "skip_unreachable_blocks" to enum reachability. Move setup of
11499 edge flags to set_all_edges_as_executable and only do it when
11500 reachability is REACHABLE_BLOCKS.
11501 * domwalk.h (enum dom_walker::reachability): New enum.
11502 (dom_walker::dom_walker): Convert bool param
11503 "skip_unreachable_blocks" to enum reachability.
11504 (set_all_edges_as_executable): New decl.
11505 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11506 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11508 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11509 but converting true to REACHABLE_BLOCKS.
11510 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11512 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11513 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11514 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11516 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11517 if check_all_array_refs will be called.
11519 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11521 * tree.c (selftest::test_location_wrappers): Add more test
11524 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11526 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11527 (selftest::test_bit_in_range): Likewise.
11529 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11532 * doc/sourcebuild.texi (vect_float): Say that the selector
11533 only describes the situation when -funsafe-math-optimizations is on.
11534 (vect_float_strict): Document.
11536 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11538 PR tree-optimization/83965
11539 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11540 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11541 instead of checking only for a reduction.
11542 (vect_recog_widen_sum_pattern): Likewise.
11544 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11546 * predict.c (probably_never_executed): Only use precise profile info.
11547 (compute_function_frequency): Skip after inlining hack since we now
11548 have quality checking.
11550 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11552 * profile-count.h (profile_probability::very_unlikely,
11553 profile_probability::unlikely, profile_probability::even): Set
11554 precision to guessed.
11556 2018-01-23 Richard Biener <rguenther@suse.de>
11558 PR tree-optimization/83963
11559 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11560 Properly terminate dominator walk when crossing the exit edge not
11561 when visiting its source block.
11563 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11566 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11567 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11569 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11571 PR tree-optimization/83957
11572 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11573 semicolon after for body surrounded by braces.
11575 PR tree-optimization/83081
11576 * profile-count.h (profile_probability::split): New method.
11577 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11578 Use profile_probability::split.
11579 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11580 when splitting a single conditional jump into 2.
11582 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11584 PR tree-optimization/69452
11585 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11588 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11590 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11591 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11592 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11594 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11596 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11598 * config/rl78/rl78.md (movdi): New define_expand.
11599 * config/rl78/rl78.c (rl78_split_movdi): New function.
11601 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11604 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11606 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11607 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11608 128-bit to produce an UNSPEC move to get the double word with the
11609 signbit and then a shift directly to do signbit.
11610 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11611 implementation with a new version that just does either a direct
11612 move or a regular move. Move memory interface to separate insns.
11613 Move insns so they are next to the expander.
11614 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11615 with signbit move. Split big and little endian case.
11616 (signbit<mode>2_dm_mem_le): Likewise.
11617 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11618 (signbit<mode>2_dm2): Likewise.
11620 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11622 * config/rl78/rl78.md (anddi3): New define_expand.
11624 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11626 * config/rl78/rl78.md (umindi3): New define_expand.
11628 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11630 * config/rl78/rl78.md (smindi3): New define_expand.
11632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11634 * config/rl78/rl78.md (smaxdi3): New define_expand.
11636 2018-01-22 Carl Love <cel@us.ibm.com>
11638 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11639 LVX_V1TI): Add macro expansion.
11640 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11641 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11642 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11643 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11644 Change check to determine if the instruction is a byte reversing
11645 entry. Fix typo in comment.
11646 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11647 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11648 Add def_builtin calls for new builtins.
11649 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11650 Add define_insn expansion.
11652 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11654 * config/rl78/rl78.md (umaxdi3): New define_expand.
11656 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11658 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11659 for non-QImode registers.
11661 2018-01-22 Richard Biener <rguenther@suse.de>
11663 PR tree-optimization/83963
11664 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11665 including the loop exit block.
11666 (scop_detection::merge_sese): Likewise.
11667 (scop_detection::add_scop): Do it here instead.
11669 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11671 * doc/sourcebuild.texi (arm_softfloat): Document.
11673 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11676 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11677 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11678 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11680 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11681 David Edelsohn <dje.gcc@gmail.com>
11684 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11685 Change "crset eq" to "crset 2".
11686 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11687 (*call_indirect_aix<mode>_nospec): Likewise.
11688 (*call_value_indirect_aix<mode>_nospec): Likewise.
11689 (*call_indirect_elfv2<mode>_nospec): Likewise.
11690 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11691 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11692 change assembly output from . to $.
11693 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11694 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11695 (*tablejump<mode>_internal1_nospec): Likewise.
11697 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11700 * config/sh/sh_optimize_sett_clrt.cc:
11701 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11703 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11705 PR tree-optimization/83940
11706 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11707 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11708 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11709 use it to pass back the definition type.
11710 (vect_check_store_rhs): Likewise rhs_dt_out.
11711 (vect_build_gather_load_calls): Add a mask_dt argument and use
11712 it instead of a call to vect_is_simple_use.
11713 (vectorizable_store): Update calls to vect_check_load_store_mask
11714 and vect_check_store_rhs. Use the dt returned by the latter instead
11715 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11716 instead of calls to vect_is_simple_use. Pass the scalar rather
11717 than the vector operand to vect_is_simple_use when handling
11718 second and subsequent copies of an rhs value.
11719 (vectorizable_load): Update calls to vect_check_load_store_mask
11720 and vect_build_gather_load_calls. Use the cached mask_dt and
11721 gs_info.offset_dt instead of calls to vect_is_simple_use.
11723 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11725 PR middle-end/83945
11726 * tree-emutls.c: Include gimplify.h.
11727 (lower_emutls_2): New function.
11728 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11729 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11730 it before further processing.
11733 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11734 UINTVAL (trueop1) instead of INTVAL (op1).
11736 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11740 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11741 INCOMING_FRAME_SP_OFFSET if not defined.
11742 (scan_trace): Add ENTRY argument. If true and
11743 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11744 emit a note to adjust the CFA offset.
11745 (create_cfi_notes): Adjust scan_trace callers.
11746 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11747 INCOMING_FRAME_SP_OFFSET in the CIE.
11748 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11749 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11751 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11752 * doc/tm.texi: Regenerated.
11754 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11756 PR rtl-optimization/83147
11757 * lra-constraints.c (remove_inheritance_pseudos): Use
11758 lra_substitute_pseudo_within_insn.
11760 2018-01-19 Tom de Vries <tom@codesourcery.com>
11761 Cesar Philippidis <cesar@codesourcery.com>
11764 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11766 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11769 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11770 spaces for function labels.
11772 2018-01-19 Martin Liska <mliska@suse.cz>
11774 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11775 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11776 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11777 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11778 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11779 (PRED_CONST_RETURN): Change from 69 to 65.
11780 (PRED_NULL_RETURN): Change from 91 to 71.
11781 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11782 (PRED_LOOP_GUARD): Change from 66 to 73.
11784 2018-01-19 Martin Liska <mliska@suse.cz>
11786 * predict.c (predict_insn_def): Add new assert.
11787 (struct branch_predictor): Change type to signed integer.
11788 (test_prediction_value_range): Amend test to cover
11789 PROB_UNINITIALIZED.
11790 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11791 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11792 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11793 (PRED_LOOP_IV_COMPARE): Likewise.
11794 * predict.h (PROB_UNINITIALIZED): Define new constant.
11796 2018-01-19 Martin Liska <mliska@suse.cz>
11798 * predict.c (dump_prediction): Add new format for
11799 analyze_brprob.py script which is enabled with -details
11801 * profile-count.h (precise_p): New function.
11803 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11805 PR tree-optimization/83922
11806 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11807 there are no statements that need masking.
11808 (vect_active_double_reduction_p): New function.
11809 (vect_analyze_loop_operations): Use it when handling phis that
11810 are not in the loop header.
11812 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11814 PR tree-optimization/83914
11815 * tree-vect-loop.c (vectorizable_induction): Don't convert
11816 init_expr or apply the peeling adjustment for inductions
11817 that are nested within the vectorized loop.
11819 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11821 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11824 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11828 * function.h (gimplify_parameters): Add gimple_seq * argument.
11829 * function.c: Include gimple.h and options.h.
11830 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11831 for the added local temporaries if needed.
11832 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11833 if there are any parameter cleanups, wrap whole body into a
11834 try/finally with the cleanups.
11836 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11839 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11840 Use GET_MODE_CLASS for scalar floating point.
11842 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11846 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11847 Fix call of call_cgraph_insertion_hooks.
11849 2018-01-18 Martin Sebor <msebor@redhat.com>
11851 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11853 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11856 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11859 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11862 * common.opt: (-ffile-prefix-map): New option.
11863 * opts.c (common_handle_option): Defer it.
11864 * opts-global.c (handle_common_deferred_options): Handle it.
11865 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11866 * file-prefix-map.h: New file.
11867 (remap_debug_filename, add_debug_prefix_map): ...here.
11868 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11869 * final.c (debug_prefix_map, add_debug_prefix_map
11870 remap_debug_filename): Move to...
11871 * file-prefix-map.c: New file.
11872 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11873 generalize, get rid of alloca(), use strrchr() instead of strchr().
11874 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11875 Implement in terms of add_prefix_map().
11876 (remap_macro_filename, remap_debug_filename): Implement in term of
11878 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11879 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11880 * dbxout.c: Include file-prefix-map.h.
11881 * varasm.c: Likewise.
11882 * vmsdbgout.c: Likewise.
11883 * xcoffout.c: Likewise.
11884 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11885 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11886 * doc/invoke.texi (-ffile-prefix-map): Document.
11887 (-fdebug-prefix-map): Update description.
11889 2018-01-18 Martin Liska <mliska@suse.cz>
11891 * config/i386/i386.c (indirect_thunk_name): Document that also
11893 (output_indirect_thunk): Document why both instructions
11894 (pause and lfence) are generated.
11896 2018-01-18 Richard Biener <rguenther@suse.de>
11898 PR tree-optimization/83887
11899 * graphite-scop-detection.c
11900 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11901 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11902 (scop_detection::merge_sese): Re-implement with a flood-fill
11903 algorithm that properly finds a SESE region if it exists.
11905 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11908 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11909 pointer_diff optimizations use view_convert instead of convert.
11911 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11913 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11914 Generate different code for -mno-speculate-indirect-jumps.
11915 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11916 (*call_indirect_aix<mode>): Disable for
11917 -mno-speculate-indirect-jumps.
11918 (*call_indirect_aix<mode>_nospec): New define_insn.
11919 (*call_value_indirect_aix<mode>): Disable for
11920 -mno-speculate-indirect-jumps.
11921 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11922 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11923 -mno-speculate-indirect-jumps.
11924 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11926 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11928 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11929 long double type, set the flags for noting the default long double
11930 type, even if we don't pass or return a long double type.
11932 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11935 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11938 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11940 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11941 support for merge[hl].
11942 (fold_mergehl_helper): New helper function.
11943 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11944 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11945 (altivec_vmrglw_direct): Add xxmrglw insn.
11947 2018-01-17 Andrew Waterman <andrew@sifive.com>
11949 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11950 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11952 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11955 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11956 call the lto_location_cache before reading the
11957 DECL_SOURCE_LOCATION of the types.
11959 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11960 Richard Sandiford <richard.sandiford@linaro.org>
11962 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11963 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11964 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11965 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11966 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11968 * config/aarch64/constraints.md (aarch64_movti_operand):
11970 * config/aarch64/predicates.md (Uti): Add new constraint.
11972 2018-01-17 Carl Love <cel@us.ibm.com>
11974 * config/rs6000/vsx.md (define_expand xl_len_r,
11975 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11976 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11978 (define_expand, define_insn): Move the shift left from the
11979 define_insn to the define_expand for lxvl and stxvl instructions.
11980 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11981 and XL_LEN_R definitions to PURE.
11983 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11985 * config/i386/i386.c (indirect_thunk_name): Declare regno
11986 as unsigned int. Compare regno with INVALID_REGNUM.
11987 (output_indirect_thunk): Ditto.
11988 (output_indirect_thunk_function): Ditto.
11989 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11990 in the call to output_indirect_thunk_function.
11992 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
11994 PR middle-end/83884
11995 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
11996 rather than the size of inner_type to determine the stack slot size
11997 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
11999 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12002 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12005 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12007 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12008 endian Linux systems to optionally enable multilibs for selecting
12009 the long double type if the user configured an explicit type.
12010 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12011 have no long double multilibs if not defined.
12012 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12013 warn if the user used -mabi={ieee,ibm}longdouble and we built
12014 multilibs for long double.
12015 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12016 appropriate multilib option.
12017 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12019 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12020 for building long double multilibs.
12021 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12023 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12025 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12028 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12030 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12033 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12036 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12039 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12041 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12042 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12044 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12046 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12047 different rtl trees depending on TARGET_64BIT.
12048 (rs6000_gen_lvx): Likewise.
12050 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12052 * config/visium/visium.md (nop): Tweak comment.
12053 (hazard_nop): Likewise.
12055 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12057 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12058 -mspeculate-indirect-jumps.
12059 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12060 for -mno-speculate-indirect-jumps.
12061 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12062 (*call_value_indirect_elfv2<mode>): Disable for
12063 -mno-speculate-indirect-jumps.
12064 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12065 (indirect_jump): Emit different RTL for
12066 -mno-speculate-indirect-jumps.
12067 (*indirect_jump<mode>): Disable for
12068 -mno-speculate-indirect-jumps.
12069 (*indirect_jump<mode>_nospec): New define_insn.
12070 (tablejump): Emit different RTL for
12071 -mno-speculate-indirect-jumps.
12072 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12073 (tablejumpsi_nospec): New define_expand.
12074 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12075 (tablejumpdi_nospec): New define_expand.
12076 (*tablejump<mode>_internal1): Disable for
12077 -mno-speculate-indirect-jumps.
12078 (*tablejump<mode>_internal1_nospec): New define_insn.
12079 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12082 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12084 * caller-save.c (insert_save): Drop unnecessary parameter. All
12087 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12088 Richard Biener <rguenth@suse.de>
12091 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12092 return early, inline manually is_gimple_sizepos. Make sure if we
12093 call gimplify_expr we don't end up with a gimple constant.
12094 * tree.c (variably_modified_type_p): Don't return true for
12095 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12096 * gimplify.h (is_gimple_sizepos): Remove.
12098 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12100 PR tree-optimization/83857
12101 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12102 vectorizable_live_operation for pure SLP statements.
12103 (vectorizable_live_operation): Handle PHIs.
12105 2018-01-16 Richard Biener <rguenther@suse.de>
12107 PR tree-optimization/83867
12108 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12109 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12111 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12114 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12115 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12116 If off is not INTEGER_CST, issue a may not be aligned warning
12117 rather than isn't aligned. Use isn%'t rather than isn't.
12118 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12120 <case MULT_EXPR>: Improve the case when bottom and one of the
12121 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12122 operand, in that case check if the other operand is multiple of
12123 bottom divided by the INTEGER_CST operand.
12125 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12128 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12129 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12130 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12131 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12132 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12133 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12134 (pa_function_arg_size): New function.
12136 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12138 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12139 in a separate statement.
12141 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12143 PR tree-optimization/83847
12144 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12145 group gathers and scatters.
12147 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12149 PR rtl-optimization/86620
12150 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12152 PR rtl-optimization/83213
12153 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12154 to last if both are JUMP_INSNs.
12156 PR tree-optimization/83843
12157 * gimple-ssa-store-merging.c
12158 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12159 store_immediate_info for bswap/nop orig_stores.
12161 2018-01-15 Andrew Waterman <andrew@sifive.com>
12163 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12165 <UDIV>: Increase cost if !TARGET_DIV.
12167 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12169 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12170 (define_attr "cr_logical_3op"): New.
12171 (cceq_ior_compare): Adjust.
12172 (cceq_ior_compare_complement): Adjust.
12173 (*cceq_rev_compare): Adjust.
12174 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12175 (is_cracked_insn): Adjust.
12176 (insn_must_be_first_in_group): Adjust.
12177 * config/rs6000/40x.md: Adjust.
12178 * config/rs6000/440.md: Adjust.
12179 * config/rs6000/476.md: Adjust.
12180 * config/rs6000/601.md: Adjust.
12181 * config/rs6000/603.md: Adjust.
12182 * config/rs6000/6xx.md: Adjust.
12183 * config/rs6000/7450.md: Adjust.
12184 * config/rs6000/7xx.md: Adjust.
12185 * config/rs6000/8540.md: Adjust.
12186 * config/rs6000/cell.md: Adjust.
12187 * config/rs6000/e300c2c3.md: Adjust.
12188 * config/rs6000/e500mc.md: Adjust.
12189 * config/rs6000/e500mc64.md: Adjust.
12190 * config/rs6000/e5500.md: Adjust.
12191 * config/rs6000/e6500.md: Adjust.
12192 * config/rs6000/mpc.md: Adjust.
12193 * config/rs6000/power4.md: Adjust.
12194 * config/rs6000/power5.md: Adjust.
12195 * config/rs6000/power6.md: Adjust.
12196 * config/rs6000/power7.md: Adjust.
12197 * config/rs6000/power8.md: Adjust.
12198 * config/rs6000/power9.md: Adjust.
12199 * config/rs6000/rs64.md: Adjust.
12200 * config/rs6000/titan.md: Adjust.
12202 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12204 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12205 ix86_indirect_branch_register logic.
12207 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12209 * config/i386/constraints.md (Bs): Update
12210 ix86_indirect_branch_register check. Don't check
12211 ix86_indirect_branch_register with GOT_memory_operand.
12213 * config/i386/predicates.md (GOT_memory_operand): Don't check
12214 ix86_indirect_branch_register here.
12215 (GOT32_symbol_operand): Likewise.
12217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12219 * config/i386/predicates.md (constant_call_address_operand):
12220 Rewrite ix86_indirect_branch_register logic.
12221 (sibcall_insn_operand): Likewise.
12223 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12225 * config/i386/constraints.md (Bs): Replace
12226 ix86_indirect_branch_thunk_register with
12227 ix86_indirect_branch_register.
12229 * config/i386/i386.md (indirect_jump): Likewise.
12230 (tablejump): Likewise.
12231 (*sibcall_memory): Likewise.
12232 (*sibcall_value_memory): Likewise.
12233 Peepholes of indirect call and jump via memory: Likewise.
12234 * config/i386/i386.opt: Likewise.
12235 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12236 (GOT_memory_operand): Likewise.
12237 (call_insn_operand): Likewise.
12238 (sibcall_insn_operand): Likewise.
12239 (GOT32_symbol_operand): Likewise.
12241 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12243 PR middle-end/83837
12244 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12245 type rather than type addr's type points to.
12246 (expand_omp_atomic_mutex): Likewise.
12247 (expand_omp_atomic): Likewise.
12249 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12252 * config/i386/i386.c (output_indirect_thunk_function): Use
12253 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12254 for __x86_return_thunk.
12256 2018-01-15 Richard Biener <rguenther@suse.de>
12258 PR middle-end/83850
12259 * expmed.c (extract_bit_field_1): Fix typo.
12261 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12264 * config/arm/iterators.md (VF): New mode iterator.
12265 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12266 Remove integer-related logic from pattern.
12267 (neon_vabd<mode>_3): Likewise.
12269 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12271 PR middle-end/82694
12272 * common.opt (fstrict-overflow): No longer an alias.
12273 (fwrapv-pointer): New option.
12274 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12275 also for pointer types based on flag_wrapv_pointer.
12276 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12277 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12278 opts->x_flag_wrapv got set.
12279 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12280 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12281 POINTER_TYPE_OVERFLOW_UNDEFINED.
12282 * match.pd: Likewise in address comparison pattern.
12283 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12285 2018-01-15 Richard Biener <rguenther@suse.de>
12288 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12289 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12290 Reset type names to their identifier if their TYPE_DECL doesn't
12291 have linkage (and thus is used for ODR and devirt).
12292 (save_debug_info_for_decl): Remove.
12293 (save_debug_info_for_type): Likewise.
12294 (add_tree_to_fld_list): Adjust.
12295 * tree-pretty-print.c (dump_generic_node): Make dumping of
12296 type names more robust.
12298 2018-01-15 Richard Biener <rguenther@suse.de>
12300 * BASE-VER: Bump to 8.0.1.
12302 2018-01-14 Martin Sebor <msebor@redhat.com>
12305 * builtins.c (check_access): Avoid warning when the no-warning bit
12308 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12310 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12311 * ira-color (allocno_hard_regs_compare): Likewise.
12313 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12316 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12317 Use .pushsection/.popsection.
12319 2018-01-14 Martin Sebor <msebor@redhat.com>
12322 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12324 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12326 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12327 entry from extra_headers.
12328 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12329 extra_headers, make the list bitwise identical to the i?86-*-* one.
12331 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12333 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12334 -mcmodel=large with -mindirect-branch=thunk,
12335 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12336 -mfunction-return=thunk-extern.
12337 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12338 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12339 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12341 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12343 * config/i386/i386.c (print_reg): Print the name of the full
12344 integer register without '%'.
12345 (ix86_print_operand): Handle 'V'.
12346 * doc/extend.texi: Document 'V' modifier.
12348 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12350 * config/i386/constraints.md (Bs): Disallow memory operand for
12351 -mindirect-branch-register.
12353 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12354 (GOT_memory_operand): Likewise.
12355 (call_insn_operand): Likewise.
12356 (sibcall_insn_operand): Likewise.
12357 (GOT32_symbol_operand): Likewise.
12358 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12359 for -mindirect-branch-register.
12360 (tablejump): Likewise.
12361 (*sibcall_memory): Likewise.
12362 (*sibcall_value_memory): Likewise.
12363 Disallow peepholes of indirect call and jump via memory for
12364 -mindirect-branch-register.
12365 (*call_pop): Replace m with Bw.
12366 (*call_value_pop): Likewise.
12367 (*sibcall_pop_memory): Replace m with Bs.
12368 * config/i386/i386.opt (mindirect-branch-register): New option.
12369 * doc/invoke.texi: Document -mindirect-branch-register option.
12371 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12373 * config/i386/i386-protos.h (ix86_output_function_return): New.
12374 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12375 set function_return_type.
12376 (indirect_thunk_name): Add ret_p to indicate thunk for function
12378 (output_indirect_thunk_function): Pass false to
12379 indirect_thunk_name.
12380 (ix86_output_indirect_branch_via_reg): Likewise.
12381 (ix86_output_indirect_branch_via_push): Likewise.
12382 (output_indirect_thunk_function): Create alias for function
12383 return thunk if regno < 0.
12384 (ix86_output_function_return): New function.
12385 (ix86_handle_fndecl_attribute): Handle function_return.
12386 (ix86_attribute_table): Add function_return.
12387 * config/i386/i386.h (machine_function): Add
12388 function_return_type.
12389 * config/i386/i386.md (simple_return_internal): Use
12390 ix86_output_function_return.
12391 (simple_return_internal_long): Likewise.
12392 * config/i386/i386.opt (mfunction-return=): New option.
12393 (indirect_branch): Mention -mfunction-return=.
12394 * doc/extend.texi: Document function_return function attribute.
12395 * doc/invoke.texi: Document -mfunction-return= option.
12397 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12399 * config/i386/i386-opts.h (indirect_branch): New.
12400 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12401 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12402 with local indirect jump when converting indirect call and jump.
12403 (ix86_set_indirect_branch_type): New.
12404 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12405 (indirectlabelno): New.
12406 (indirect_thunk_needed): Likewise.
12407 (indirect_thunk_bnd_needed): Likewise.
12408 (indirect_thunks_used): Likewise.
12409 (indirect_thunks_bnd_used): Likewise.
12410 (INDIRECT_LABEL): Likewise.
12411 (indirect_thunk_name): Likewise.
12412 (output_indirect_thunk): Likewise.
12413 (output_indirect_thunk_function): Likewise.
12414 (ix86_output_indirect_branch_via_reg): Likewise.
12415 (ix86_output_indirect_branch_via_push): Likewise.
12416 (ix86_output_indirect_branch): Likewise.
12417 (ix86_output_indirect_jmp): Likewise.
12418 (ix86_code_end): Call output_indirect_thunk_function if needed.
12419 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12421 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12422 (ix86_attribute_table): Add indirect_branch.
12423 * config/i386/i386.h (machine_function): Add indirect_branch_type
12424 and has_local_indirect_jump.
12425 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12427 (tablejump): Likewise.
12428 (*indirect_jump): Use ix86_output_indirect_jmp.
12429 (*tablejump_1): Likewise.
12430 (simple_return_indirect_internal): Likewise.
12431 * config/i386/i386.opt (mindirect-branch=): New option.
12432 (indirect_branch): New.
12435 (thunk-inline): Likewise.
12436 (thunk-extern): Likewise.
12437 * doc/extend.texi: Document indirect_branch function attribute.
12438 * doc/invoke.texi: Document -mindirect-branch= option.
12440 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12443 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12445 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12447 * ipa-inline.c (want_inline_small_function_p): Return false if
12448 inlining has already failed with CIF_FINAL_ERROR.
12449 (update_caller_keys): Call want_inline_small_function_p before
12451 (update_callee_keys): Likewise.
12453 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12455 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12457 (rs6000_quadword_masked_address_p): Likewise.
12458 (quad_aligned_load_p): Likewise.
12459 (quad_aligned_store_p): Likewise.
12460 (const_load_sequence_p): Add comment to describe the outer-most loop.
12461 (mimic_memory_attributes_and_flags): New function.
12462 (rs6000_gen_stvx): Likewise.
12463 (replace_swapped_aligned_store): Likewise.
12464 (rs6000_gen_lvx): Likewise.
12465 (replace_swapped_aligned_load): Likewise.
12466 (replace_swapped_load_constant): Capitalize argument name in
12467 comment describing this function.
12468 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12469 and stores that access quad-word aligned addresses and replace
12470 with stvx or lvx instructions when appropriate.
12471 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12472 New function prototype.
12473 (rs6000_quadword_masked_address_p): Likewise.
12474 (rs6000_gen_lvx): Likewise.
12475 (rs6000_gen_stvx): Likewise.
12476 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12477 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12478 when memory address is aligned.
12479 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12480 this split to select lvx instruction when memory address is aligned.
12481 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12482 instruction when memory address is aligned.
12483 (*vsx_le_perm_load_v16qi): Likewise.
12484 (four unnamed splitters): Modify to select the stvx instruction
12485 when memory is aligned.
12487 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12489 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12490 which appears in the queue multiple times.
12492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12493 Alan Hayward <alan.hayward@arm.com>
12494 David Sherwood <david.sherwood@arm.com>
12496 * tree-vectorizer.h (vec_lower_bound): New structure.
12497 (_loop_vec_info): Add check_nonzero and lower_bounds.
12498 (LOOP_VINFO_CHECK_NONZERO): New macro.
12499 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12500 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12501 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12502 fields. Make seg_len the distance travelled, not including the
12504 (dr_direction_indicator): Declare.
12505 (dr_zero_step_indicator): Likewise.
12506 (dr_known_forward_stride_p): Likewise.
12507 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12509 (runtime_alias_check_p): Allow runtime alias checks with
12511 (operator ==): Compare access_size and align.
12512 (prune_runtime_alias_test_list): Rework for new distinction between
12513 the access_size and seg_len.
12514 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12516 (get_segment_min_max): New function.
12517 (create_intersect_range_checks): Use it.
12518 (dr_step_indicator): New function.
12519 (dr_direction_indicator): Likewise.
12520 (dr_zero_step_indicator): Likewise.
12521 (dr_known_forward_stride_p): Likewise.
12522 * tree-loop-distribution.c (data_ref_segment_size): Return
12523 DR_STEP * (niters - 1).
12524 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12526 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12527 (vect_preserves_scalar_order_p): New function, split out from...
12528 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12529 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12530 (vect_vfa_access_size): New function.
12531 (vect_vfa_align): Likewise.
12532 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12533 (dump_lower_bound): New function.
12534 (vect_check_lower_bound): Likewise.
12535 (vect_small_gap_p): Likewise.
12536 (vectorizable_with_step_bound_p): Likewise.
12537 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12538 depencies if the vectorization factor is 1. Convert the checks
12539 for nonzero steps into checks on the bounds of DR_STEP. Try using
12540 a bunds check for variable steps if the minimum required step is
12541 relatively small. Update calls to the dr_with_seg_len
12542 constructor and to vect_compile_time_alias.
12543 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12545 (vect_loop_versioning): Call it.
12546 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12548 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12551 Alan Hayward <alan.hayward@arm.com>
12552 David Sherwood <david.sherwood@arm.com>
12554 * doc/sourcebuild.texi (vect_scatter_store): Document.
12555 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12557 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12559 * genopinit.c (main): Add supports_vec_scatter_store and
12560 supports_vec_scatter_store_cached to target_optabs.
12561 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12562 IFN_MASK_SCATTER_STORE.
12563 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12565 * internal-fn.h (internal_store_fn_p): Declare.
12566 (internal_fn_stored_value_index): Likewise.
12567 * internal-fn.c (scatter_store_direct): New macro.
12568 (expand_scatter_store_optab_fn): New function.
12569 (direct_scatter_store_optab_supported_p): New macro.
12570 (internal_store_fn_p): New function.
12571 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12572 IFN_MASK_SCATTER_STORE.
12573 (internal_fn_mask_index): Likewise.
12574 (internal_fn_stored_value_index): New function.
12575 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12576 for scatter stores.
12577 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12578 * optabs-query.c (supports_vec_scatter_store_p): New function.
12579 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12580 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12581 true for scatter stores.
12582 (vect_gather_scatter_fn_p): Handle scatter stores too.
12583 (vect_check_gather_scatter): Consider using scatter stores if
12584 supports_vec_scatter_store_p.
12585 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12586 scatter stores too.
12587 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12588 internal_fn_stored_value_index.
12589 (check_load_store_masking): Handle scatter stores too.
12590 (vect_get_store_rhs): Make public.
12591 (vectorizable_call): Use internal_store_fn_p.
12592 (vectorizable_store): Handle scatter store internal functions.
12593 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12594 when deciding whether the end of the group has been reached.
12595 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12596 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12597 (mask_scatter_store<mode>): New insns.
12599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12600 Alan Hayward <alan.hayward@arm.com>
12601 David Sherwood <david.sherwood@arm.com>
12603 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12604 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12605 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12607 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12608 Use vect_truncate_gather_scatter_offset if we can't treat the
12609 operation as a normal gather load or scatter store.
12610 (get_group_load_store_type): Take the gather_scatter_info
12611 as argument. Try using a gather load or scatter store for
12612 single-element groups.
12613 (get_load_store_type): Update calls to get_group_load_store_type
12614 and vect_use_strided_gather_scatters_p.
12616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12617 Alan Hayward <alan.hayward@arm.com>
12618 David Sherwood <david.sherwood@arm.com>
12620 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12621 optional tree argument.
12622 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12624 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12625 but continue to use the current value as a fallback.
12626 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12627 to compare the updates.
12628 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12629 (get_load_store_type): Use it when handling a strided access.
12630 (vect_get_strided_load_store_ops): New function.
12631 (vect_get_data_ptr_increment): Likewise.
12632 (vectorizable_load): Handle strided gather loads. Always pass
12633 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12636 Alan Hayward <alan.hayward@arm.com>
12637 David Sherwood <david.sherwood@arm.com>
12639 * doc/md.texi (gather_load@var{m}): Document.
12640 (mask_gather_load@var{m}): Likewise.
12641 * genopinit.c (main): Add supports_vec_gather_load and
12642 supports_vec_gather_load_cached to target_optabs.
12643 * optabs-tree.c (init_tree_optimization_optabs): Use
12644 ggc_cleared_alloc to allocate target_optabs.
12645 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12646 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12648 * internal-fn.h (internal_load_fn_p): Declare.
12649 (internal_gather_scatter_fn_p): Likewise.
12650 (internal_fn_mask_index): Likewise.
12651 (internal_gather_scatter_fn_supported_p): Likewise.
12652 * internal-fn.c (gather_load_direct): New macro.
12653 (expand_gather_load_optab_fn): New function.
12654 (direct_gather_load_optab_supported_p): New macro.
12655 (direct_internal_fn_optab): New function.
12656 (internal_load_fn_p): Likewise.
12657 (internal_gather_scatter_fn_p): Likewise.
12658 (internal_fn_mask_index): Likewise.
12659 (internal_gather_scatter_fn_supported_p): Likewise.
12660 * optabs-query.c (supports_at_least_one_mode_p): New function.
12661 (supports_vec_gather_load_p): Likewise.
12662 * optabs-query.h (supports_vec_gather_load_p): Declare.
12663 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12664 and memory_type field.
12665 (NUM_PATTERNS): Bump to 15.
12666 * tree-vect-data-refs.c: Include internal-fn.h.
12667 (vect_gather_scatter_fn_p): New function.
12668 (vect_describe_gather_scatter_call): Likewise.
12669 (vect_check_gather_scatter): Try using internal functions for
12670 gather loads. Recognize existing calls to a gather load function.
12671 (vect_analyze_data_refs): Consider using gather loads if
12672 supports_vec_gather_load_p.
12673 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12674 (vect_get_gather_scatter_offset_type): Likewise.
12675 (vect_convert_mask_for_vectype): Likewise.
12676 (vect_add_conversion_to_patterm): Likewise.
12677 (vect_try_gather_scatter_pattern): Likewise.
12678 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12679 (vect_vect_recog_func_ptrs): Add it.
12680 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12681 internal_fn_mask_index and internal_gather_scatter_fn_p.
12682 (check_load_store_masking): Take the gather_scatter_info as an
12683 argument and handle gather loads.
12684 (vect_get_gather_scatter_ops): New function.
12685 (vectorizable_call): Check internal_load_fn_p.
12686 (vectorizable_load): Likewise. Handle gather load internal
12688 (vectorizable_store): Update call to check_load_store_masking.
12689 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12690 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12691 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12692 (aarch64_gather_scale_operand_d): New predicates.
12693 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12694 (mask_gather_load<mode>): New insns.
12696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12697 Alan Hayward <alan.hayward@arm.com>
12698 David Sherwood <david.sherwood@arm.com>
12700 * optabs.def (fold_left_plus_optab): New optab.
12701 * doc/md.texi (fold_left_plus_@var{m}): Document.
12702 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12703 * internal-fn.c (fold_left_direct): Define.
12704 (expand_fold_left_optab_fn): Likewise.
12705 (direct_fold_left_optab_supported_p): Likewise.
12706 * fold-const-call.c (fold_const_fold_left): New function.
12707 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12708 * tree-parloops.c (valid_reduction_p): New function.
12709 (gather_scalar_reductions): Use it.
12710 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12711 (vect_finish_replace_stmt): Declare.
12712 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12713 (needs_fold_left_reduction_p): New function, split out from...
12714 (vect_is_simple_reduction): ...here. Accept reductions that
12715 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12716 (vect_force_simple_reduction): Also store the reduction type in
12717 the assignment's STMT_VINFO_REDUC_TYPE.
12718 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12719 (merge_with_identity): New function.
12720 (vect_expand_fold_left): Likewise.
12721 (vectorize_fold_left_reduction): Likewise.
12722 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12723 scalar phi in place for it. Check for target support and reject
12724 cases that would reassociate the operation. Defer the transform
12725 phase to vectorize_fold_left_reduction.
12726 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12727 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12728 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12732 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12733 call to ifc_temp_var.
12735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12736 Alan Hayward <alan.hayward@arm.com>
12737 David Sherwood <david.sherwood@arm.com>
12739 * target.def (legitimize_address_displacement): Take the original
12740 offset as a poly_int.
12741 * targhooks.h (default_legitimize_address_displacement): Update
12743 * targhooks.c (default_legitimize_address_displacement): Likewise.
12744 * doc/tm.texi: Regenerate.
12745 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12746 as an argument, moving assert of ad->disp == ad->disp_term to...
12747 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12748 Try calling targetm.legitimize_address_displacement before expanding
12749 the address rather than afterwards, and adjust for the new interface.
12750 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12751 Match the new hook interface. Handle SVE addresses.
12752 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12753 new hook interface.
12755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12757 * Makefile.in (OBJS): Add early-remat.o.
12758 * target.def (select_early_remat_modes): New hook.
12759 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12760 * doc/tm.texi: Regenerate.
12761 * targhooks.h (default_select_early_remat_modes): Declare.
12762 * targhooks.c (default_select_early_remat_modes): New function.
12763 * timevar.def (TV_EARLY_REMAT): New timevar.
12764 * passes.def (pass_early_remat): New pass.
12765 * tree-pass.h (make_pass_early_remat): Declare.
12766 * early-remat.c: New file.
12767 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12769 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12772 Alan Hayward <alan.hayward@arm.com>
12773 David Sherwood <david.sherwood@arm.com>
12775 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12776 vfm1 with a bound_epilog parameter.
12777 (vect_do_peeling): Update calls accordingly, and move the prologue
12778 call earlier in the function. Treat the base bound_epilog as 0 for
12779 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12780 this base when peeling for gaps.
12781 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12782 with fully-masked loops.
12783 (vect_estimate_min_profitable_iters): Handle the single peeled
12784 iteration in that case.
12786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12787 Alan Hayward <alan.hayward@arm.com>
12788 David Sherwood <david.sherwood@arm.com>
12790 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12791 single-element interleaving even if the size is not a power of 2.
12792 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12793 accesses for single-element interleaving if the group size is
12796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12797 Alan Hayward <alan.hayward@arm.com>
12798 David Sherwood <david.sherwood@arm.com>
12800 * doc/md.texi (fold_extract_last_@var{m}): Document.
12801 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12802 * optabs.def (fold_extract_last_optab): New optab.
12803 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12804 * internal-fn.c (fold_extract_direct): New macro.
12805 (expand_fold_extract_optab_fn): Likewise.
12806 (direct_fold_extract_optab_supported_p): Likewise.
12807 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12808 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12809 EXTRACT_LAST_REDUCTION.
12810 (get_initial_def_for_reduction): Do not create an initial vector
12811 for EXTRACT_LAST_REDUCTION reductions.
12812 (vectorizable_reduction): Leave the scalar phi in place for
12813 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12814 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12815 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12816 transform phase to vectorizable_condition.
12817 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12819 (vect_finish_stmt_generation): ...here.
12820 (vect_finish_replace_stmt): New function.
12821 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12822 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12824 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12827 Alan Hayward <alan.hayward@arm.com>
12828 David Sherwood <david.sherwood@arm.com>
12830 * doc/md.texi (extract_last_@var{m}): Document.
12831 * optabs.def (extract_last_optab): New optab.
12832 * internal-fn.def (EXTRACT_LAST): New internal function.
12833 * internal-fn.c (cond_unary_direct): New macro.
12834 (expand_cond_unary_optab_fn): Likewise.
12835 (direct_cond_unary_optab_supported_p): Likewise.
12836 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12837 loops using EXTRACT_LAST.
12838 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12839 (extract_last_<mode>): ...this optab.
12840 (vec_extract<mode><Vel>): Update accordingly.
12842 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12843 Alan Hayward <alan.hayward@arm.com>
12844 David Sherwood <david.sherwood@arm.com>
12846 * target.def (empty_mask_is_expensive): New hook.
12847 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12848 * doc/tm.texi: Regenerate.
12849 * targhooks.h (default_empty_mask_is_expensive): Declare.
12850 * targhooks.c (default_empty_mask_is_expensive): New function.
12851 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12852 if the target says that empty masks are expensive.
12853 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12855 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12858 Alan Hayward <alan.hayward@arm.com>
12859 David Sherwood <david.sherwood@arm.com>
12861 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12862 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12863 (vect_use_loop_mask_for_alignment_p): New function.
12864 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12865 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12866 niters_skip argument. Make sure that the first niters_skip elements
12867 of the first iteration are inactive.
12868 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12869 Update call to vect_set_loop_masks_directly.
12870 (get_misalign_in_elems): New function, split out from...
12871 (vect_gen_prolog_loop_niters): ...here.
12872 (vect_update_init_of_dr): Take a code argument that specifies whether
12873 the adjustment should be added or subtracted.
12874 (vect_update_init_of_drs): Likewise.
12875 (vect_prepare_for_masked_peels): New function.
12876 (vect_do_peeling): Skip prologue peeling if we're using a mask
12877 instead. Update call to vect_update_inits_of_drs.
12878 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12880 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12881 alignment. Do not include the number of peeled iterations in
12882 the minimum threshold in that case.
12883 (vectorizable_induction): Adjust the start value down by
12884 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12885 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12886 Take the number of skipped iterations into account when calculating
12888 * tree-vect-stmts.c (vect_gen_while_not): New function.
12890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12891 Alan Hayward <alan.hayward@arm.com>
12892 David Sherwood <david.sherwood@arm.com>
12894 * doc/sourcebuild.texi (vect_fully_masked): Document.
12895 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12896 default value to 0.
12897 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12899 (vect_analyze_loop_2): ...here. Don't check the vectorization
12900 factor against the number of loop iterations if the loop is
12903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12904 Alan Hayward <alan.hayward@arm.com>
12905 David Sherwood <david.sherwood@arm.com>
12907 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12908 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12909 (dump_groups): Update accordingly.
12910 (iv_use::mem_type): New member variable.
12911 (address_p): New function.
12912 (record_use): Add a mem_type argument and initialize the new
12914 (record_group_use): Add a mem_type argument. Use address_p.
12915 Remove obsolete null checks of base_object. Update call to record_use.
12916 (find_interesting_uses_op): Update call to record_group_use.
12917 (find_interesting_uses_cond): Likewise.
12918 (find_interesting_uses_address): Likewise.
12919 (get_mem_type_for_internal_fn): New function.
12920 (find_address_like_use): Likewise.
12921 (find_interesting_uses_stmt): Try find_address_like_use before
12922 calling find_interesting_uses_op.
12923 (addr_offset_valid_p): Use the iv mem_type field as the type
12924 of the addressed memory.
12925 (add_autoinc_candidates): Likewise.
12926 (get_address_cost): Likewise.
12927 (split_small_address_groups_p): Use address_p.
12928 (split_address_groups): Likewise.
12929 (add_iv_candidate_for_use): Likewise.
12930 (autoinc_possible_for_pair): Likewise.
12931 (rewrite_groups): Likewise.
12932 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12933 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12934 (get_alias_ptr_type_for_ptr_address): New function.
12935 (rewrite_use_address): Rewrite address uses in calls that were
12936 identified by find_address_like_use.
12938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12939 Alan Hayward <alan.hayward@arm.com>
12940 David Sherwood <david.sherwood@arm.com>
12942 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12944 * gimple-expr.h (is_gimple_addressable: Likewise.
12945 * gimple-expr.c (is_gimple_address): Likewise.
12946 * internal-fn.c (expand_call_mem_ref): New function.
12947 (expand_mask_load_optab_fn): Use it.
12948 (expand_mask_store_optab_fn): Likewise.
12950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12951 Alan Hayward <alan.hayward@arm.com>
12952 David Sherwood <david.sherwood@arm.com>
12954 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12955 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12956 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12957 (cond_umax@var{mode}): Document.
12958 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12959 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12960 (cond_umin_optab, cond_umax_optab): New optabs.
12961 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12962 (COND_IOR, COND_XOR): New internal functions.
12963 * internal-fn.h (get_conditional_internal_fn): Declare.
12964 * internal-fn.c (cond_binary_direct): New macro.
12965 (expand_cond_binary_optab_fn): Likewise.
12966 (direct_cond_binary_optab_supported_p): Likewise.
12967 (get_conditional_internal_fn): New function.
12968 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12969 Cope with reduction statements that are vectorized as calls rather
12971 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12972 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12973 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12974 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12975 (UNSPEC_COND_EOR): New unspecs.
12976 (optab): Add mappings for them.
12977 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12978 (sve_int_op, sve_fp_op): New int attributes.
12980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12981 Alan Hayward <alan.hayward@arm.com>
12982 David Sherwood <david.sherwood@arm.com>
12984 * optabs.def (while_ult_optab): New optab.
12985 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12986 * internal-fn.def (WHILE_ULT): New internal function.
12987 * internal-fn.h (direct_internal_fn_supported_p): New override
12988 that takes two types as argument.
12989 * internal-fn.c (while_direct): New macro.
12990 (expand_while_optab_fn): New function.
12991 (convert_optab_supported_p): Likewise.
12992 (direct_while_optab_supported_p): New macro.
12993 * wide-int.h (wi::udiv_ceil): New function.
12994 * tree-vectorizer.h (rgroup_masks): New structure.
12995 (vec_loop_masks): New typedef.
12996 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
12997 and fully_masked_p.
12998 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
12999 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13000 (vect_max_vf): New function.
13001 (slpeel_make_loop_iterate_ntimes): Delete.
13002 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13003 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13004 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13005 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13006 internal-fn.h, stor-layout.h and optabs-query.h.
13007 (vect_set_loop_mask): New function.
13008 (add_preheader_seq): Likewise.
13009 (add_header_seq): Likewise.
13010 (interleave_supported_p): Likewise.
13011 (vect_maybe_permute_loop_masks): Likewise.
13012 (vect_set_loop_masks_directly): Likewise.
13013 (vect_set_loop_condition_masked): Likewise.
13014 (vect_set_loop_condition_unmasked): New function, split out from
13015 slpeel_make_loop_iterate_ntimes.
13016 (slpeel_make_loop_iterate_ntimes): Rename to..
13017 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13018 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13019 (vect_do_peeling): Update call accordingly.
13020 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13022 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13023 mask_compare_type, can_fully_mask_p and fully_masked_p.
13024 (release_vec_loop_masks): New function.
13025 (_loop_vec_info): Use it to free the loop masks.
13026 (can_produce_all_loop_masks_p): New function.
13027 (vect_get_max_nscalars_per_iter): Likewise.
13028 (vect_verify_full_masking): Likewise.
13029 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13030 retries, and free the mask rgroups before retrying. Check loop-wide
13031 reasons for disallowing fully-masked loops. Make the final decision
13032 about whether use a fully-masked loop or not.
13033 (vect_estimate_min_profitable_iters): Do not assume that peeling
13034 for the number of iterations will be needed for fully-masked loops.
13035 (vectorizable_reduction): Disable fully-masked loops.
13036 (vectorizable_live_operation): Likewise.
13037 (vect_halve_mask_nunits): New function.
13038 (vect_double_mask_nunits): Likewise.
13039 (vect_record_loop_mask): Likewise.
13040 (vect_get_loop_mask): Likewise.
13041 (vect_transform_loop): Handle the case in which the final loop
13042 iteration might handle a partial vector. Call vect_set_loop_condition
13043 instead of slpeel_make_loop_iterate_ntimes.
13044 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13045 (check_load_store_masking): New function.
13046 (prepare_load_store_mask): Likewise.
13047 (vectorizable_store): Handle fully-masked loops.
13048 (vectorizable_load): Likewise.
13049 (supportable_widening_operation): Use vect_halve_mask_nunits for
13051 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13052 (vect_gen_while): New function.
13053 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13054 (aarch64_uqdec<mode>): New insn.
13056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13057 Alan Hayward <alan.hayward@arm.com>
13058 David Sherwood <david.sherwood@arm.com>
13060 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13061 (reduc_xor_scal_optab): New optabs.
13062 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13063 (reduc_xor_scal_@var{m}): Document.
13064 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13065 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13066 internal functions.
13067 * fold-const-call.c (fold_const_call): Handle them.
13068 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13069 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13070 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13071 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13072 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13073 (UNSPEC_XORV): New unspecs.
13074 (optab): Add entries for them.
13075 (BITWISEV): New int iterator.
13076 (bit_reduc_op): New int attributes.
13078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13079 Alan Hayward <alan.hayward@arm.com>
13080 David Sherwood <david.sherwood@arm.com>
13082 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13083 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13084 * optabs.def (vec_shl_insert_optab): New optab.
13085 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13086 (duplicate_and_interleave): Likewise.
13087 * tree-vect-loop.c: Include internal-fn.h.
13088 (neutral_op_for_slp_reduction): New function, split out from
13089 get_initial_defs_for_reduction.
13090 (get_initial_def_for_reduction): Handle option 2 for variable-length
13091 vectors by loading the neutral value into a vector and then shifting
13092 the initial value into element 0.
13093 (get_initial_defs_for_reduction): Replace the code argument with
13094 the neutral value calculated by neutral_op_for_slp_reduction.
13095 Use gimple_build_vector for constant-length vectors.
13096 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13097 but the first group_size elements have a neutral value.
13098 Use duplicate_and_interleave otherwise.
13099 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13100 Update call to get_initial_defs_for_reduction. Handle SLP
13101 reductions for variable-length vectors by creating one vector
13102 result for each scalar result, with the elements associated
13103 with other scalar results stubbed out with the neutral value.
13104 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13105 Require IFN_VEC_SHL_INSERT for double reductions on
13106 variable-length vectors, or SLP reductions that have
13107 a neutral value. Require can_duplicate_and_interleave_p
13108 support for variable-length unchained SLP reductions if there
13109 is no neutral value, such as for MIN/MAX reductions. Also require
13110 the number of vector elements to be a multiple of the number of
13111 SLP statements when doing variable-length unchained SLP reductions.
13112 Update call to vect_create_epilog_for_reduction.
13113 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13114 and remove initial values.
13115 (duplicate_and_interleave): Make public.
13116 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13117 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13120 Alan Hayward <alan.hayward@arm.com>
13121 David Sherwood <david.sherwood@arm.com>
13123 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13124 (can_duplicate_and_interleave_p): New function.
13125 (vect_get_and_check_slp_defs): Take the vector of statements
13126 rather than just the current one. Remove excess parentheses.
13127 Restriction rejectinon of vect_constant_def and vect_external_def
13128 for variable-length vectors to boolean types, or types for which
13129 can_duplicate_and_interleave_p is false.
13130 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13131 (duplicate_and_interleave): New function.
13132 (vect_get_constant_vectors): Use gimple_build_vector for
13133 constant-length vectors and suitable variable-length constant
13134 vectors. Use duplicate_and_interleave for other variable-length
13135 vectors. Don't defer the update when inserting new statements.
13137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13138 Alan Hayward <alan.hayward@arm.com>
13139 David Sherwood <david.sherwood@arm.com>
13141 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13142 min_profitable_iters doesn't go negative.
13144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13145 Alan Hayward <alan.hayward@arm.com>
13146 David Sherwood <david.sherwood@arm.com>
13148 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13149 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13150 * optabs.def (vec_mask_load_lanes_optab): New optab.
13151 (vec_mask_store_lanes_optab): Likewise.
13152 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13153 (MASK_STORE_LANES): Likewise.
13154 * internal-fn.c (mask_load_lanes_direct): New macro.
13155 (mask_store_lanes_direct): Likewise.
13156 (expand_mask_load_optab_fn): Handle masked operations.
13157 (expand_mask_load_lanes_optab_fn): New macro.
13158 (expand_mask_store_optab_fn): Handle masked operations.
13159 (expand_mask_store_lanes_optab_fn): New macro.
13160 (direct_mask_load_lanes_optab_supported_p): Likewise.
13161 (direct_mask_store_lanes_optab_supported_p): Likewise.
13162 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13164 (vect_load_lanes_supported): Likewise.
13165 * tree-vect-data-refs.c (strip_conversion): New function.
13166 (can_group_stmts_p): Likewise.
13167 (vect_analyze_data_ref_accesses): Use it instead of checking
13168 for a pair of assignments.
13169 (vect_store_lanes_supported): Take a masked_p parameter.
13170 (vect_load_lanes_supported): Likewise.
13171 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13172 vect_store_lanes_supported and vect_load_lanes_supported.
13173 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13174 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13175 parameter. Don't allow gaps for masked accesses.
13176 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13177 and vect_load_lanes_supported.
13178 (get_load_store_type): Take a masked_p parameter and update
13179 call to get_group_load_store_type.
13180 (vectorizable_store): Update call to get_load_store_type.
13181 Handle IFN_MASK_STORE_LANES.
13182 (vectorizable_load): Update call to get_load_store_type.
13183 Handle IFN_MASK_LOAD_LANES.
13185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13186 Alan Hayward <alan.hayward@arm.com>
13187 David Sherwood <david.sherwood@arm.com>
13189 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13191 * config/aarch64/aarch64-protos.h
13192 (aarch64_sve_struct_memory_operand_p): Declare.
13193 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13194 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13195 (VPRED, vpred): Handle SVE structure modes.
13196 * config/aarch64/constraints.md (Utx): New constraint.
13197 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13198 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13199 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13200 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13201 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13202 structure modes. Split into pieces after RA.
13203 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13204 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13206 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13207 SVE structure modes.
13208 (aarch64_classify_address): Likewise.
13209 (sizetochar): Move earlier in file.
13210 (aarch64_print_operand): Handle SVE register lists.
13211 (aarch64_array_mode): New function.
13212 (aarch64_sve_struct_memory_operand_p): Likewise.
13213 (TARGET_ARRAY_MODE): Redefine.
13215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13216 Alan Hayward <alan.hayward@arm.com>
13217 David Sherwood <david.sherwood@arm.com>
13219 * target.def (array_mode): New target hook.
13220 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13221 * doc/tm.texi: Regenerate.
13222 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13223 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13224 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13225 targetm.array_mode.
13226 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13230 Alan Hayward <alan.hayward@arm.com>
13231 David Sherwood <david.sherwood@arm.com>
13233 * fold-const.c (fold_binary_loc): Check the argument types
13234 rather than the result type when testing for a vector operation.
13236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13238 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13239 * doc/tm.texi: Regenerate.
13241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13242 Alan Hayward <alan.hayward@arm.com>
13243 David Sherwood <david.sherwood@arm.com>
13245 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13246 (sve): Document new AArch64 extension.
13247 * doc/md.texi (w): Extend the description of the AArch64
13248 constraint to include SVE vectors.
13249 (Upl, Upa): Document new AArch64 predicate constraints.
13250 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13252 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13253 (msve-vector-bits=): New option.
13254 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13255 SVE when these are disabled.
13256 (sve): New extension.
13257 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13258 modes. Adjust their number of units based on aarch64_sve_vg.
13259 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13260 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13261 aarch64_addr_query_type.
13262 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13263 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13264 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13265 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13266 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13267 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13268 (aarch64_simd_imm_zero_p): Delete.
13269 (aarch64_check_zero_based_sve_index_immediate): Declare.
13270 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13271 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13272 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13273 (aarch64_sve_float_mul_immediate_p): Likewise.
13274 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13275 rather than an rtx.
13276 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13277 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13278 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13279 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13280 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13281 (aarch64_regmode_natural_size): Likewise.
13282 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13283 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13285 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13286 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13287 for VG and the SVE predicate registers.
13288 (V_ALIASES): Add a "z"-prefixed alias.
13289 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13290 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13291 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13292 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13293 (REG_CLASS_NAMES): Add entries for them.
13294 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13295 and the predicate registers.
13296 (aarch64_sve_vg): Declare.
13297 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13298 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13299 (REGMODE_NATURAL_SIZE): Define.
13300 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13302 * config/aarch64/aarch64.c: Include cfgrtl.h.
13303 (simd_immediate_info): Add a constructor for series vectors,
13304 and an associated step field.
13305 (aarch64_sve_vg): New variable.
13306 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13307 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13308 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13309 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13310 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13311 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13312 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13313 (aarch64_get_mask_mode): New functions.
13314 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13315 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13316 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13317 predicate modes and predicate registers. Explicitly restrict
13318 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13319 to store a vector mode if it is recognized by
13320 aarch64_classify_vector_mode.
13321 (aarch64_regmode_natural_size): New function.
13322 (aarch64_hard_regno_caller_save_mode): Return the original mode
13324 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13325 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13326 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13327 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13329 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13330 does not overlap dest if the function is frame-related. Handle
13332 (aarch64_split_add_offset): New function.
13333 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13334 them aarch64_add_offset.
13335 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13336 and update call to aarch64_sub_sp.
13337 (aarch64_add_cfa_expression): New function.
13338 (aarch64_expand_prologue): Pass extra temporary registers to the
13339 functions above. Handle the case in which we need to emit new
13340 DW_CFA_expressions for registers that were originally saved
13341 relative to the stack pointer, but now have to be expressed
13342 relative to the frame pointer.
13343 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13345 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13346 IP0 and IP1 values for SVE frames.
13347 (aarch64_expand_vec_series): New function.
13348 (aarch64_expand_sve_widened_duplicate): Likewise.
13349 (aarch64_expand_sve_const_vector): Likewise.
13350 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13351 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13352 into the register, rather than emitting a SET directly.
13353 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13354 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13355 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13356 (offset_9bit_signed_scaled_p): New functions.
13357 (aarch64_replicate_bitmask_imm): New function.
13358 (aarch64_bitmask_imm): Use it.
13359 (aarch64_cannot_force_const_mem): Reject expressions involving
13360 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13361 (aarch64_classify_index): Handle SVE indices, by requiring
13362 a plain register index with a scale that matches the element size.
13363 (aarch64_classify_address): Handle SVE addresses. Assert that
13364 the mode of the address is VOIDmode or an integer mode.
13365 Update call to aarch64_classify_symbol.
13366 (aarch64_classify_symbolic_expression): Update call to
13367 aarch64_classify_symbol.
13368 (aarch64_const_vec_all_in_range_p): New function.
13369 (aarch64_print_vector_float_operand): Likewise.
13370 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13371 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13372 and the FP immediates 1.0 and 0.5.
13373 (aarch64_print_address_internal): Handle SVE addresses.
13374 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13375 (aarch64_regno_regclass): Handle predicate registers.
13376 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13378 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13379 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13380 (aarch64_convert_sve_vector_bits): New function.
13381 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13382 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13383 rather than an rtx.
13384 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13385 Handle SVE vector and predicate modes. Accept VL-based constants
13386 that need only one temporary register, and VL offsets that require
13387 no temporary registers.
13388 (aarch64_conditional_register_usage): Mark the predicate registers
13389 as fixed if SVE isn't available.
13390 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13391 Return true for SVE vector and predicate modes.
13392 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13393 rather than an unsigned int. Handle SVE modes.
13394 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13396 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13398 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13399 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13400 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13401 (aarch64_sve_float_mul_immediate_p): New functions.
13402 (aarch64_sve_valid_immediate): New function.
13403 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13404 Explicitly reject structure modes. Check for INDEX constants.
13405 Handle PTRUE and PFALSE constants.
13406 (aarch64_check_zero_based_sve_index_immediate): New function.
13407 (aarch64_simd_imm_zero_p): Delete.
13408 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13409 vector modes. Accept constants in the range of CNT[BHWD].
13410 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13411 ask for an Advanced SIMD mode.
13412 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13413 (aarch64_simd_vector_alignment): Handle SVE predicates.
13414 (aarch64_vectorize_preferred_vector_alignment): New function.
13415 (aarch64_simd_vector_alignment_reachable): Use it instead of
13417 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13418 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13420 (MAX_VECT_LEN): Delete.
13421 (expand_vec_perm_d): Add a vec_flags field.
13422 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13423 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13424 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13426 (aarch64_evpc_rev): Rename to...
13427 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13428 (aarch64_evpc_rev_global): New function.
13429 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13430 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13432 (aarch64_evpc_sve_tbl): New function.
13433 (aarch64_expand_vec_perm_const_1): Update after rename of
13434 aarch64_evpc_rev. Handle SVE permutes too, trying
13435 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13436 than aarch64_evpc_tbl.
13437 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13438 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13439 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13440 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13441 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13442 (aarch64_expand_sve_vcond): New functions.
13443 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13444 of aarch64_vector_mode_p.
13445 (aarch64_dwarf_poly_indeterminate_value): New function.
13446 (aarch64_compute_pressure_classes): Likewise.
13447 (aarch64_can_change_mode_class): Likewise.
13448 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13449 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13450 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13451 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13452 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13453 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13454 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13455 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13457 (Dn, Dl, Dr): Accept const as well as const_vector.
13458 (Dz): Likewise. Compare against CONST0_RTX.
13459 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13460 of "vector" where appropriate.
13461 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13462 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13463 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13464 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13465 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13466 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13467 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13468 (v_int_equiv): Extend to SVE modes.
13469 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13471 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13472 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13473 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13474 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13475 (SVE_COND_FP_CMP): New int iterators.
13476 (perm_hilo): Handle the new unpack unspecs.
13477 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13479 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13480 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13481 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13482 (aarch64_equality_operator, aarch64_constant_vector_operand)
13483 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13484 (aarch64_sve_nonimmediate_operand): Likewise.
13485 (aarch64_sve_general_operand): Likewise.
13486 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13487 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13488 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13489 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13490 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13491 (aarch64_sve_float_arith_immediate): Likewise.
13492 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13493 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13494 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13495 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13496 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13497 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13498 (aarch64_sve_float_arith_operand): Likewise.
13499 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13500 (aarch64_sve_float_mul_operand): Likewise.
13501 (aarch64_sve_vec_perm_operand): Likewise.
13502 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13503 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13504 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13505 as well as const_vector.
13506 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13507 in file. Use CONST0_RTX and CONSTM1_RTX.
13508 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13509 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13510 Use aarch64_simd_imm_zero.
13511 * config/aarch64/aarch64-sve.md: New file.
13512 * config/aarch64/aarch64.md: Include it.
13513 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13514 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13515 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13516 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13517 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13518 (sve): New attribute.
13519 (enabled): Disable instructions with the sve attribute unless
13521 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13522 aarch64_expand_mov_immediate.
13523 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13524 CNT[BHSD] immediates.
13525 (movti): Split CONST_POLY_INT moves into two halves.
13526 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13527 Split additions that need a temporary here if the destination
13528 is the stack pointer.
13529 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13530 (*add<mode>3_poly_1): New instruction.
13531 (set_clobber_cc): New expander.
13533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13535 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13536 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13537 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13538 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13539 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13540 Change innermode from fixed_mode_size to machine_mode.
13541 (simplify_subreg): Update call accordingly. Handle a constant-sized
13542 subreg of a variable-length CONST_VECTOR.
13544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13545 Alan Hayward <alan.hayward@arm.com>
13546 David Sherwood <david.sherwood@arm.com>
13548 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13549 (add_offset_to_base): New function, split out from...
13550 (create_mem_ref): ...here. When handling a scale other than 1,
13551 check first whether the address is valid without the offset.
13552 Add it into the base if so, leaving the index and scale as-is.
13554 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13557 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13558 fold_for_warn before checking if arg2 is INTEGER_CST.
13560 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13562 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13563 (store_multiple_operation): Delete.
13564 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13565 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13566 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13567 guarded by TARGET_STRING.
13568 (rs6000_output_load_multiple): Delete.
13569 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13570 OPTION_MASK_STRING / TARGET_STRING handling.
13571 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13572 (const rs6000_opt_masks) <"string">: Change mask to 0.
13573 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13574 (MASK_STRING): Delete.
13575 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13577 (load_multiple): Delete.
13584 (store_multiple): Delete.
13591 (movmemsi_8reg): Delete.
13592 (corresponding unnamed define_insn): Delete.
13593 (movmemsi_6reg): Delete.
13594 (corresponding unnamed define_insn): Delete.
13595 (movmemsi_4reg): Delete.
13596 (corresponding unnamed define_insn): Delete.
13597 (movmemsi_2reg): Delete.
13598 (corresponding unnamed define_insn): Delete.
13599 (movmemsi_1reg): Delete.
13600 (corresponding unnamed define_insn): Delete.
13601 * config/rs6000/rs6000.opt (mno-string): New.
13602 (mstring): Replace by deprecation warning stub.
13603 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13605 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13607 * regrename.c (regrename_do_replace): If replacing the same
13608 reg multiple times, try to reuse last created gen_raw_REG.
13611 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13612 main to workaround a bug in GDB.
13614 2018-01-12 Tom de Vries <tom@codesourcery.com>
13617 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13619 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13621 PR rtl-optimization/80481
13622 * ira-color.c (get_cap_member): New function.
13623 (allocnos_conflict_by_live_ranges_p): Use it.
13624 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13625 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13627 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13630 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13631 (*saddl_se_1): Ditto.
13632 (*ssubsi_1): Ditto.
13633 (*ssubl_se_1): Ditto.
13635 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13637 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13638 rather than wi::to_widest for DR_INITs.
13639 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13640 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13641 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13643 (vect_analyze_group_access_1): Note that here.
13645 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13647 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13648 polynomial type sizes.
13650 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13652 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13653 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13654 (gimple_add_tmp_var): Likewise.
13656 2018-01-12 Martin Liska <mliska@suse.cz>
13658 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13659 (gimple_alloc_sizes): Likewise.
13660 (dump_gimple_statistics): Use PRIu64 in printf format.
13661 * gimple.h: Change uint64_t to int.
13663 2018-01-12 Martin Liska <mliska@suse.cz>
13665 * tree-core.h: Use uint64_t instead of int.
13666 * tree.c (tree_node_counts): Likewise.
13667 (tree_node_sizes): Likewise.
13668 (dump_tree_statistics): Use PRIu64 in printf format.
13670 2018-01-12 Martin Liska <mliska@suse.cz>
13672 * Makefile.in: As qsort_chk is implemented in vec.c, add
13673 vec.o to linkage of gencfn-macros.
13674 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13675 passing the info to record_node_allocation_statistics.
13676 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13678 * ggc-common.c (struct ggc_usage): Add operator== and use
13679 it in operator< and compare function.
13680 * mem-stats.h (struct mem_usage): Likewise.
13681 * vec.c (struct vec_usage): Remove operator< and compare
13682 function. Can be simply inherited.
13684 2018-01-12 Martin Jambor <mjambor@suse.cz>
13687 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13688 * tree-ssa-math-opts.c: Include domwalk.h.
13689 (convert_mult_to_fma_1): New function.
13690 (fma_transformation_info): New type.
13691 (fma_deferring_state): Likewise.
13692 (cancel_fma_deferring): New function.
13693 (result_of_phi): Likewise.
13694 (last_fma_candidate_feeds_initial_phi): Likewise.
13695 (convert_mult_to_fma): Added deferring logic, split actual
13696 transformation to convert_mult_to_fma_1.
13697 (math_opts_dom_walker): New type.
13698 (math_opts_dom_walker::after_dom_children): New method, body moved
13699 here from pass_optimize_widening_mul::execute, added deferring logic
13701 (pass_optimize_widening_mul::execute): Moved most of code to
13702 math_opts_dom_walker::after_dom_children.
13703 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13704 * config/i386/i386.c (ix86_option_override_internal): Added
13705 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13707 2018-01-12 Richard Biener <rguenther@suse.de>
13710 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13711 inline instance vars.
13713 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13716 * config/rx/rx.c (rx_is_restricted_memory_address):
13717 Handle SUBREG case.
13719 2018-01-12 Richard Biener <rguenther@suse.de>
13721 PR tree-optimization/80846
13722 * target.def (split_reduction): New target hook.
13723 * targhooks.c (default_split_reduction): New function.
13724 * targhooks.h (default_split_reduction): Declare.
13725 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13726 target requests first reduce vectors by combining low and high
13728 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13729 (get_vectype_for_scalar_type_and_size): Export.
13730 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13731 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13732 * doc/tm.texi: Regenerate.
13733 * config/i386/i386.c (ix86_split_reduction): Implement
13734 TARGET_VECTORIZE_SPLIT_REDUCTION.
13736 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13739 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13740 in PIC mode except for TARGET_VXWORKS_RTP.
13741 * config/sparc/sparc.c: Include cfgrtl.h.
13742 (TARGET_INIT_PIC_REG): Define.
13743 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13744 (sparc_pic_register_p): New predicate.
13745 (sparc_legitimate_address_p): Use it.
13746 (sparc_legitimize_pic_address): Likewise.
13747 (sparc_delegitimize_address): Likewise.
13748 (sparc_mode_dependent_address_p): Likewise.
13749 (gen_load_pcrel_sym): Remove 4th parameter.
13750 (load_got_register): Adjust call to above. Remove obsolete stuff.
13751 (sparc_expand_prologue): Do not call load_got_register here.
13752 (sparc_flat_expand_prologue): Likewise.
13753 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13754 (sparc_use_pseudo_pic_reg): New function.
13755 (sparc_init_pic_reg): Likewise.
13756 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13757 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13759 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13761 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13762 Add item for branch_cost.
13764 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13766 PR rtl-optimization/83565
13767 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13768 not extend the result to a larger mode for rotate operations.
13769 (num_sign_bit_copies1): Likewise.
13771 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13774 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13776 Use values-Xc.o for -pedantic.
13777 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13779 2018-01-12 Martin Liska <mliska@suse.cz>
13782 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13784 (possible_polymorphic_call_targets): Use it.
13785 (ipa_devirt): Likewise.
13787 2018-01-12 Martin Liska <mliska@suse.cz>
13789 * profile-count.h (enum profile_quality): Use 0 as invalid
13790 enum value of profile_quality.
13792 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13794 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13795 -mext-string options.
13797 2018-01-12 Richard Biener <rguenther@suse.de>
13799 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13800 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13801 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13803 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13805 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13807 * configure.ac (--with-long-double-format): Add support for the
13808 configuration option to change the default long double format on
13810 * config.gcc (powerpc*-linux*-*): Likewise.
13811 * configure: Regenerate.
13812 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13813 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13814 used without modification.
13816 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13818 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13819 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13820 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13821 MISC_BUILTIN_SPEC_BARRIER.
13822 (rs6000_init_builtins): Likewise.
13823 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13825 (speculation_barrier): New define_insn.
13826 * doc/extend.texi: Document __builtin_speculation_barrier.
13828 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13831 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13832 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13833 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13835 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13836 integral modes instead of "ss" and "sd".
13837 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13838 vectors with 32-bit and 64-bit elements.
13839 (vecdupssescalarmodesuffix): New mode attribute.
13840 (vec_dup<mode>): Use it.
13842 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13845 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13846 frame if argument is passed on stack.
13848 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13851 * ree.c (combine_reaching_defs): Optimize also
13852 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13853 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13855 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13857 PR middle-end/83189
13858 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13860 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13862 PR middle-end/83718
13863 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13864 after they are computed.
13866 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13868 PR tree-optimization/83695
13869 * gimple-loop-linterchange.cc
13870 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13871 reset cached scev information after interchange.
13872 (pass_linterchange::execute): Remove call to scev_reset_htab.
13874 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13876 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13877 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13878 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13879 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13880 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13881 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13882 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13883 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13884 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13885 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13886 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13887 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13888 (V_lane_reg): Likewise.
13889 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13891 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13892 (vfmal_lane_low<mode>_intrinsic,
13893 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13894 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13895 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13896 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13897 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13898 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13900 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13902 * config/arm/arm-cpus.in (fp16fml): New feature.
13903 (ALL_SIMD): Add fp16fml.
13904 (armv8.2-a): Add fp16fml as an option.
13905 (armv8.3-a): Likewise.
13906 (armv8.4-a): Add fp16fml as part of fp16.
13907 * config/arm/arm.h (TARGET_FP16FML): Define.
13908 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13910 * config/arm/arm-modes.def (V2HF): Define.
13911 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13912 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13913 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13914 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13915 vfmsl_low, vfmsl_high): New set of builtins.
13916 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13917 (vfml_op): New code attribute.
13918 (VFMLHALVES): New int iterator.
13919 (VFML, VFMLSEL): New mode attributes.
13920 (V_reg): Define mapping for V2HF.
13921 (V_hi, V_lo): New mode attributes.
13922 (VF_constraint): Likewise.
13923 (vfml_half, vfml_half_selector): New int attributes.
13924 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13926 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13927 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13929 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13930 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13931 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13932 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13934 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13935 Document new effective target and option set.
13937 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13939 * config/arm/arm-cpus.in (armv8_4): New feature.
13940 (ARMv8_4a): New fgroup.
13941 (armv8.4-a): New arch.
13942 * config/arm/arm-tables.opt: Regenerate.
13943 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13944 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13945 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13946 Add matching rules for -march=armv8.4-a and extensions.
13947 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13949 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13952 * config/rx/rx.md (BW): New mode attribute.
13953 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13955 2018-01-11 Richard Biener <rguenther@suse.de>
13957 PR tree-optimization/83435
13958 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13959 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13960 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13962 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13963 Alan Hayward <alan.hayward@arm.com>
13964 David Sherwood <david.sherwood@arm.com>
13966 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13968 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13969 (aarch64_print_address_internal): Use it to check for a zero offset.
13971 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13972 Alan Hayward <alan.hayward@arm.com>
13973 David Sherwood <david.sherwood@arm.com>
13975 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13976 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13977 Return a poly_int64 rather than a HOST_WIDE_INT.
13978 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13979 rather than a HOST_WIDE_INT.
13980 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13981 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13982 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13983 final_offset from HOST_WIDE_INT to poly_int64.
13984 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13985 to_constant when getting the number of units in an Advanced SIMD
13987 (aarch64_builtin_vectorized_function): Check for a constant number
13989 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13991 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
13992 attribute instead of GET_MODE_NUNITS.
13993 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
13994 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
13995 GET_MODE_SIZE for fixed-size registers.
13996 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
13997 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
13998 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
13999 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14000 (aarch64_print_operand, aarch64_print_address_internal)
14001 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14002 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14003 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14004 Handle polynomial GET_MODE_SIZE.
14005 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14006 wider than SImode without modification.
14007 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14008 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14009 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14010 passing and returning SVE modes.
14011 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14012 rather than GEN_INT.
14013 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14014 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14015 (aarch64_allocate_and_probe_stack_space): Likewise.
14016 (aarch64_layout_frame): Cope with polynomial offsets.
14017 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14018 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14019 polynomial offsets.
14020 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14021 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14022 poly_int64 rather than a HOST_WIDE_INT.
14023 (aarch64_get_separate_components, aarch64_process_components)
14024 (aarch64_expand_prologue, aarch64_expand_epilogue)
14025 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14026 (aarch64_anchor_offset): New function, split out from...
14027 (aarch64_legitimize_address): ...here.
14028 (aarch64_builtin_vectorization_cost): Handle polynomial
14029 TYPE_VECTOR_SUBPARTS.
14030 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14032 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14033 number of elements from the PARALLEL rather than the mode.
14034 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14035 rather than GET_MODE_BITSIZE.
14036 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14037 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14038 (aarch64_expand_vec_perm_const_1): Handle polynomial
14039 d->perm.length () and d->perm elements.
14040 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14041 Apply to_constant to d->perm elements.
14042 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14043 polynomial CONST_VECTOR_NUNITS.
14044 (aarch64_move_pointer): Take amount as a poly_int64 rather
14046 (aarch64_progress_pointer): Avoid temporary variable.
14047 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14048 the mode attribute instead of GET_MODE.
14050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14051 Alan Hayward <alan.hayward@arm.com>
14052 David Sherwood <david.sherwood@arm.com>
14054 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14055 x exists before using it.
14056 (aarch64_add_constant_internal): Rename to...
14057 (aarch64_add_offset_1): ...this. Replace regnum with separate
14058 src and dest rtxes. Handle the case in which they're different,
14059 including when the offset is zero. Replace scratchreg with an rtx.
14060 Use 2 additions if there is no spare register into which we can
14061 move a 16-bit constant.
14062 (aarch64_add_constant): Delete.
14063 (aarch64_add_offset): Replace reg with separate src and dest
14064 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14065 Use aarch64_add_offset_1.
14066 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14067 an rtx rather than an int. Take the delta as a poly_int64
14068 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14069 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14070 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14071 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14072 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14073 and aarch64_add_sp.
14074 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14075 aarch64_add_constant.
14077 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14079 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14080 Use scalar_float_mode.
14082 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14084 * config/aarch64/aarch64-simd.md
14085 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14086 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14087 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14088 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14089 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14090 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14091 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14092 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14093 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14094 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14096 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14099 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14100 targ_options->x_arm_arch_string is non NULL.
14102 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14104 * config/aarch64/aarch64.h
14105 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14107 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14110 * expmed.c (emit_store_flag_force): Swap if const op0
14111 and change VOIDmode to mode of op0.
14113 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14115 PR rtl-optimization/83761
14116 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14117 than bytes to mode_for_size.
14119 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14121 PR middle-end/83189
14122 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14123 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14126 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14128 PR middle-end/83575
14129 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14130 when in layout mode.
14131 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14132 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14135 2018-01-10 Michael Collison <michael.collison@arm.com>
14137 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14138 * config/aarch64/aarch64-option-extension.def: Add
14139 AARCH64_OPT_EXTENSION of 'fp16fml'.
14140 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14141 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14142 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14143 * config/aarch64/constraints.md (Ui7): New constraint.
14144 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14145 (VFMLA_SEL_W): Ditto.
14148 (VFMLA16_LOW): New int iterator.
14149 (VFMLA16_HIGH): Ditto.
14150 (UNSPEC_FMLAL): New unspec.
14151 (UNSPEC_FMLSL): Ditto.
14152 (UNSPEC_FMLAL2): Ditto.
14153 (UNSPEC_FMLSL2): Ditto.
14154 (f16mac): New code attribute.
14155 * config/aarch64/aarch64-simd-builtins.def
14156 (aarch64_fmlal_lowv2sf): Ditto.
14157 (aarch64_fmlsl_lowv2sf): Ditto.
14158 (aarch64_fmlalq_lowv4sf): Ditto.
14159 (aarch64_fmlslq_lowv4sf): Ditto.
14160 (aarch64_fmlal_highv2sf): Ditto.
14161 (aarch64_fmlsl_highv2sf): Ditto.
14162 (aarch64_fmlalq_highv4sf): Ditto.
14163 (aarch64_fmlslq_highv4sf): Ditto.
14164 (aarch64_fmlal_lane_lowv2sf): Ditto.
14165 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14166 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14167 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14168 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14169 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14170 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14171 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14172 (aarch64_fmlal_lane_highv2sf): Ditto.
14173 (aarch64_fmlsl_lane_highv2sf): Ditto.
14174 (aarch64_fmlal_laneq_highv2sf): Ditto.
14175 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14176 (aarch64_fmlalq_lane_highv4sf): Ditto.
14177 (aarch64_fmlsl_lane_highv4sf): Ditto.
14178 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14179 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14180 * config/aarch64/aarch64-simd.md:
14181 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14182 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14183 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14184 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14185 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14186 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14187 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14188 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14189 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14190 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14191 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14192 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14193 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14194 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14195 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14196 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14197 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14198 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14199 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14200 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14201 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14202 (vfmlsl_low_u32): Ditto.
14203 (vfmlalq_low_u32): Ditto.
14204 (vfmlslq_low_u32): Ditto.
14205 (vfmlal_high_u32): Ditto.
14206 (vfmlsl_high_u32): Ditto.
14207 (vfmlalq_high_u32): Ditto.
14208 (vfmlslq_high_u32): Ditto.
14209 (vfmlal_lane_low_u32): Ditto.
14210 (vfmlsl_lane_low_u32): Ditto.
14211 (vfmlal_laneq_low_u32): Ditto.
14212 (vfmlsl_laneq_low_u32): Ditto.
14213 (vfmlalq_lane_low_u32): Ditto.
14214 (vfmlslq_lane_low_u32): Ditto.
14215 (vfmlalq_laneq_low_u32): Ditto.
14216 (vfmlslq_laneq_low_u32): Ditto.
14217 (vfmlal_lane_high_u32): Ditto.
14218 (vfmlsl_lane_high_u32): Ditto.
14219 (vfmlal_laneq_high_u32): Ditto.
14220 (vfmlsl_laneq_high_u32): Ditto.
14221 (vfmlalq_lane_high_u32): Ditto.
14222 (vfmlslq_lane_high_u32): Ditto.
14223 (vfmlalq_laneq_high_u32): Ditto.
14224 (vfmlslq_laneq_high_u32): Ditto.
14225 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14226 (AARCH64_FL_FOR_ARCH8_4): New.
14227 (AARCH64_ISA_F16FML): New ISA flag.
14228 (TARGET_F16FML): New feature flag for fp16fml.
14229 (doc/invoke.texi): Document new fp16fml option.
14231 2018-01-10 Michael Collison <michael.collison@arm.com>
14233 * config/aarch64/aarch64-builtins.c:
14234 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14235 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14236 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14237 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14238 (AARCH64_ISA_SHA3): New ISA flag.
14239 (TARGET_SHA3): New feature flag for sha3.
14240 * config/aarch64/iterators.md (sha512_op): New int attribute.
14241 (CRYPTO_SHA512): New int iterator.
14242 (UNSPEC_SHA512H): New unspec.
14243 (UNSPEC_SHA512H2): Ditto.
14244 (UNSPEC_SHA512SU0): Ditto.
14245 (UNSPEC_SHA512SU1): Ditto.
14246 * config/aarch64/aarch64-simd-builtins.def
14247 (aarch64_crypto_sha512hqv2di): New builtin.
14248 (aarch64_crypto_sha512h2qv2di): Ditto.
14249 (aarch64_crypto_sha512su0qv2di): Ditto.
14250 (aarch64_crypto_sha512su1qv2di): Ditto.
14251 (aarch64_eor3qv8hi): Ditto.
14252 (aarch64_rax1qv2di): Ditto.
14253 (aarch64_xarqv2di): Ditto.
14254 (aarch64_bcaxqv8hi): Ditto.
14255 * config/aarch64/aarch64-simd.md:
14256 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14257 (aarch64_crypto_sha512su0qv2di): Ditto.
14258 (aarch64_crypto_sha512su1qv2di): Ditto.
14259 (aarch64_eor3qv8hi): Ditto.
14260 (aarch64_rax1qv2di): Ditto.
14261 (aarch64_xarqv2di): Ditto.
14262 (aarch64_bcaxqv8hi): Ditto.
14263 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14264 (vsha512h2q_u64): Ditto.
14265 (vsha512su0q_u64): Ditto.
14266 (vsha512su1q_u64): Ditto.
14267 (veor3q_u16): Ditto.
14268 (vrax1q_u64): Ditto.
14269 (vxarq_u64): Ditto.
14270 (vbcaxq_u16): Ditto.
14271 * config/arm/types.md (crypto_sha512): New type attribute.
14272 (crypto_sha3): Ditto.
14273 (doc/invoke.texi): Document new sha3 option.
14275 2018-01-10 Michael Collison <michael.collison@arm.com>
14277 * config/aarch64/aarch64-builtins.c:
14278 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14279 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14280 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14281 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14282 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14283 (AARCH64_ISA_SM4): New ISA flag.
14284 (TARGET_SM4): New feature flag for sm4.
14285 * config/aarch64/aarch64-simd-builtins.def
14286 (aarch64_sm3ss1qv4si): Ditto.
14287 (aarch64_sm3tt1aq4si): Ditto.
14288 (aarch64_sm3tt1bq4si): Ditto.
14289 (aarch64_sm3tt2aq4si): Ditto.
14290 (aarch64_sm3tt2bq4si): Ditto.
14291 (aarch64_sm3partw1qv4si): Ditto.
14292 (aarch64_sm3partw2qv4si): Ditto.
14293 (aarch64_sm4eqv4si): Ditto.
14294 (aarch64_sm4ekeyqv4si): Ditto.
14295 * config/aarch64/aarch64-simd.md:
14296 (aarch64_sm3ss1qv4si): Ditto.
14297 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14298 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14299 (aarch64_sm4eqv4si): Ditto.
14300 (aarch64_sm4ekeyqv4si): Ditto.
14301 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14302 (sm3part_op): Ditto.
14303 (CRYPTO_SM3TT): Ditto.
14304 (CRYPTO_SM3PART): Ditto.
14305 (UNSPEC_SM3SS1): New unspec.
14306 (UNSPEC_SM3TT1A): Ditto.
14307 (UNSPEC_SM3TT1B): Ditto.
14308 (UNSPEC_SM3TT2A): Ditto.
14309 (UNSPEC_SM3TT2B): Ditto.
14310 (UNSPEC_SM3PARTW1): Ditto.
14311 (UNSPEC_SM3PARTW2): Ditto.
14312 (UNSPEC_SM4E): Ditto.
14313 (UNSPEC_SM4EKEY): Ditto.
14314 * config/aarch64/constraints.md (Ui2): New constraint.
14315 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14316 * config/arm/types.md (crypto_sm3): New type attribute.
14317 (crypto_sm4): Ditto.
14318 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14319 (vsm3tt1aq_u32): Ditto.
14320 (vsm3tt1bq_u32): Ditto.
14321 (vsm3tt2aq_u32): Ditto.
14322 (vsm3tt2bq_u32): Ditto.
14323 (vsm3partw1q_u32): Ditto.
14324 (vsm3partw2q_u32): Ditto.
14325 (vsm4eq_u32): Ditto.
14326 (vsm4ekeyq_u32): Ditto.
14327 (doc/invoke.texi): Document new sm4 option.
14329 2018-01-10 Michael Collison <michael.collison@arm.com>
14331 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14332 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14333 (AARCH64_FL_FOR_ARCH8_4): New.
14334 (AARCH64_FL_V8_4): New flag.
14335 (doc/invoke.texi): Document new armv8.4-a option.
14337 2018-01-10 Michael Collison <michael.collison@arm.com>
14339 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14340 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14341 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14342 * config/aarch64/aarch64-option-extension.def: Add
14343 AARCH64_OPT_EXTENSION of 'sha2'.
14344 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14345 (crypto): Disable sha2 and aes if crypto disabled.
14346 (crypto): Enable aes and sha2 if enabled.
14347 (simd): Disable sha2 and aes if simd disabled.
14348 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14350 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14351 (TARGET_SHA2): New feature flag for sha2.
14352 (TARGET_AES): New feature flag for aes.
14353 * config/aarch64/aarch64-simd.md:
14354 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14355 conditional on TARGET_AES.
14356 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14357 (aarch64_crypto_sha1hsi): Make pattern conditional
14359 (aarch64_crypto_sha1hv4si): Ditto.
14360 (aarch64_be_crypto_sha1hv4si): Ditto.
14361 (aarch64_crypto_sha1su1v4si): Ditto.
14362 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14363 (aarch64_crypto_sha1su0v4si): Ditto.
14364 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14365 (aarch64_crypto_sha256su0v4si): Ditto.
14366 (aarch64_crypto_sha256su1v4si): Ditto.
14367 (doc/invoke.texi): Document new aes and sha2 options.
14369 2018-01-10 Martin Sebor <msebor@redhat.com>
14371 PR tree-optimization/83781
14372 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14375 2018-01-11 Martin Sebor <msebor@gmail.com>
14376 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14378 PR tree-optimization/83501
14379 PR tree-optimization/81703
14381 * tree-ssa-strlen.c (get_string_cst): Rename...
14382 (get_string_len): ...to this. Handle global constants.
14383 (handle_char_store): Adjust.
14385 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14386 Jim Wilson <jimw@sifive.com>
14388 * config/riscv/riscv-protos.h (riscv_output_return): New.
14389 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14390 (riscv_attribute_table, riscv_output_return),
14391 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14392 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14393 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14395 (riscv_expand_prologue): Add early return for naked function.
14396 (riscv_expand_epilogue): Likewise.
14397 (riscv_function_ok_for_sibcall): Return false for naked function.
14398 (riscv_set_current_function): New.
14399 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14400 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14401 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14402 * doc/extend.texi (RISC-V Function Attributes): New.
14404 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14406 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14407 check for 128-bit long double before checking TCmode.
14408 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14409 128-bit long doubles before checking TFmode or TCmode.
14410 (FLOAT128_IBM_P): Likewise.
14412 2018-01-10 Martin Sebor <msebor@redhat.com>
14414 PR tree-optimization/83671
14415 * builtins.c (c_strlen): Unconditionally return zero for the empty
14417 Use -Warray-bounds for warnings.
14418 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14419 for non-constant array indices with COMPONENT_REF, arrays of
14420 arrays, and pointers to arrays.
14421 (gimple_fold_builtin_strlen): Determine and set length range for
14422 non-constant character arrays.
14424 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14426 PR middle-end/81897
14427 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14430 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14432 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14434 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14437 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14438 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14439 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14440 indexed_or_indirect_operand predicate.
14441 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14442 (*vsx_le_perm_load_v8hi): Likewise.
14443 (*vsx_le_perm_load_v16qi): Likewise.
14444 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14445 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14446 (*vsx_le_perm_store_v8hi): Likewise.
14447 (*vsx_le_perm_store_v16qi): Likewise.
14448 (eight unnamed splitters): Likewise.
14450 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14452 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14453 * config/rs6000/emmintrin.h: Likewise.
14454 * config/rs6000/mmintrin.h: Likewise.
14455 * config/rs6000/xmmintrin.h: Likewise.
14457 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14460 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14462 * tree.c (tree_nop_conversion): Return true for location wrapper
14464 (maybe_wrap_with_location): New function.
14465 (selftest::check_strip_nops): New function.
14466 (selftest::test_location_wrappers): New function.
14467 (selftest::tree_c_tests): Call it.
14468 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14469 (maybe_wrap_with_location): New decl.
14470 (EXPR_LOCATION_WRAPPER_P): New macro.
14471 (location_wrapper_p): New inline function.
14472 (tree_strip_any_location_wrapper): New inline function.
14474 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14477 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14478 stack_realign_offset for the largest alignment of stack slot
14480 (ix86_find_max_used_stack_alignment): New function.
14481 (ix86_finalize_stack_frame_flags): Use it. Set
14482 max_used_stack_alignment if we don't realign stack.
14483 * config/i386/i386.h (machine_function): Add
14484 max_used_stack_alignment.
14486 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14488 * config/arm/arm.opt (-mbranch-cost): New option.
14489 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14492 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14495 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14496 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14498 2018-01-10 Richard Biener <rguenther@suse.de>
14501 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14502 early out so it also covers the case where we have a non-NULL
14505 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14507 PR tree-optimization/83753
14508 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14509 for non-strided grouped accesses if the number of elements is 1.
14511 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14514 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14515 * i386.h (TARGET_USE_GATHER): Define.
14516 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14518 2018-01-10 Martin Liska <mliska@suse.cz>
14521 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14522 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14524 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14525 CLEANUP_NO_PARTITIONING is not set.
14527 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14529 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14530 for vectors, as a partial revert of r254296.
14531 * rtl.h (const_vec_p): Delete.
14532 (const_vec_duplicate_p): Don't test for vector CONSTs.
14533 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14534 * expmed.c (make_tree): Likewise.
14537 * common.md (E, F): Use CONSTANT_P instead of checking for
14539 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14540 checking for CONST_VECTOR.
14542 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14544 PR middle-end/83575
14545 * predict.c (force_edge_cold): Handle in more sane way edges
14546 with no prediction.
14548 2018-01-09 Carl Love <cel@us.ibm.com>
14550 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14552 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14553 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14554 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14555 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14556 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14557 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14558 * config/rs6000/rs6000-protos.h: Add extern defition for
14559 rs6000_generate_float2_double_code.
14560 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14562 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14563 (float2_v2df): Add define_expand.
14565 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14568 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14569 op_mode in the force_to_mode call.
14571 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14573 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14574 instead of checking each element individually.
14575 (aarch64_evpc_uzp): Likewise.
14576 (aarch64_evpc_zip): Likewise.
14577 (aarch64_evpc_ext): Likewise.
14578 (aarch64_evpc_rev): Likewise.
14579 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14580 instead of checking each element individually. Return true without
14582 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14583 whether all selected elements come from the same input, instead of
14584 checking each element individually. Remove calls to gen_rtx_REG,
14585 start_sequence and end_sequence and instead assert that no rtl is
14588 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14590 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14591 order of HIGH and CONST checks.
14593 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14595 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14596 if the destination isn't an SSA_NAME.
14598 2018-01-09 Richard Biener <rguenther@suse.de>
14600 PR tree-optimization/83668
14601 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14603 (canonicalize_loop_form): ... here, renamed from ...
14604 (canonicalize_loop_closed_ssa_form): ... this and amended to
14605 swap successor edges for loop exit blocks to make us use
14606 the RPO order we need for initial schedule generation.
14608 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14610 PR tree-optimization/64811
14611 * match.pd: When optimizing comparisons with Inf, avoid
14612 introducing or losing exceptions from comparisons with NaN.
14614 2018-01-09 Martin Liska <mliska@suse.cz>
14617 * asan.c (shadow_mem_size): Add gcc_assert.
14619 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14621 Don't save registers in main().
14624 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14625 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14626 * config/avr/avr.c (avr_set_current_function): Don't error if
14627 naked, OS_task or OS_main are specified at the same time.
14628 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14630 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14632 * common/config/avr/avr-common.c (avr_option_optimization_table):
14633 Switch on -mmain-is-OS_task for optimizing compilations.
14635 2018-01-09 Richard Biener <rguenther@suse.de>
14637 PR tree-optimization/83572
14638 * graphite.c: Include cfganal.h.
14639 (graphite_transform_loops): Connect infinite loops to exit
14640 and remove fake edges at the end.
14642 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14644 * ipa-inline.c (edge_badness): Revert accidental checkin.
14646 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14649 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14650 symbols; not inline clones.
14652 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14655 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14656 hard registers. Formatting fixes.
14658 PR preprocessor/83722
14659 * gcc.c (try_generate_repro): Pass
14660 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14661 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14664 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14665 Kito Cheng <kito.cheng@gmail.com>
14667 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14668 (riscv_leaf_function_p): Delete.
14669 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14671 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14673 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14675 (do_ifelse): New function.
14676 (do_isel): New function.
14677 (do_sub3): New function.
14678 (do_add3): New function.
14679 (do_load_mask_compare): New function.
14680 (do_overlap_load_compare): New function.
14681 (expand_compare_loop): New function.
14682 (expand_block_compare): Call expand_compare_loop() when appropriate.
14683 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14684 option description.
14685 (-mblock-compare-inline-loop-limit): New option.
14687 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14690 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14691 Reverse order of second and third operands in first alternative.
14692 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14693 of first and second elements in UNSPEC_VPERMR vector.
14694 (altivec_expand_vec_perm_le): Likewise.
14696 2018-01-08 Jeff Law <law@redhat.com>
14698 PR rtl-optimizatin/81308
14699 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14700 (process_switch): If group_case_labels makes a change, then set
14702 (pass_convert_switch::execute): If a switch is converted, then
14703 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14705 PR rtl-optimization/81308
14706 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14709 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14711 PR target/83663 - Revert r255946
14712 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14713 generation for cases where splatting a value is not useful.
14714 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14715 across a vec_duplicate and a paradoxical subreg forming a vector
14716 mode to a vec_concat.
14718 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14720 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14721 -march=armv8.3-a variants.
14722 * config/arm/t-multilib: Likewise.
14723 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14725 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14727 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14729 (cceq_ior_compare_complement): Give it a name so I can use it, and
14730 change boolean_or_operator predicate to boolean_operator so it can
14731 be used to generate a crand.
14732 (eqne): New code iterator.
14733 (bd/bd_neg): New code_attrs.
14734 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14735 a single define_insn.
14736 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14737 decrement (bdnzt/bdnzf/bdzt/bdzf).
14738 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14739 with the new names of the branch decrement patterns, and added the
14740 names of the branch decrement conditional patterns.
14742 2018-01-08 Richard Biener <rguenther@suse.de>
14744 PR tree-optimization/83563
14745 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14748 2018-01-08 Richard Biener <rguenther@suse.de>
14750 PR middle-end/83713
14751 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14753 2018-01-08 Richard Biener <rguenther@suse.de>
14755 PR tree-optimization/83685
14756 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14757 references to abnormals.
14759 2018-01-08 Richard Biener <rguenther@suse.de>
14762 * dwarf2out.c (output_indirect_strings): Handle empty
14763 skeleton_debug_str_hash.
14764 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14766 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14768 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14769 (emit_store_direct): Likewise.
14770 (arc_trampoline_adjust_address): Likewise.
14771 (arc_asm_trampoline_template): New function.
14772 (arc_initialize_trampoline): Use asm_trampoline_template.
14773 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14774 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14775 * config/arc/arc.md (flush_icache): Delete pattern.
14777 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14779 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14780 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14783 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14786 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14787 by not USED_FOR_TARGET.
14788 (make_pass_resolve_sw_modes): Likewise.
14790 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14792 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14795 2018-01-08 Richard Biener <rguenther@suse.de>
14797 PR middle-end/83580
14798 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14800 2018-01-08 Richard Biener <rguenther@suse.de>
14802 PR middle-end/83517
14803 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14805 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14807 PR middle-end/81897
14808 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14809 basic blocks with a small number of successors.
14810 (convert_control_dep_chain_into_preds): Improve handling of
14812 (dump_predicates): Split apart into...
14813 (dump_pred_chain): ...here...
14814 (dump_pred_info): ...and here.
14815 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14816 (can_chain_union_be_invalidated_p): Improve check for invalidation
14818 (uninit_uses_cannot_happen): Avoid unnecessary if
14819 convert_control_dep_chain_into_preds yielded nothing.
14821 2018-01-06 Martin Sebor <msebor@redhat.com>
14823 PR tree-optimization/83640
14824 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14825 subtracting negative offset from size.
14826 (builtin_access::overlap): Adjust offset bounds of the access to fall
14827 within the size of the object if possible.
14829 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14831 PR rtl-optimization/83699
14832 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14833 extract_bit_field_as_subreg to cases in which the extracted
14834 value is also a vector.
14836 * lra-constraints.c (process_alt_operands): Test for the equivalence
14837 substitutions when detecting a possible reload cycle.
14839 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14842 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14843 by default if flag_selective_schedling{,2}. Formatting fixes.
14845 PR rtl-optimization/83682
14846 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14847 if it has non-VECTOR_MODE element mode.
14848 (vec_duplicate_p): Likewise.
14850 PR middle-end/83694
14851 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14852 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14854 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14857 * config/i386/i386-builtin.def
14858 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14859 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14860 Require also OPTION_MASK_ISA_AVX512F in addition to
14861 OPTION_MASK_ISA_GFNI.
14862 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14863 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14864 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14865 to OPTION_MASK_ISA_GFNI.
14866 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14867 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14868 OPTION_MASK_ISA_AVX512BW.
14869 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14870 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14871 addition to OPTION_MASK_ISA_GFNI.
14872 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14873 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14874 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14875 to OPTION_MASK_ISA_GFNI.
14876 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14877 a requirement for all ISAs rather than any of them with a few
14879 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14881 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14882 bitmasks to be enabled with 3 exceptions, instead of requiring any
14883 enabled ISA with lots of exceptions.
14884 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14885 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14886 Change avx512bw in isa attribute to avx512f.
14887 * config/i386/sgxintrin.h: Add license boilerplate.
14888 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14889 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14890 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14891 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14893 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14894 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14895 temporarily sse2 rather than sse if not enabled already.
14898 * config/i386/sse.md (VI248_VLBW): Rename to ...
14899 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14900 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14901 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14902 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14903 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14904 mode iterator instead of VI248_VLBW.
14906 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14908 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14909 (record_modified): Skip clobbers; add debug output.
14910 (param_change_prob): Use sreal frequencies.
14912 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14914 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14915 punt for user-aligned variables.
14917 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14919 * tree-chrec.c (chrec_contains_symbols): Return true for
14922 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14925 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14926 of (x|y) == x for BICS pattern.
14928 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14930 PR tree-optimization/83605
14931 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14932 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14935 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14937 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14938 * config/epiphany/rtems.h: New file.
14940 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14941 Uros Bizjak <ubizjak@gmail.com>
14944 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14945 QIreg_operand instead of register_operand predicate.
14946 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14947 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14948 comments instead of -fmitigate[-_]rop.
14950 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14953 * cgraphunit.c (symbol_table::compile): Switch to text_section
14954 before calling assembly_start debug hook.
14955 * run-rtl-passes.c (run_rtl_passes): Likewise.
14958 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14960 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14961 range_int_cst_p rather than !symbolic_range_p before calling
14962 extract_range_from_multiplicative_op_1.
14964 2018-01-04 Jeff Law <law@redhat.com>
14966 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14967 redundant test in assertion.
14969 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14971 * doc/rtl.texi: Document machine_mode wrapper classes.
14973 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14975 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14976 using tree_to_uhwi.
14978 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14980 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14981 the VEC_PERM_EXPR fold to fail.
14983 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14986 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14987 to switched_sections.
14989 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14992 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
14993 test for d.testing.
14995 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
14998 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
14999 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15001 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15004 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15005 is BLKmode and bitpos not zero or mode change is needed.
15007 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15010 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15013 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15016 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15017 instead of MULT rtx. Update all corresponding splitters.
15018 (*saddl_se): Ditto.
15019 (*ssub<modesuffix>): Ditto.
15020 (*ssubl_se): Ditto.
15021 (*cmp_sadd_di): Update split patterns.
15022 (*cmp_sadd_si): Ditto.
15023 (*cmp_sadd_sidi): Ditto.
15024 (*cmp_ssub_di): Ditto.
15025 (*cmp_ssub_si): Ditto.
15026 (*cmp_ssub_sidi): Ditto.
15027 * config/alpha/predicates.md (const23_operand): New predicate.
15028 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15029 Look for ASHIFT, not MULT inner operand.
15030 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15032 2018-01-04 Martin Liska <mliska@suse.cz>
15034 PR gcov-profile/83669
15035 * gcov.c (output_intermediate_file): Add version to intermediate
15037 * doc/gcov.texi: Document new field 'version' in intermediate
15038 file format. Fix location of '-k' option of gcov command.
15040 2018-01-04 Martin Liska <mliska@suse.cz>
15043 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15045 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15047 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15049 2018-01-03 Martin Sebor <msebor@redhat.com>
15051 PR tree-optimization/83655
15052 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15053 checking calls with invalid arguments.
15055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15057 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15058 (vectorizable_mask_load_store): Delete.
15059 (vectorizable_call): Return false for masked loads and stores.
15060 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15061 instead of gimple_assign_rhs1.
15062 (vectorizable_load): Handle IFN_MASK_LOAD.
15063 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15067 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15069 (vectorizable_mask_load_store): ...here.
15070 (vectorizable_load): ...and here.
15072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15074 * tree-vect-stmts.c (vect_build_all_ones_mask)
15075 (vect_build_zero_merge_argument): New functions, split out from...
15076 (vectorizable_load): ...here.
15078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15080 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15082 (vectorizable_mask_load_store): ...here.
15083 (vectorizable_store): ...and here.
15085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15087 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15089 (vectorizable_mask_load_store): ...here.
15091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15093 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15094 (vect_model_store_cost): Take a vec_load_store_type instead of a
15096 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15097 (vect_model_store_cost): Take a vec_load_store_type instead of a
15099 (vectorizable_mask_load_store): Update accordingly.
15100 (vectorizable_store): Likewise.
15101 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15105 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15106 IFN_MASK_LOAD calls here rather than...
15107 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15110 Alan Hayward <alan.hayward@arm.com>
15111 David Sherwood <david.sherwood@arm.com>
15113 * expmed.c (extract_bit_field_1): For vector extracts,
15114 fall back to extract_bit_field_as_subreg if vec_extract
15117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15118 Alan Hayward <alan.hayward@arm.com>
15119 David Sherwood <david.sherwood@arm.com>
15121 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15122 they are variable or constant sized.
15123 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15124 slots for constant-sized data.
15126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15127 Alan Hayward <alan.hayward@arm.com>
15128 David Sherwood <david.sherwood@arm.com>
15130 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15131 handling COND_EXPRs with boolean comparisons, try to find a better
15132 basis for the mask type than the boolean itself.
15134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15136 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15137 is calculated and how it can be overridden.
15138 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15139 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15141 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15145 Alan Hayward <alan.hayward@arm.com>
15146 David Sherwood <david.sherwood@arm.com>
15148 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15149 Remove the mode argument.
15150 (aarch64_simd_valid_immediate): Remove the mode and inverse
15152 * config/aarch64/iterators.md (bitsize): New iterator.
15153 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15154 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15155 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15156 aarch64_simd_valid_immediate.
15157 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15158 (aarch64_reg_or_bic_imm): Likewise.
15159 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15160 with an insn_type enum and msl with a modifier_type enum.
15161 Replace element_width with a scalar_mode. Change the shift
15162 to unsigned int. Add constructors for scalar_float_mode and
15163 scalar_int_mode elements.
15164 (aarch64_vect_float_const_representable_p): Delete.
15165 (aarch64_can_const_movi_rtx_p)
15166 (aarch64_simd_scalar_immediate_valid_for_move)
15167 (aarch64_simd_make_constant): Update call to
15168 aarch64_simd_valid_immediate.
15169 (aarch64_advsimd_valid_immediate_hs): New function.
15170 (aarch64_advsimd_valid_immediate): Likewise.
15171 (aarch64_simd_valid_immediate): Remove mode and inverse
15172 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15173 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15174 and aarch64_float_const_representable_p on the result.
15175 (aarch64_output_simd_mov_immediate): Remove mode argument.
15176 Update call to aarch64_simd_valid_immediate and use of
15177 simd_immediate_info.
15178 (aarch64_output_scalar_simd_mov_immediate): Update call
15181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15182 Alan Hayward <alan.hayward@arm.com>
15183 David Sherwood <david.sherwood@arm.com>
15185 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15186 (mode_nunits): Likewise CONST_MODE_NUNITS.
15187 * machmode.def (ADJUST_NUNITS): Document.
15188 * genmodes.c (mode_data::need_nunits_adj): New field.
15189 (blank_mode): Update accordingly.
15190 (adj_nunits): New variable.
15191 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15193 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15194 listed in adj_nunits.
15195 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15196 listed in adj_nunits. Don't emit case statements for such modes.
15197 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15198 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15199 nothing if adj_nunits is nonnull.
15200 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15201 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15202 (emit_mode_fbit): Update use of print_maybe_const_decl.
15203 (emit_move_size): Likewise. Treat the array as non-const
15205 (emit_mode_adjustments): Handle adj_nunits.
15207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15209 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15210 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15211 (VECTOR_MODES): Use it.
15212 (make_vector_modes): Take the prefix as an argument.
15214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15215 Alan Hayward <alan.hayward@arm.com>
15216 David Sherwood <david.sherwood@arm.com>
15218 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15219 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15220 for MODE_VECTOR_BOOL.
15221 * machmode.def (VECTOR_BOOL_MODE): Document.
15222 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15223 (make_vector_bool_mode): New function.
15224 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15226 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15227 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15229 * stor-layout.c (int_mode_for_mode): Likewise.
15230 * tree.c (build_vector_type_for_mode): Likewise.
15231 * varasm.c (output_constant_pool_2): Likewise.
15232 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15233 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15234 for MODE_VECTOR_BOOL.
15235 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15236 of mode class checks.
15237 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15238 instead of a list of mode class checks.
15239 (expand_vector_scalar_condition): Likewise.
15240 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15243 Alan Hayward <alan.hayward@arm.com>
15244 David Sherwood <david.sherwood@arm.com>
15246 * machmode.h (mode_size): Change from unsigned short to
15248 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15249 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15250 or if measurement_type is not polynomial.
15251 (fixed_size_mode::includes_p): Check for constant-sized modes.
15252 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15253 return a poly_uint16 rather than an unsigned short.
15254 (emit_mode_size): Change the type of mode_size from unsigned short
15255 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15256 (emit_mode_adjustments): Cope with polynomial vector sizes.
15257 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15259 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15261 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15262 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15263 * caller-save.c (setup_save_areas): Likewise.
15264 (replace_reg_with_saved_mem): Likewise.
15265 * calls.c (emit_library_call_value_1): Likewise.
15266 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15267 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15268 (gen_lowpart_for_combine): Likewise.
15269 * convert.c (convert_to_integer_1): Likewise.
15270 * cse.c (equiv_constant, cse_insn): Likewise.
15271 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15272 (cselib_subst_to_values): Likewise.
15273 * dce.c (word_dce_process_block): Likewise.
15274 * df-problems.c (df_word_lr_mark_ref): Likewise.
15275 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15276 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15277 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15278 (rtl_for_decl_location): Likewise.
15279 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15280 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15281 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15282 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15283 (expand_expr_real_1): Likewise.
15284 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15285 (pad_below): Likewise.
15286 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15287 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15288 * ira.c (get_subreg_tracking_sizes): Likewise.
15289 * ira-build.c (ira_create_allocno_objects): Likewise.
15290 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15291 (ira_sort_regnos_for_alter_reg): Likewise.
15292 * ira-costs.c (record_operand_costs): Likewise.
15293 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15294 (resolve_simple_move): Likewise.
15295 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15296 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15297 (lra_constraints): Likewise.
15298 (CONST_POOL_OK_P): Reject variable-sized modes.
15299 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15300 (add_pseudo_to_slot, lra_spill): Likewise.
15301 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15302 * optabs-query.c (get_best_extraction_insn): Likewise.
15303 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15304 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15305 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15306 * recog.c (offsettable_address_addr_space_p): Likewise.
15307 * regcprop.c (maybe_mode_change): Likewise.
15308 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15309 * regrename.c (build_def_use): Likewise.
15310 * regstat.c (dump_reg_info): Likewise.
15311 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15312 (find_reloads, find_reloads_subreg_address): Likewise.
15313 * reload1.c (eliminate_regs_1): Likewise.
15314 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15315 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15316 (simplify_binary_operation_1, simplify_subreg): Likewise.
15317 * targhooks.c (default_function_arg_padding): Likewise.
15318 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15319 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15320 (verify_gimple_assign_ternary): Likewise.
15321 * tree-inline.c (estimate_move_cost): Likewise.
15322 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15323 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15324 (get_address_cost_ainc): Likewise.
15325 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15326 (vect_supportable_dr_alignment): Likewise.
15327 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15328 (vectorizable_reduction): Likewise.
15329 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15330 (vectorizable_operation, vectorizable_load): Likewise.
15331 * tree.c (build_same_sized_truth_vector_type): Likewise.
15332 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15333 * var-tracking.c (emit_note_insn_var_location): Likewise.
15334 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15335 (ADDR_VEC_ALIGN): Likewise.
15337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15338 Alan Hayward <alan.hayward@arm.com>
15339 David Sherwood <david.sherwood@arm.com>
15341 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15343 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15344 or if measurement_type is polynomial.
15345 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15346 * combine.c (make_extraction): Likewise.
15347 * dse.c (find_shift_sequence): Likewise.
15348 * dwarf2out.c (mem_loc_descriptor): Likewise.
15349 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15350 (extract_bit_field, extract_low_bits): Likewise.
15351 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15352 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15353 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15354 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15355 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15356 * reload.c (find_reloads): Likewise.
15357 * reload1.c (alter_reg): Likewise.
15358 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15359 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15360 * tree-if-conv.c (predicate_mem_writes): Likewise.
15361 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15362 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15363 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15364 * valtrack.c (dead_debug_insert_temp): Likewise.
15365 * varasm.c (mergeable_constant_section): Likewise.
15366 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15369 Alan Hayward <alan.hayward@arm.com>
15370 David Sherwood <david.sherwood@arm.com>
15372 * expr.c (expand_assignment): Cope with polynomial mode sizes
15373 when assigning to a CONCAT.
15375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15376 Alan Hayward <alan.hayward@arm.com>
15377 David Sherwood <david.sherwood@arm.com>
15379 * machmode.h (mode_precision): Change from unsigned short to
15381 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15383 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15384 or if measurement_type is not polynomial.
15385 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15386 in which the mode is already known to be a scalar_int_mode.
15387 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15388 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15390 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15391 for GET_MODE_PRECISION.
15392 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15393 for GET_MODE_PRECISION.
15394 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15396 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15397 (expand_field_assignment, make_extraction): Likewise.
15398 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15399 (get_last_value): Likewise.
15400 * convert.c (convert_to_integer_1): Likewise.
15401 * cse.c (cse_insn): Likewise.
15402 * expr.c (expand_expr_real_1): Likewise.
15403 * lra-constraints.c (simplify_operand_subreg): Likewise.
15404 * optabs-query.c (can_atomic_load_p): Likewise.
15405 * optabs.c (expand_atomic_load): Likewise.
15406 (expand_atomic_store): Likewise.
15407 * ree.c (combine_reaching_defs): Likewise.
15408 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15409 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15410 * tree.h (type_has_mode_precision_p): Likewise.
15411 * ubsan.c (instrument_si_overflow): Likewise.
15413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15414 Alan Hayward <alan.hayward@arm.com>
15415 David Sherwood <david.sherwood@arm.com>
15417 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15418 polynomial numbers of units.
15419 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15420 (valid_vector_subparts_p): New function.
15421 (build_vector_type): Remove temporary shim and take the number
15422 of units as a poly_uint64 rather than an int.
15423 (build_opaque_vector_type): Take the number of units as a
15424 poly_uint64 rather than an int.
15425 * tree.c (build_vector_from_ctor): Handle polynomial
15426 TYPE_VECTOR_SUBPARTS.
15427 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15428 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15429 (build_vector_from_val): If the number of units is variable,
15430 use build_vec_duplicate_cst for constant operands and
15431 VEC_DUPLICATE_EXPR otherwise.
15432 (make_vector_type): Remove temporary is_constant ().
15433 (build_vector_type, build_opaque_vector_type): Take the number of
15434 units as a poly_uint64 rather than an int.
15435 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15437 * cfgexpand.c (expand_debug_expr): Likewise.
15438 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15439 (store_constructor, expand_expr_real_1): Likewise.
15440 (const_scalar_mask_from_tree): Likewise.
15441 * fold-const-call.c (fold_const_reduction): Likewise.
15442 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15443 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15444 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15445 (fold_relational_const): Likewise.
15446 (native_interpret_vector): Likewise. Change the size from an
15447 int to an unsigned int.
15448 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15449 TYPE_VECTOR_SUBPARTS.
15450 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15451 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15452 duplicating a non-constant operand into a variable-length vector.
15453 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15454 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15455 * ipa-icf.c (sem_variable::equals): Likewise.
15456 * match.pd: Likewise.
15457 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15458 * print-tree.c (print_node): Likewise.
15459 * stor-layout.c (layout_type): Likewise.
15460 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15461 * tree-cfg.c (verify_gimple_comparison): Likewise.
15462 (verify_gimple_assign_binary): Likewise.
15463 (verify_gimple_assign_ternary): Likewise.
15464 (verify_gimple_assign_single): Likewise.
15465 * tree-pretty-print.c (dump_generic_node): Likewise.
15466 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15467 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15468 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15469 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15470 (vect_shift_permute_load_chain): Likewise.
15471 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15472 (expand_vector_condition, optimize_vector_constructor): Likewise.
15473 (lower_vec_perm, get_compute_type): Likewise.
15474 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15475 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15476 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15477 (vect_recog_mask_conversion_pattern): Likewise.
15478 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15479 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15480 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15481 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15482 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15483 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15484 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15485 (supportable_widening_operation): Likewise.
15486 (supportable_narrowing_operation): Likewise.
15487 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15489 * varasm.c (output_constant): Likewise.
15491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15492 Alan Hayward <alan.hayward@arm.com>
15493 David Sherwood <david.sherwood@arm.com>
15495 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15496 so that both the length == 3 and length != 3 cases set up their
15497 own permute vectors. Add comments explaining why we know the
15498 number of elements is constant.
15499 (vect_permute_load_chain): Likewise.
15501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15502 Alan Hayward <alan.hayward@arm.com>
15503 David Sherwood <david.sherwood@arm.com>
15505 * machmode.h (mode_nunits): Change from unsigned char to
15507 (ONLY_FIXED_SIZE_MODES): New macro.
15508 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15509 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15510 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15512 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15513 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15514 or if measurement_type is not polynomial.
15515 * genmodes.c (ZERO_COEFFS): New macro.
15516 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15518 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15519 Use ZERO_COEFFS when emitting initializers.
15520 * data-streamer.h (bp_pack_poly_value): New function.
15521 (bp_unpack_poly_value): Likewise.
15522 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15523 for GET_MODE_NUNITS.
15524 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15525 for GET_MODE_NUNITS.
15526 * tree.c (make_vector_type): Remove temporary shim and make
15527 the real function take the number of units as a poly_uint64
15528 rather than an int.
15529 (build_vector_type_for_mode): Handle polynomial nunits.
15530 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15531 * emit-rtl.c (const_vec_series_p_1): Likewise.
15532 (gen_rtx_CONST_VECTOR): Likewise.
15533 * fold-const.c (test_vec_duplicate_folding): Likewise.
15534 * genrecog.c (validate_pattern): Likewise.
15535 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15536 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15537 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15538 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15539 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15540 * rtlanal.c (subreg_get_info): Likewise.
15541 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15542 (vect_grouped_load_supported): Likewise.
15543 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15544 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15545 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15546 (simplify_const_unary_operation, simplify_binary_operation_1)
15547 (simplify_const_binary_operation, simplify_ternary_operation)
15548 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15549 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15550 instead of CONST_VECTOR_NUNITS.
15551 * varasm.c (output_constant_pool_2): Likewise.
15552 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15553 explicit-encoded elements in the XVEC for variable-length vectors.
15555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15557 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15560 Alan Hayward <alan.hayward@arm.com>
15561 David Sherwood <david.sherwood@arm.com>
15563 * coretypes.h (fixed_size_mode): Declare.
15564 (fixed_size_mode_pod): New typedef.
15565 * builtins.h (target_builtins::x_apply_args_mode)
15566 (target_builtins::x_apply_result_mode): Change type to
15567 fixed_size_mode_pod.
15568 * builtins.c (apply_args_size, apply_result_size, result_vector)
15569 (expand_builtin_apply_args_1, expand_builtin_apply)
15570 (expand_builtin_return): Update accordingly.
15572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15574 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15575 * cselib.c (cselib_hash_rtx): Likewise.
15576 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15577 CONST_VECTOR encoding.
15579 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15580 Jeff Law <law@redhat.com>
15583 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15584 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15585 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15586 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15589 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15590 explicitly probe *sp in a noreturn function if there were any callee
15591 register saves or frame pointer is needed.
15593 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15596 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15597 BLKmode for ternary, binary or unary expressions.
15600 * var-tracking.c (delete_vta_debug_insn): New inline function.
15601 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15602 insns from get_insns () to NULL instead of each bb separately.
15603 Use delete_vta_debug_insn. No longer static.
15604 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15605 delete_vta_debug_insns callers.
15606 * rtl.h (delete_vta_debug_insns): Declare.
15607 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15608 instead of variable_tracking_main.
15610 2018-01-03 Martin Sebor <msebor@redhat.com>
15612 PR tree-optimization/83603
15613 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15614 arguments past the endof the argument list in functions declared
15615 without a prototype.
15616 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15617 Avoid checking when arguments are null.
15619 2018-01-03 Martin Sebor <msebor@redhat.com>
15622 * doc/extend.texi (attribute const): Fix a typo.
15623 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15624 issuing -Wsuggest-attribute for void functions.
15626 2018-01-03 Martin Sebor <msebor@redhat.com>
15628 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15629 offset_int::from instead of wide_int::to_shwi.
15630 (maybe_diag_overlap): Remove assertion.
15631 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15632 * gimple-ssa-sprintf.c (format_directive): Same.
15633 (parse_directive): Same.
15634 (sprintf_dom_walker::compute_format_length): Same.
15635 (try_substitute_return_value): Same.
15637 2018-01-03 Jeff Law <law@redhat.com>
15639 PR middle-end/83654
15640 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15641 non-constant residual for zero at runtime and avoid probing in
15642 that case. Reorganize code for trailing problem to mirror handling
15645 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15647 PR tree-optimization/83501
15648 * tree-ssa-strlen.c (get_string_cst): New.
15649 (handle_char_store): Call get_string_cst.
15651 2018-01-03 Martin Liska <mliska@suse.cz>
15653 PR tree-optimization/83593
15654 * tree-ssa-strlen.c: Include tree-cfg.h.
15655 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15656 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15657 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15659 (strlen_dom_walker::before_dom_children): Call
15660 gimple_purge_dead_eh_edges. Dump tranformation with details
15662 (strlen_dom_walker::before_dom_children): Update call by adding
15663 new argument cleanup_eh.
15664 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15666 2018-01-03 Martin Liska <mliska@suse.cz>
15669 * cif-code.def (VARIADIC_THUNK): New enum value.
15670 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15673 2018-01-03 Jan Beulich <jbeulich@suse.com>
15675 * sse.md (mov<mode>_internal): Tighten condition for when to use
15676 vmovdqu<ssescalarsize> for TI and OI modes.
15678 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15680 Update copyright years.
15682 2018-01-03 Martin Liska <mliska@suse.cz>
15685 * ipa-visibility.c (function_and_variable_visibility): Skip
15686 functions with noipa attribure.
15688 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15690 * gcc.c (process_command): Update copyright notice dates.
15691 * gcov-dump.c (print_version): Ditto.
15692 * gcov.c (print_version): Ditto.
15693 * gcov-tool.c (print_version): Ditto.
15694 * gengtype.c (create_file): Ditto.
15695 * doc/cpp.texi: Bump @copying's copyright year.
15696 * doc/cppinternals.texi: Ditto.
15697 * doc/gcc.texi: Ditto.
15698 * doc/gccint.texi: Ditto.
15699 * doc/gcov.texi: Ditto.
15700 * doc/install.texi: Ditto.
15701 * doc/invoke.texi: Ditto.
15703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15705 * vector-builder.h (vector_builder::m_full_nelts): Change from
15706 unsigned int to poly_uint64.
15707 (vector_builder::full_nelts): Update prototype accordingly.
15708 (vector_builder::new_vector): Likewise.
15709 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15710 (vector_builder::operator ==): Likewise.
15711 (vector_builder::finalize): Likewise.
15712 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15713 Take the number of elements as a poly_uint64 rather than an
15715 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15716 from unsigned int to poly_uint64.
15717 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15718 (vec_perm_indices::new_vector): Likewise.
15719 (vec_perm_indices::length): Likewise.
15720 (vec_perm_indices::nelts_per_input): Likewise.
15721 (vec_perm_indices::input_nelts): Likewise.
15722 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15723 number of elements per input as a poly_uint64 rather than an
15724 unsigned int. Use the original encoding for variable-length
15725 vectors, rather than clamping each individual element.
15726 For the second and subsequent elements in each pattern,
15727 clamp the step and base before clamping their sum.
15728 (vec_perm_indices::series_p): Handle polynomial element counts.
15729 (vec_perm_indices::all_in_range_p): Likewise.
15730 (vec_perm_indices_to_tree): Likewise.
15731 (vec_perm_indices_to_rtx): Likewise.
15732 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15733 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15734 (tree_vector_builder::new_binary_operation): Handle polynomial
15735 element counts. Return false if we need to know the number
15736 of elements at compile time.
15737 * fold-const.c (fold_vec_perm): Punt if the number of elements
15738 isn't known at compile time.
15740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15742 * vec-perm-indices.h (vec_perm_builder): Change element type
15743 from HOST_WIDE_INT to poly_int64.
15744 (vec_perm_indices::element_type): Update accordingly.
15745 (vec_perm_indices::clamp): Handle polynomial element_types.
15746 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15747 (vec_perm_indices::all_in_range_p): Likewise.
15748 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15750 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15751 polynomial vec_perm_indices element types.
15752 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15753 * fold-const.c (fold_vec_perm): Likewise.
15754 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15755 * tree-vect-generic.c (lower_vec_perm): Likewise.
15756 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15757 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15758 element type to HOST_WIDE_INT.
15760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15761 Alan Hayward <alan.hayward@arm.com>
15762 David Sherwood <david.sherwood@arm.com>
15764 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15765 rather than an int. Use plus_constant.
15766 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15767 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15770 Alan Hayward <alan.hayward@arm.com>
15771 David Sherwood <david.sherwood@arm.com>
15773 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15774 a HOST_WIDE_INT to a poly_int64.
15776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15777 Alan Hayward <alan.hayward@arm.com>
15778 David Sherwood <david.sherwood@arm.com>
15780 * calls.c (load_register_parameters): Cope with polynomial
15781 mode sizes. Require a constant size for BLKmode parameters
15782 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15783 forces a parameter to be padded at the lsb end in order to
15784 fill a complete number of words, require the parameter size
15785 to be ordered wrt UNITS_PER_WORD.
15787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15788 Alan Hayward <alan.hayward@arm.com>
15789 David Sherwood <david.sherwood@arm.com>
15791 * reload1.c (spill_stack_slot_width): Change element type
15792 from unsigned int to poly_uint64_pod.
15793 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
15800 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
15807 operands whose sizes aren't ordered.
15808 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15815 that the mode size is in the set {1, 2, 4, 8, 16}.
15817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15818 Alan Hayward <alan.hayward@arm.com>
15819 David Sherwood <david.sherwood@arm.com>
15821 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15822 Use plus_constant instead of gen_rtx_PLUS.
15824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15825 Alan Hayward <alan.hayward@arm.com>
15826 David Sherwood <david.sherwood@arm.com>
15828 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15829 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15830 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15831 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15832 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15833 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15834 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15835 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15836 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15837 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15839 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15840 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15841 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15842 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15843 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15844 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15845 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15846 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15847 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15848 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15850 * expr.c (emit_move_resolve_push): Treat the input and result
15851 of PUSH_ROUNDING as a poly_int64.
15852 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15853 (emit_push_insn): Likewise.
15854 * lra-eliminations.c (mark_not_eliminable): Likewise.
15855 * recog.c (push_operand): Likewise.
15856 * reload1.c (elimination_effects): Likewise.
15857 * rtlanal.c (nonzero_bits1): Likewise.
15858 * calls.c (store_one_arg): Likewise. Require the padding to be
15859 known at compile time.
15861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15862 Alan Hayward <alan.hayward@arm.com>
15863 David Sherwood <david.sherwood@arm.com>
15865 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15866 Use plus_constant instead of gen_rtx_PLUS.
15868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15869 Alan Hayward <alan.hayward@arm.com>
15870 David Sherwood <david.sherwood@arm.com>
15872 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15873 rather than an int.
15875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15876 Alan Hayward <alan.hayward@arm.com>
15877 David Sherwood <david.sherwood@arm.com>
15879 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15880 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15881 via stack temporaries. Treat the mode size as polynomial too.
15883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15884 Alan Hayward <alan.hayward@arm.com>
15885 David Sherwood <david.sherwood@arm.com>
15887 * expr.c (expand_expr_real_2): When handling conversions involving
15888 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15889 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15890 as a poly_uint64 too.
15892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15893 Alan Hayward <alan.hayward@arm.com>
15894 David Sherwood <david.sherwood@arm.com>
15896 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
15903 REGMODE_NATURAL_SIZE.
15904 * expmed.c (store_bit_field_1): Likewise.
15905 * expr.c (store_constructor): Likewise.
15906 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15907 and polynomial REGMODE_NATURAL_SIZE.
15908 (gen_lowpart_common): Likewise.
15909 * reginfo.c (record_subregs_of_mode): Likewise.
15910 * rtlanal.c (read_modify_subreg_p): Likewise.
15912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15913 Alan Hayward <alan.hayward@arm.com>
15914 David Sherwood <david.sherwood@arm.com>
15916 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15917 numbers of elements.
15919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15920 Alan Hayward <alan.hayward@arm.com>
15921 David Sherwood <david.sherwood@arm.com>
15923 * match.pd: Cope with polynomial numbers of vector elements.
15925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15926 Alan Hayward <alan.hayward@arm.com>
15927 David Sherwood <david.sherwood@arm.com>
15929 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15930 in a POINTER_PLUS_EXPR.
15932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15933 Alan Hayward <alan.hayward@arm.com>
15934 David Sherwood <david.sherwood@arm.com>
15936 * omp-simd-clone.c (simd_clone_subparts): New function.
15937 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15938 (ipa_simd_modify_function_body): Likewise.
15940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15941 Alan Hayward <alan.hayward@arm.com>
15942 David Sherwood <david.sherwood@arm.com>
15944 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15945 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15946 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15947 (expand_vector_condition, vector_element): Likewise.
15948 (subparts_gt): New function.
15949 (get_compute_type): Use subparts_gt.
15950 (count_type_subparts): Delete.
15951 (expand_vector_operations_1): Use subparts_gt instead of
15952 count_type_subparts.
15954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15955 Alan Hayward <alan.hayward@arm.com>
15956 David Sherwood <david.sherwood@arm.com>
15958 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15959 (vect_compile_time_alias): ...this new function. Do the calculation
15960 on poly_ints rather than trees.
15961 (vect_prune_runtime_alias_test_list): Update call accordingly.
15963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15964 Alan Hayward <alan.hayward@arm.com>
15965 David Sherwood <david.sherwood@arm.com>
15967 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
15969 (vect_schedule_slp_instance): Likewise.
15971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15972 Alan Hayward <alan.hayward@arm.com>
15973 David Sherwood <david.sherwood@arm.com>
15975 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15976 constant and extern definitions for variable-length vectors.
15977 (vect_get_constant_vectors): Note that the number of units
15978 is known to be constant.
15980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15981 Alan Hayward <alan.hayward@arm.com>
15982 David Sherwood <david.sherwood@arm.com>
15984 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
15985 of units as polynomial. Choose between WIDE and NARROW based
15988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15989 Alan Hayward <alan.hayward@arm.com>
15990 David Sherwood <david.sherwood@arm.com>
15992 * tree-vect-stmts.c (simd_clone_subparts): New function.
15993 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
15995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15996 Alan Hayward <alan.hayward@arm.com>
15997 David Sherwood <david.sherwood@arm.com>
15999 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16000 vectors as polynomial. Use build_index_vector for
16001 IFN_GOMP_SIMD_LANE.
16003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16004 Alan Hayward <alan.hayward@arm.com>
16005 David Sherwood <david.sherwood@arm.com>
16007 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16008 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16009 for variable-length vectors.
16010 (vectorizable_mask_load_store): Treat the number of units as
16011 polynomial, asserting that it is constant if the condition has
16012 already been enforced.
16013 (vectorizable_store, vectorizable_load): Likewise.
16015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16016 Alan Hayward <alan.hayward@arm.com>
16017 David Sherwood <david.sherwood@arm.com>
16019 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16020 of units as polynomial. Punt if we can't tell at compile time
16021 which vector contains the final result.
16023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16024 Alan Hayward <alan.hayward@arm.com>
16025 David Sherwood <david.sherwood@arm.com>
16027 * tree-vect-loop.c (vectorizable_induction): Treat the number
16028 of units as polynomial. Punt on SLP inductions. Use an integer
16029 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16030 cast of such a series for variable-length floating-point
16033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16034 Alan Hayward <alan.hayward@arm.com>
16035 David Sherwood <david.sherwood@arm.com>
16037 * tree.h (build_index_vector): Declare.
16038 * tree.c (build_index_vector): New function.
16039 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16040 of units as polynomial, forcibly converting it to a constant if
16041 vectorizable_reduction has already enforced the condition.
16042 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16043 to create a {1,2,3,...} vector.
16044 (vectorizable_reduction): Treat the number of units as polynomial.
16045 Choose vectype_in based on the largest scalar element size rather
16046 than the smallest number of units. Enforce the restrictions
16049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16050 Alan Hayward <alan.hayward@arm.com>
16051 David Sherwood <david.sherwood@arm.com>
16053 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16054 number of units as polynomial.
16056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16057 Alan Hayward <alan.hayward@arm.com>
16058 David Sherwood <david.sherwood@arm.com>
16060 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16061 * target.def (autovectorize_vector_sizes): Return the vector sizes
16062 by pointer, using vector_sizes rather than a bitmask.
16063 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16064 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16065 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16067 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16068 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16069 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16070 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16071 * omp-general.c (omp_max_vf): Likewise.
16072 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16073 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16074 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16075 * tree-vect-slp.c (vect_slp_bb): Likewise.
16076 * doc/tm.texi: Regenerate.
16077 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16079 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16080 the vector size as a poly_uint64 rather than an unsigned int.
16081 (current_vector_size): Change from an unsigned int to a poly_uint64.
16082 (get_vectype_for_scalar_type): Update accordingly.
16083 * tree.h (build_truth_vector_type): Take the size and number of
16084 units as a poly_uint64 rather than an unsigned int.
16085 (build_vector_type): Add a temporary overload that takes
16086 the number of units as a poly_uint64 rather than an unsigned int.
16087 * tree.c (make_vector_type): Likewise.
16088 (build_truth_vector_type): Take the number of units as a poly_uint64
16089 rather than an unsigned int.
16091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16092 Alan Hayward <alan.hayward@arm.com>
16093 David Sherwood <david.sherwood@arm.com>
16095 * target.def (get_mask_mode): Take the number of units and length
16096 as poly_uint64s rather than unsigned ints.
16097 * targhooks.h (default_get_mask_mode): Update accordingly.
16098 * targhooks.c (default_get_mask_mode): Likewise.
16099 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16100 * doc/tm.texi: Regenerate.
16102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16103 Alan Hayward <alan.hayward@arm.com>
16104 David Sherwood <david.sherwood@arm.com>
16106 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16107 * omp-general.c (omp_max_vf): Likewise.
16108 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16109 (expand_omp_simd): Handle polynomial safelen.
16110 * omp-low.c (omplow_simd_context): Add a default constructor.
16111 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16112 (lower_rec_simd_input_clauses): Update accordingly.
16113 (lower_rec_input_clauses): Likewise.
16115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16116 Alan Hayward <alan.hayward@arm.com>
16117 David Sherwood <david.sherwood@arm.com>
16119 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16120 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16121 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16122 (vect_analyze_slp_cost): Likewise.
16123 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16124 (vect_model_load_cost): Likewise.
16126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16127 Alan Hayward <alan.hayward@arm.com>
16128 David Sherwood <david.sherwood@arm.com>
16130 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16131 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16132 from an unsigned int * to a poly_uint64_pod *.
16133 (calculate_unrolling_factor): New function.
16134 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16137 Alan Hayward <alan.hayward@arm.com>
16138 David Sherwood <david.sherwood@arm.com>
16140 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16141 from an unsigned int to a poly_uint64.
16142 (_loop_vec_info::slp_unrolling_factor): Likewise.
16143 (_loop_vec_info::vectorization_factor): Change from an int
16145 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16146 (vect_get_num_vectors): New function.
16147 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16148 (vect_get_num_copies): Use vect_get_num_vectors.
16149 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16150 to an unsigned int *.
16151 (vect_analyze_data_refs): Change min_vf from an int * to a
16153 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16154 than an unsigned HOST_WIDE_INT.
16155 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16156 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16157 to an unsigned int *.
16158 (vect_analyze_data_ref_dependences): Likewise.
16159 (vect_compute_data_ref_alignment): Handle polynomial vf.
16160 (vect_enhance_data_refs_alignment): Likewise.
16161 (vect_prune_runtime_alias_test_list): Likewise.
16162 (vect_shift_permute_load_chain): Likewise.
16163 (vect_supportable_dr_alignment): Likewise.
16164 (dependence_distance_ge_vf): Take the vectorization factor as a
16165 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16166 (vect_analyze_data_refs): Change min_vf from an int * to a
16168 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16169 vfm1 as a poly_uint64 rather than an int. Make the same change
16170 for the returned bound_scalar.
16171 (vect_gen_vector_loop_niters): Handle polynomial vf.
16172 (vect_do_peeling): Likewise. Update call to
16173 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16174 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16176 * tree-vect-loop.c (vect_determine_vectorization_factor)
16177 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16178 (vect_get_known_peeling_cost): Likewise.
16179 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16180 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16181 (vect_transform_loop): Likewise. Use the lowest possible VF when
16182 updating the upper bounds of the loop.
16183 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16184 rather than an int.
16185 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16186 polynomial unroll factors.
16187 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16188 (vect_make_slp_decision): Likewise.
16189 (vect_supported_load_permutation_p): Likewise, and polynomial
16191 (vect_analyze_slp_cost): Handle polynomial vf.
16192 (vect_slp_analyze_node_operations): Likewise.
16193 (vect_slp_analyze_bb_1): Likewise.
16194 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16195 than an unsigned HOST_WIDE_INT.
16196 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16197 (vectorizable_load): Handle polynomial vf.
16198 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16200 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16203 Alan Hayward <alan.hayward@arm.com>
16204 David Sherwood <david.sherwood@arm.com>
16206 * match.pd: Handle bit operations involving three constants
16207 and try to fold one pair.
16209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16211 * tree-vect-loop-manip.c: Include gimple-fold.h.
16212 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16213 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16214 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16215 Add a path that uses a step of VF instead of 1, but disable it
16217 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16218 and niters_no_overflow parameters. Update calls to
16219 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16220 Create a new SSA name if the latter choses to use a ste other
16221 than zero, and return it via niters_vector_mult_vf_var.
16222 * tree-vect-loop.c (vect_transform_loop): Update calls to
16223 vect_do_peeling, vect_gen_vector_loop_niters and
16224 slpeel_make_loop_iterate_ntimes.
16225 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16226 (vect_gen_vector_loop_niters): Update declarations after above changes.
16228 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16230 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16231 128-bit round to integer instructions.
16232 (ceil<mode>2): Likewise.
16233 (btrunc<mode>2): Likewise.
16234 (round<mode>2): Likewise.
16236 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16238 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16239 unaligned VSX load/store on P8/P9.
16240 (expand_block_clear): Allow the use of unaligned VSX
16241 load/store on P8/P9.
16243 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16245 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16247 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16248 swap associated with both a load and a store.
16250 2018-01-02 Andrew Waterman <andrew@sifive.com>
16252 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16253 * config/riscv/riscv.md (clear_cache): Use it.
16255 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16257 * web.c: Remove out-of-date comment.
16259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16261 * expr.c (fixup_args_size_notes): Check that any existing
16262 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16263 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16264 (emit_single_push_insn): ...here.
16266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16268 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16269 (const_vector_encoded_nelts): New function.
16270 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16271 (const_vector_int_elt, const_vector_elt): Declare.
16272 * emit-rtl.c (const_vector_int_elt_1): New function.
16273 (const_vector_elt): Likewise.
16274 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16275 of CONST_VECTOR_ELT.
16277 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16279 * expr.c: Include rtx-vector-builder.h.
16280 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16281 directly on the tree encoding.
16282 (const_vector_from_tree): Likewise.
16283 * optabs.c: Include rtx-vector-builder.h.
16284 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16285 sequence of "u" values.
16286 * vec-perm-indices.c: Include rtx-vector-builder.h.
16287 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16288 directly on the vec_perm_indices encoding.
16290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16292 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16293 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16294 * rtx-vector-builder.h: New file.
16295 * rtx-vector-builder.c: Likewise.
16296 * rtl.h (rtx_def::u2): Add a const_vector field.
16297 (CONST_VECTOR_NPATTERNS): New macro.
16298 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16299 (CONST_VECTOR_DUPLICATE_P): Likewise.
16300 (CONST_VECTOR_STEPPED_P): Likewise.
16301 (CONST_VECTOR_ENCODED_ELT): Likewise.
16302 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16303 (unwrap_const_vec_duplicate): Likewise.
16304 (const_vec_series_p): Check for a non-duplicated vector encoding.
16305 Say that the function only returns true for integer vectors.
16306 * emit-rtl.c: Include rtx-vector-builder.h.
16307 (gen_const_vec_duplicate_1): Delete.
16308 (gen_const_vector): Call gen_const_vec_duplicate instead of
16309 gen_const_vec_duplicate_1.
16310 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16311 (gen_const_vec_duplicate): Use rtx_vector_builder.
16312 (gen_const_vec_series): Likewise.
16313 (gen_rtx_CONST_VECTOR): Likewise.
16314 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16315 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16316 Build a new vector rather than modifying a CONST_VECTOR in-place.
16317 (handle_special_swappables): Update call accordingly.
16318 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16319 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16320 Build a new vector rather than modifying a CONST_VECTOR in-place.
16321 (handle_special_swappables): Update call accordingly.
16323 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16325 * simplify-rtx.c (simplify_const_binary_operation): Use
16326 CONST_VECTOR_ELT instead of XVECEXP.
16328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16330 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16331 the selector elements to be different from the data elements
16332 if the selector is a VECTOR_CST.
16333 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16334 ssizetype for the selector.
16336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16338 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16339 before testing each element individually.
16340 * tree-vect-generic.c (lower_vec_perm): Likewise.
16342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16344 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16345 * selftest-run-tests.c (selftest::run_tests): Call it.
16346 * vector-builder.h (vector_builder::operator ==): New function.
16347 (vector_builder::operator !=): Likewise.
16348 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16349 (vec_perm_indices::all_from_input_p): New function.
16350 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16351 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16352 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16353 instead of reading the VECTOR_CST directly. Detect whether both
16354 vector inputs are the same before constructing the vec_perm_indices,
16355 and update the number of inputs argument accordingly. Use the
16356 utility functions added above. Only construct sel2 if we need to.
16358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16360 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16361 the broadcast of the low byte.
16362 (expand_mult_highpart): Use an explicit encoding for the permutes.
16363 * optabs-query.c (can_mult_highpart_p): Likewise.
16364 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16365 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16366 (vectorizable_bswap): Likewise.
16367 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16368 explicit encoding for the power-of-2 permutes.
16369 (vect_permute_store_chain): Likewise.
16370 (vect_grouped_load_supported): Likewise.
16371 (vect_permute_load_chain): Likewise.
16373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16375 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16376 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16377 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16378 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16379 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16380 (vect_gen_perm_mask_any): Likewise.
16382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16384 * int-vector-builder.h: New file.
16385 * vec-perm-indices.h: Include int-vector-builder.h.
16386 (vec_perm_indices): Redefine as an int_vector_builder.
16387 (auto_vec_perm_indices): Delete.
16388 (vec_perm_builder): Redefine as a stand-alone class.
16389 (vec_perm_indices::vec_perm_indices): New function.
16390 (vec_perm_indices::clamp): Likewise.
16391 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16392 (vec_perm_indices::new_vector): New function.
16393 (vec_perm_indices::new_expanded_vector): Update for new
16394 vec_perm_indices class.
16395 (vec_perm_indices::rotate_inputs): New function.
16396 (vec_perm_indices::all_in_range_p): Operate directly on the
16397 encoded form, without computing elided elements.
16398 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16399 encoding. Update for new vec_perm_indices class.
16400 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16401 the given vec_perm_builder.
16402 (expand_vec_perm_var): Update vec_perm_builder constructor.
16403 (expand_mult_highpart): Use vec_perm_builder instead of
16404 auto_vec_perm_indices.
16405 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16406 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16407 or double series encoding as appropriate.
16408 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16409 vec_perm_indices instead of auto_vec_perm_indices.
16410 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16411 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16412 (vect_permute_store_chain): Likewise.
16413 (vect_grouped_load_supported): Likewise.
16414 (vect_permute_load_chain): Likewise.
16415 (vect_shift_permute_load_chain): Likewise.
16416 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16417 (vect_transform_slp_perm_load): Likewise.
16418 (vect_schedule_slp_instance): Likewise.
16419 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16420 (vectorizable_mask_load_store): Likewise.
16421 (vectorizable_bswap): Likewise.
16422 (vectorizable_store): Likewise.
16423 (vectorizable_load): Likewise.
16424 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16425 vec_perm_indices instead of auto_vec_perm_indices. Use
16426 tree_to_vec_perm_builder to read the vector from a tree.
16427 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16428 vec_perm_builder instead of a vec_perm_indices.
16429 (have_whole_vector_shift): Use vec_perm_builder and
16430 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16431 truncation to calc_vec_perm_mask_for_shift.
16432 (vect_create_epilog_for_reduction): Likewise.
16433 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16434 from auto_vec_perm_indices to vec_perm_indices.
16435 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16436 instead of changing individual elements.
16437 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16438 the vector in d.perm.
16439 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16440 from auto_vec_perm_indices to vec_perm_indices.
16441 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16442 instead of changing individual elements.
16443 (arm_vectorize_vec_perm_const): Use new_vector to install
16444 the vector in d.perm.
16445 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16446 Update vec_perm_builder constructor.
16447 (rs6000_expand_interleave): Likewise.
16448 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16449 (rs6000_expand_interleave): Likewise.
16451 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16453 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16454 to qimode could truncate the indices.
16455 * optabs.c (expand_vec_perm_var): Likewise.
16457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16459 * Makefile.in (OBJS): Add vec-perm-indices.o.
16460 * vec-perm-indices.h: New file.
16461 * vec-perm-indices.c: Likewise.
16462 * target.h (vec_perm_indices): Replace with a forward class
16464 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16465 * optabs.h: Include vec-perm-indices.h.
16466 (expand_vec_perm): Delete.
16467 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16468 (expand_vec_perm_const): Declare.
16469 * target.def (vec_perm_const_ok): Replace with...
16470 (vec_perm_const): ...this new hook.
16471 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16472 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16473 * doc/tm.texi: Regenerate.
16474 * optabs.def (vec_perm_const): Delete.
16475 * doc/md.texi (vec_perm_const): Likewise.
16476 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16477 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16478 expand_vec_perm for constant permutation vectors. Assert that
16479 the mode of variable permutation vectors is the integer equivalent
16480 of the mode that is being permuted.
16481 * optabs-query.h (selector_fits_mode_p): Declare.
16482 * optabs-query.c: Include vec-perm-indices.h.
16483 (selector_fits_mode_p): New function.
16484 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16485 is defined, instead of checking whether the vec_perm_const_optab
16486 exists. Use targetm.vectorize.vec_perm_const instead of
16487 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16488 fit in the vector mode before using a variable permute.
16489 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16490 vec_perm_indices instead of an rtx.
16491 (expand_vec_perm): Replace with...
16492 (expand_vec_perm_const): ...this new function. Take the selector
16493 as a vec_perm_indices rather than an rtx. Also take the mode of
16494 the selector. Update call to shift_amt_for_vec_perm_mask.
16495 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16496 Use vec_perm_indices::new_expanded_vector to expand the original
16497 selector into bytes. Check whether the indices fit in the vector
16498 mode before using a variable permute.
16499 (expand_vec_perm_var): Make global.
16500 (expand_mult_highpart): Use expand_vec_perm_const.
16501 * fold-const.c: Includes vec-perm-indices.h.
16502 * tree-ssa-forwprop.c: Likewise.
16503 * tree-vect-data-refs.c: Likewise.
16504 * tree-vect-generic.c: Likewise.
16505 * tree-vect-loop.c: Likewise.
16506 * tree-vect-slp.c: Likewise.
16507 * tree-vect-stmts.c: Likewise.
16508 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16510 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16511 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16512 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16513 (aarch64_vectorize_vec_perm_const): ...this new function.
16514 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16515 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16516 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16517 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16518 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16519 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16520 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16522 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16523 check for NEON modes.
16524 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16525 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16526 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16527 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16529 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16530 the old VEC_PERM_CONST conditions.
16531 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16532 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16533 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16534 (ia64_vectorize_vec_perm_const_ok): Merge into...
16535 (ia64_vectorize_vec_perm_const): ...this new function.
16536 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16537 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16538 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16539 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16540 * config/mips/mips.c (mips_expand_vec_perm_const)
16541 (mips_vectorize_vec_perm_const_ok): Merge into...
16542 (mips_vectorize_vec_perm_const): ...this new function.
16543 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16544 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16545 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16546 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16547 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16548 (rs6000_expand_vec_perm_const): Delete.
16549 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16551 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16552 (altivec_expand_vec_perm_const_le): Take each operand individually.
16553 Operate on constant selectors rather than rtxes.
16554 (altivec_expand_vec_perm_const): Likewise. Update call to
16555 altivec_expand_vec_perm_const_le.
16556 (rs6000_expand_vec_perm_const): Delete.
16557 (rs6000_vectorize_vec_perm_const_ok): Delete.
16558 (rs6000_vectorize_vec_perm_const): New function.
16559 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16560 an element count and rtx array.
16561 (rs6000_expand_extract_even): Update call accordingly.
16562 (rs6000_expand_interleave): Likewise.
16563 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16564 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16565 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16566 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16567 (rs6000_expand_vec_perm_const): Delete.
16568 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16569 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16570 (altivec_expand_vec_perm_const_le): Take each operand individually.
16571 Operate on constant selectors rather than rtxes.
16572 (altivec_expand_vec_perm_const): Likewise. Update call to
16573 altivec_expand_vec_perm_const_le.
16574 (rs6000_expand_vec_perm_const): Delete.
16575 (rs6000_vectorize_vec_perm_const_ok): Delete.
16576 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16577 reference to the SPE evmerge intructions.
16578 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16579 an element count and rtx array.
16580 (rs6000_expand_extract_even): Update call accordingly.
16581 (rs6000_expand_interleave): Likewise.
16582 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16583 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16585 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16587 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16589 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16590 vector mode and that that mode matches the mode of the data
16592 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16593 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16594 directly using expand_vec_perm_1 when forcing selectors into
16596 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16600 * optabs-query.h (can_vec_perm_p): Delete.
16601 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16602 * optabs-query.c (can_vec_perm_p): Split into...
16603 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16604 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16605 particular selector is valid.
16606 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16607 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16608 (vect_grouped_load_supported): Likewise.
16609 (vect_shift_permute_load_chain): Likewise.
16610 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16611 (vect_transform_slp_perm_load): Likewise.
16612 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16613 (vectorizable_bswap): Likewise.
16614 (vect_gen_perm_mask_checked): Likewise.
16615 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16616 implementations of variable permutation vectors into account
16617 when deciding which selector to use.
16618 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16619 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16620 with a false third argument.
16621 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16622 to test whether the constant selector is valid and can_vec_perm_var_p
16623 to test whether a variable selector is valid.
16625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16627 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16628 * optabs-query.c (can_vec_perm_p): Likewise.
16629 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16630 instead of vec_perm_indices.
16631 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16632 (vect_gen_perm_mask_checked): Likewise,
16633 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16634 (vect_gen_perm_mask_checked): Likewise,
16636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16638 * optabs-query.h (qimode_for_vec_perm): Declare.
16639 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16640 (qimode_for_vec_perm): ...this new function.
16641 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16643 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16645 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16646 does not have a conditional at the top.
16648 2018-01-02 Richard Biener <rguenther@suse.de>
16650 * ipa-inline.c (big_speedup_p): Fix expression.
16652 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16655 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16658 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16662 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16663 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16664 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16665 cond_taken_branch_cost 3->4.
16667 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16669 PR tree-optimization/83581
16670 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16671 TODO_cleanup_cfg if any changes have been made.
16673 PR middle-end/83608
16674 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16675 convert_modes if target mode has the right side, but different mode
16678 PR middle-end/83609
16679 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16680 last argument when extracting from CONCAT. If either from_real or
16681 from_imag is NULL, use expansion through memory. If result is not
16682 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16683 the parts directly to inner mode, if even that fails, use expansion
16686 PR middle-end/83623
16687 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16688 check for bswap in mode rather than HImode and use that in expand_unop
16691 Copyright (C) 2018 Free Software Foundation, Inc.
16693 Copying and distribution of this file, with or without modification,
16694 are permitted in any medium without royalty provided the copyright
16695 notice and this notice are preserved.