1 2018-06-15 Richard Biener <rguenther@suse.de>
4 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
5 leave useless conversion stripping to force_gimple_operand_gsi.
6 (gimplify_build2): Likewise.
7 (gimplify_build1): Likewise.
9 2018-06-15 Richard Biener <rguenther@suse.de>
12 * tree-cfg.c (move_stmt_op): unshare invariant addresses
13 before adjusting their block.
15 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
17 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
18 multilibs for *-*-rtems*.
19 * config/riscv/t-rtems: New file.
21 2018-06-14 Jakub Jelinek <jakub@redhat.com>
24 * match.pd ((A +- CST1) +- CST2): Punt if last resort
25 unsigned_type_for returns NULL.
28 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
29 subregs of multi-word pseudos unless the float mode has word size.
31 2018-06-14 Richard Biener <rguenther@suse.de>
34 * tree-vect-generic.c (build_word_mode_vector_type): Remove
35 duplicate and harmful type_hash_canon.
36 * tree.c (type_hash_canon): Assert we didn't find ourselves.
38 2018-06-14 Richard Biener <rguenther@suse.de>
41 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
44 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
46 * config/rtems.h (STDINT_LONG32): Define.
48 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
49 Prachi Godbole <prachi.godbole@imgtec.com>
51 * config/mips/mips-cpus.def: Define P6600.
52 * config/mips/mips-tables.opt: Regenerate.
53 * config/mips/mips.c (mips_ucbranch_type): New enum.
54 (mips_rtx_cost_data): Add support for P6600.
55 (mips_issue_rate): Likewise.
56 (mips_multipass_dfa_lookahead): Likewise.
57 (mips_avoid_hazard): Likewise.
58 (mips_reorg_process_insns): Likewise.
59 (mips_classify_branch_p6600): New function.
60 * config/mips/mips.h (TUNE_P6600): New define.
61 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
62 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
63 * config/mips/mips.md: Include p6600.md.
64 (processor): Add p6600.
65 * config/mips/p6600.md: New file.
66 * doc/invoke.texi: Add p6600 to supported architectures.
68 2018-06-13 Martin Sebor <msebor@redhat.com>
70 PR tree-optimization/86114
71 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
73 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
75 2018-06-13 Richard Biener <rguenther@suse.de>
77 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
78 Properly set vector type of the intermediate stmt.
79 * tree-vect-stmts.c (vectorizable_operation): The destination
80 var always has vectype_out type.
82 2018-06-13 Jeff Law <law@redhat.com>
84 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
85 integer 0 for argument to print_rtl_with_bb.
86 (rl78_reorg): Likewise.
88 2018-06-13 David Malcolm <dmalcolm@redhat.com>
90 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
91 from rtx to rtx_insn *.
92 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
94 (add_sched_insns_for_speculation): Likewise for local "target",
95 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
96 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
97 from rtx_insn ** to rtx_code_label **.
98 (reorg_emit_nops): Likewise.
99 (c6x_reorg): Likewise for local "call_labels".
100 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
102 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
103 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
104 the loops over LABEL_REFS.
105 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
107 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
108 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
109 (split_branches): Strengthen local "olabel" from rtx to
110 rtx_insn *, adding a safe_as_a cast.
111 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
113 (add_insn_after): Likewise for first two params.
114 (add_insn_before): Likewise.
115 (remove_insn): Likewise for param.
116 (emit_pattern_before_noloc): Likewise for second and third params.
117 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
118 (emit_call_insn_before_noloc): Likewise.
119 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
121 (emit_barrier_before): Likewise.
122 (emit_label_before): Strengthen "label" param from "rtx" to
123 "rtx_code_label *". Strengthen "before" param from "rtx" to
125 (emit_insn_after_1): Strengthen "after" param from "rtx" to
127 (emit_pattern_after_noloc): Likewise.
128 (emit_insn_after_noloc): Likewise.
129 (emit_jump_insn_after_noloc): Likewise.
130 (emit_call_insn_after_noloc): Likewise.
131 (emit_debug_insn_after_noloc): Likewise.
132 (emit_barrier_after): Likewise.
133 (emit_label_after): Likewise for both params.
134 (emit_pattern_after_setloc): Likewise for "after" param. Convert
135 "loc" param from "int" to "location_t".
136 (emit_insn_after_setloc): Likewise.
137 (emit_jump_insn_after_setloc): Likewise.
138 (emit_call_insn_after_setloc): Likewise.
139 (emit_debug_insn_after_setloc): Likewise.
140 (emit_pattern_before_setloc): Likewise for "before" param. Convert
141 "loc" param from "int" to "location_t".
142 (emit_pattern_before): Convert NULL_RTX to NULL.
143 (emit_insn_before_setloc): Convert "loc" param from "int" to
145 (emit_jump_insn_before_setloc): Likewise.
146 (emit_call_insn_before_setloc): Likewise.
147 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
148 rtx_insn *. Convert "loc" param from "int" to "location_t".
149 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
150 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
151 Convert 3rd param from "int" to "location_t".
152 (emit_barrier_before, emit_barrier_after, next_real_insn):
153 Strengthen param from rtx to rtx_insn *.
154 (emit_label_before): Strengthen 1st param from "rtx" to
155 "rtx_code_label *". Strengthen 2nd param from "rtx" to
157 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
158 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
159 Strengthen 2nd param from "rtx" to "rtx_insn *".
160 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
161 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
162 Likewise. Convert 3rd param from "int" to "location_t".
163 (emit_label_after): Strengthen 1st param from "rtx" to
165 (next_real_insn, remove_insn): Strengthen param from "rtx" to
167 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
168 from "rtx" to "rtx_insn *".
170 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
172 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
173 bodies streamed in with -Q.
174 * dumpfile.c (dump_files): Add lto-stream-out dump file.
175 * dumpfile.h (tree_dump_index): Add lto_stream_out.
176 * gimple-streamer-out.c: Include gimple-pretty-print.h
177 (output_bb): Dump stmts streamed.
178 * lto-section-out.c: Include print-tree.h
179 (lto_begin_section): Dump sections created.
180 (lto_output_decl_index): Dump decl encoded.
181 * lto-streamer-out.c: Include print-tree.h
182 (create_output_block): Dump output block created.
183 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
184 (output_function): Dump function output.
185 (output_constructor): Dump constructor streamed.
186 (write_global_stream): Output indexes encoded.
187 (produce_asm_for_decls): Dump streams encoded.
188 * lto-streamer.c (streamer_dump_file): New global var.
189 * lto-streamer.h (streamer_dump_file): Declare.
190 * passes.c (ipa_write_summaries): Initialize streamer dump.
191 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
194 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
197 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
198 offsets for register save directives. Emit a second batch of save
199 directives, if need be, when the function accesses prior frames.
201 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
203 * config/arc/fpu.md (fmasf4): Force operand to register.
206 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
208 * config/arc/arc-protos.h (arc_pad_return): Remove.
209 * config/arc/arc.c (machine_function): Remove force_short_suffix
211 (arc_print_operand): Adjust printing of '&'.
212 (arc_verify_short): Remove conditional printing of short suffix.
213 (arc_final_prescan_insn): Remove reference to size_reason.
214 (pad_return): New function.
215 (arc_reorg): Call pad_return.
216 (arc_pad_return): Remove.
217 (arc_init_machine_status): Remove reference to force_short_suffix.
218 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
219 (attr length): When attribute iscompact is true force to 2
220 regardless; in the case of maybe check if we want to force the
221 instruction to have 4 bytes length.
222 (nopv): Change it to generate 4 byte long nop as well.
223 (blockage): New pattern.
224 (simple_return): Remove call to arc_pad_return.
225 (p_return_i): Likewise.
227 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
229 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
231 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
233 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
236 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
238 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
239 for ARC700 and ARCv2.
241 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
244 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
245 operands[2] instead of operands[1].
248 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
250 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
251 case, check whether the outer register overlaps an unallocatable
252 register, not just whether it fits the required class.
254 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
256 * poly-int.h (can_div_trunc_p): Add new overload in which all values
258 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
259 (memrefs_conflict_p): Likewise.
260 (init_alias_analysis): Likewise.
261 * cfgexpand.c (expand_debug_expr): Likewise.
262 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
263 * cse.c (fold_rtx): Likewise.
264 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
265 * expr.c (emit_block_move_hints): Likewise.
266 (clear_storage_hints, push_block, emit_push_insn): Likewise.
267 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
268 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
269 (emit_group_store): Likewise.
270 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
271 to read the PRE/POST_MODIFY increment.
272 * calls.c (store_one_arg): Use strip_offset.
273 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
275 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
277 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
278 (simplify_binary_operation_1): Extend CONST_INT handling to
280 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
281 than a HOST_WIDE_INT.
282 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
284 (adjust_mems, add_stores): Update accodingly.
285 (vt_canonicalize_addr): Track polynomial offsets.
286 (emit_note_insn_var_location): Likewise.
287 (vt_add_function_parameter): Likewise.
288 (vt_initialize): Likewise.
290 2018-06-12 Jeff Law <law@redhat.com>
292 * config.gcc (alpha*-*-freebsd*): Remove.
293 * config/alpha/freebsd.h: Remove.
295 2018-06-12 David Malcolm <dmalcolm@redhat.com>
298 * spellcheck-tree.c (levenshtein_distance): Rename to...
299 (get_edit_distance): ...this, and update for underlying renaming.
300 * spellcheck-tree.h (levenshtein_distance): Rename to...
301 (get_edit_distance): ...this.
302 * spellcheck.c (levenshtein_distance): Rename to...
303 (get_edit_distance): ...this. Convert from Levenshtein distance
304 to Damerau-Levenshtein distance by supporting transpositions of
305 adjacent characters. Rename "v1" to "v_next" and "v0" to
307 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
308 (selftest::test_edit_distance_unit_test_oneway): ...this, and
309 update for underlying renaming.
310 (selftest::levenshtein_distance_unit_test): Rename to...
311 (selftest::test_get_edit_distance_unit): ...this, and update for
313 (selftest::test_find_closest_string): Add example from PR 69968
314 where transposition helps
315 (selftest::test_metric_conditions): Update for renaming.
316 (selftest::test_metric_conditions): Likewise.
317 (selftest::spellcheck_c_tests): Likewise.
318 * spellcheck.h (levenshtein_distance): Rename both overloads to...
319 (get_edit_distance): ...this.
320 (best_match::consider): Update for renaming.
322 2018-06-12 Martin Sebor <msebor@redhat.com>
324 PR tree-optimization/85259
325 * builtins.c (compute_objsize): Handle constant offsets.
326 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
327 true iff a warning has been issued.
328 * gimple.h (gimple_nonartificial_location): New function.
329 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
330 gimple_nonartificial_location and handle -Wno-system-headers.
331 (handle_builtin_stxncpy): Same.
333 2018-06-12 Martin Sebor <msebor@redhat.com>
336 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
338 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
340 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
341 BUILTIN_VEC_XST entries for pointer to double and long long.
343 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
346 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
348 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
351 2018-06-12 Martin Liska <mliska@suse.cz>
353 * doc/options.texi: Document IntegerRange.
355 2018-06-12 Martin Liska <mliska@suse.cz>
357 * config/i386/i386.opt: Make MPX-related options as Deprecated.
358 * opt-functions.awk: Handle Deprecated flag.
359 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
361 (read_cmdline_option): Report warning for a deprecated option.
362 * opts.h (struct cl_option): Add new field cl_deprecated.
363 (CL_ERR_DEPRECATED): New.
365 2018-06-12 Martin Liska <mliska@suse.cz>
367 * doc/options.texi: Document Deprecated option flag.
369 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
371 * config/arc/arc-arch.h (arc_extras): New enum.
372 (arc_cpu_t):Add field extra.
373 (arc_cpu_types): Consider the extras.
374 * config/arc/arc-cpus.def: Add extras info.
375 * config/arc/arc-opts.h (processor_type): Consider extra field.
376 * config/arc/arc.c (arc_override_options): Handle extra field.
378 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
380 * config/arc/arc-arch.h: Update ARC_OPTX macro.
381 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
383 * config/arc/arc.c (arc_init): Update pic warning.
384 (irq_range): Update irq range parsing warnings.
385 (arc_override_options): Update various warning messages.
386 (arc_handle_aux_attribute): Likewise.
388 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
390 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
392 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
394 * doc/sourcebuild.texi: Document usage of line number 0 in verify
395 compiler messages directives.
397 2018-06-12 Matthew Fortune <mfortune@gmail.com>
399 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
400 * config/mips/mips-tables.opt: Regenerate.
401 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
403 * doc/invoke.texi: Document -march=i6500.
405 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
407 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
408 (i6400_gpmul): Add cpu_unit.
409 (i6400_gpdiv): Likewise.
410 (i6400_msa_add_d): Update reservations.
411 (i6400_msa_int_add) Likewise.
412 (i6400_msa_short_logic3) Likewise.
413 (i6400_msa_short_logic2) Likewise.
414 (i6400_msa_short_logic) Likewise.
415 (i6400_msa_move) Likewise.
416 (i6400_msa_cmp) Likewise.
417 (i6400_msa_short_float2) Likewise.
418 (i6400_msa_div_d) Likewise.
419 (i6400_msa_long_logic1) Likewise.
420 (i6400_msa_long_logic2) Likewise.
421 (i6400_msa_mult) Likewise.
422 (i6400_msa_long_float2) Likewise.
423 (i6400_msa_long_float4) Likewise.
424 (i6400_msa_long_float5) Likewise.
425 (i6400_msa_long_float8) Likewise.
426 (i6400_fpu_fadd): Include frint type.
427 (i6400_fpu_store): New define_insn_reservation.
428 (i6400_fpu_load): Likewise.
429 (i6400_fpu_move): Likewise.
430 (i6400_fpu_fcmp): Likewise.
431 (i6400_fpu_fmadd): Likewise.
432 (i6400_int_mult): Include imul3nc type and update reservation.
433 (i6400_int_div): Include idiv3 type and update reservation.
434 (i6400_int_load): Update to check type not move_type.
435 (i6400_int_store): Likewise.
436 (i6400_int_prefetch): Set zero latency.
438 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
440 * gcc.c: Document new %@{...} sequence.
441 (LINK_COMMAND_SPEC): Use it for the -L switches.
442 (cpp_unique_options): Use it for the -I switches.
443 (at_file_argbuf): New global variable.
444 (in_at_file): Likewise.
445 (alloc_args): Create at_file_argbuf.
446 (clear_args): Truncate at_file_argbuf.
447 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
448 (open_at_file): New function.
449 (close_at_file): Likewise.
450 (create_at_file): Delete.
451 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
454 (validate_switches_from_spec): Deal with %@{...} sequence.
455 (validate_switches): Likewise.
456 (driver::finalize): Call clear_args.
458 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
460 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
462 2018-06-11 Martin Sebor <msebor@redhat.com>
464 * doc/invoke.texi (-Wall): List -Wc++17-compat.
465 (Wno-class-memaccess): Add @opindex.
466 (Wno-templates, Wno-multiple-inheritance): Same.
467 (Wno-virtual-inheritance, Wno-namespaces): Same.
468 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
469 (Wno-format-overflow, Wno-format-truncation): Same.
470 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
471 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
472 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
473 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
474 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
475 (Wno-misspelled-isr): Same.
477 2018-06-11 Martin Sebor <msebor@redhat.com>
479 * PR tree-optimization/86083
480 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
482 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
484 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
486 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
489 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
490 on the correct operand.
491 (*movdi_internal64): Ditto.
493 2018-06-11 Martin Liska <mliska@suse.cz>
495 PR tree-optimization/86089
496 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
498 2018-06-11 Julia Koval <julia.koval@intel.com>
500 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
501 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
502 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
504 2018-06-11 Olivier Hainque <hainque@adacore.com>
506 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
507 for Ada with strict dwarf2.
509 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
512 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
515 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
517 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
519 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
521 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
524 2018-06-08 Martin Liska <mliska@suse.cz>
526 * tree-cfg.h (debug_function): Fix argument type to match
529 2018-06-08 Martin Liska <mliska@suse.cz>
531 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
532 Remove usage of MPX-related (and removed) fields.
533 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
535 2018-06-08 David Malcolm <dmalcolm@redhat.com>
537 * cfg.c (debug): Use TDF_NONE rather than 0.
538 * cfghooks.c (debug): Likewise.
539 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
540 (struct dump_option_value_info): Convert to...
541 (struct kv_pair): ...this template type.
542 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
544 (optinfo_verbosity_options): Likewise.
545 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
547 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
548 than int for "optgroup_flags" param.
549 (dump_generic_expr_loc): Use dump_flags_t rather than int for
551 (dump_dec): Likewise.
552 (dump_finish): Use TDF_NONE rather than 0.
553 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
554 rather than int for "optgroup_flags" param. Use TDF_NONE rather
555 than 0. Update for change to option_ptr.
556 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
557 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
558 0. Update for changes to optinfo_verbosity_options and
560 (opt_info_switch_p): Convert optgroup_flags from int to
562 (dump_basic_block): Use dump_flags_t rather than int
563 for "dump_kind" param.
564 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
565 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
566 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
567 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
568 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
569 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
570 TDF_NONE): Convert from macros to...
571 (enum dump_flag): ...this new enum.
572 (dump_flags_t): Update to use enum.
573 (operator|, operator&, operator~, operator|=, operator&=):
574 Implement for dump_flags_t.
575 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
576 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
577 Convert from macros to...
578 (enum optgroup_flag): ...this new enum.
579 (optgroup_flags_t): New typedef.
580 (operator|, operator|=): Implement for optgroup_flags_t.
581 (struct dump_file_info): Convert field "alt_flags" to
582 dump_flags_t. Convert field "optgroup_flags" to
584 (dump_basic_block): Use dump_flags_t rather than int for param.
585 (dump_generic_expr_loc): Likewise.
586 (dump_dec): Likewise.
587 (dump_register): Convert param "optgroup_flags" to
589 (opt_info_enable_passes): Likewise.
590 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
592 * gimple-pretty-print.c (debug): Likewise.
593 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
594 (merged_store_group::apply_stores): Likewise.
595 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
596 * gimple.c (verify_gimple_pp): Likewise.
597 * graphite-poly.c (print_pbb_body): Likewise.
598 * passes.c (pass_manager::register_one_dump_file): Convert
599 local "optgroup_flags" to optgroup_flags_t.
600 * print-tree.c (print_node): Use TDF_NONE rather than 0.
602 (debug_body): Likewise.
603 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
605 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
607 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
608 (convert_mult_to_fma): Likewise.
609 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
610 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
611 * tree-vect-data-refs.c (dump_lower_bound): Convert param
612 "dump_kind" to dump_flags_t.
614 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
616 * config/rs6000/rs6000.c (min, max): Delete.
618 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
620 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
621 -mabi=spe and -mabi=no-spe.
623 2018-06-08 Martin Liska <mliska@suse.cz>
625 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
626 where we expect an existing summary.
628 2018-06-08 Martin Liska <mliska@suse.cz>
630 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
631 * ipa-inline.h (estimate_edge_growth): Likewise.
633 2018-06-08 Martin Liska <mliska@suse.cz>
635 * cgraph.c (function_version_hasher::hash): Use
636 cgraph_node::get_uid ().
637 (function_version_hasher::equal):
638 * cgraph.h (cgraph_node::get_uid): New method.
639 * ipa-inline.c (update_caller_keys): Use
640 cgraph_node::get_uid ().
641 (update_callee_keys): Likewise.
642 * ipa-utils.c (searchc): Likewise.
643 (ipa_reduced_postorder): Likewise.
644 * lto-cgraph.c (input_node): Likewise.
645 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
646 * symbol-summary.h (symtab_insertion): Likewise.
647 (symtab_removal): Likewise.
648 (symtab_duplication): Likewise.
649 * tree-pretty-print.c (dump_function_header): Likewise.
650 * tree-sra.c (convert_callers_for_node): Likewise.
652 2018-06-08 Martin Liska <mliska@suse.cz>
654 * cgraph.c (symbol_table::create_edge): Always assign a new
656 (symbol_table::free_edge): Do not recycle numbers.
657 * cgraph.h (cgraph_edge::get): New method.
658 * symbol-summary.h (symtab_removal): Use it.
659 (symtab_duplication): Likewise.
660 (call_summary::hashable_uid): Remove.
662 2018-06-08 Martin Liska <mliska@suse.cz>
664 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
665 (initialize_growth_caches): Remove.
666 (free_growth_caches): Likewise.
667 (do_estimate_edge_time): Use edge_growth_cache.
668 (do_estimate_edge_size): Likewise.
669 (do_estimate_edge_hints): Likewise.
670 * ipa-inline.c (reset_edge_caches): Likewise.
671 (recursive_inlining): Likewise.
672 (inline_small_functions): Likewise.
673 * ipa-inline.h (initialize_growth_caches): Remove.
674 (estimate_edge_size): Likewise.
675 (estimate_edge_time): Likewise.
676 (estimate_edge_hints): Likewise.
677 (reset_edge_growth_cache): Likewise.
678 * symbol-summary.h (call_summary::remove): New method.
680 2018-06-08 Martin Liska <mliska@suse.cz>
682 * ipa-cp.c (class edge_clone_summary): New summary.
683 (grow_edge_clone_vectors): Remove.
684 (ipcp_edge_duplication_hook): Remove.
685 (class edge_clone_summary_t): New call_summary class.
686 (ipcp_edge_removal_hook): Remove.
687 (edge_clone_summary_t::duplicate): New function.
688 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
689 (create_specialized_node): Likewise.
690 (ipcp_driver): Initialize edge_clone_summaries and do not
693 2018-06-08 Martin Liska <mliska@suse.cz>
695 * symbol-summary.h (get): New function.
696 (call_summary::m_initialize_when_cloning): New class member.
698 2018-06-08 Martin Liska <mliska@suse.cz>
700 * cgraph.c (cgraph_node::remove): Do not recycle uid.
701 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
702 (symbol_table::allocate_cgraph_symbol): Do not set uid.
703 * passes.c (uid_hash_t): Record removed_nodes by their uids.
704 (remove_cgraph_node_from_order): Use the removed_nodes set.
705 (do_per_function_toporder): Likwise.
706 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
707 instead of summary_uid.
708 (symtab_removal): Likewise.
709 (symtab_duplication): Likewise.
711 2018-06-08 Martin Liska <mliska@suse.cz>
713 * ipa-cp.c (ipcp_store_bits_results): Use
714 ipcp_transformation_sum.
715 (ipcp_store_vr_results): Likewise.
716 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
718 (ipcp_transformation_initialize): ... this.
719 (ipa_set_node_agg_value_chain):
720 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
721 (write_ipcp_transformation_info): Likewise.
722 (read_ipcp_transformation_info): Likewise.
723 (ipcp_update_bits): Likewise.
724 (ipcp_update_vr): Likewise.
725 (ipcp_transform_function): Likewise.
726 * ipa-prop.h: Rename ipcp_transformation_summary to
728 (class ipcp_transformation_t): New function summary.
729 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
730 (ipa_get_agg_replacements_for_node): Likewise.
732 2018-06-08 Martin Liska <mliska@suse.cz>
734 * ipa-pure-const.c (struct funct_state_d): Do it class instead
736 (class funct_state_summary_t): New function_summary class.
737 (has_function_state): Remove.
738 (get_function_state): Likewise.
739 (set_function_state): Likewise.
740 (add_new_function): Likewise.
741 (funct_state_summary_t::insert): New function.
742 (duplicate_node_data): Remove.
743 (remove_node_data): Remove.
744 (funct_state_summary_t::duplicate): New function.
745 (register_hooks): Create new funct_state_summaries.
746 (pure_const_generate_summary): Use it.
747 (pure_const_write_summary): Likewise.
748 (pure_const_read_summary): Likewise.
749 (propagate_pure_const): Likewise.
750 (propagate_nothrow): Likewise.
751 (dump_malloc_lattice): Likewise.
752 (propagate_malloc): Likewise.
753 (execute): Do not register hooks, just remove summary
755 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
758 2018-06-08 Martin Liska <mliska@suse.cz>
760 * ipa-reference.c (remove_node_data): Remove.
761 (duplicate_node_data): Likewise.
762 (class ipa_ref_var_info_summary_t): New class.
763 (class ipa_ref_opt_summary_t): Likewise.
764 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
765 (get_reference_optimization_summary): Use
766 ipa_ref_opt_sum_summaries.
767 (set_reference_vars_info): Remove.
768 (set_reference_optimization_summary): Likewise.
769 (ipa_init): Create summaries.
770 (init_function_info): Use function summary.
771 (ipa_ref_opt_summary_t::duplicate): New function.
772 (ipa_ref_opt_summary_t::remove): New function.
773 (get_read_write_all_from_node): Fix GNU coding style.
774 (propagate): Use function summary.
775 (write_node_summary_p): Fix GNU coding style.
776 (stream_out_bitmap): Likewise.
777 (ipa_reference_read_optimization_summary): Use function summary.
778 (ipa_reference_c_finalize): Do not release hooks.
780 2018-06-08 Martin Liska <mliska@suse.cz>
782 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
783 (analyze_function_body): Extract multiple calls of get_create.
784 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
785 * ipa-inline.c (recursive_inlining): Use ::get method.
786 * ipa-inline.h (estimate_edge_growth): Likewise.
788 2018-06-08 Martin Liska <mliska@suse.cz>
790 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
792 (hsa_function_summary::hsa_function_summary): Use the new enum
794 (hsa_gpu_implementation_p): Use hsa_summaries::get.
795 * hsa-gen.c (hsa_get_host_function): Likewise.
796 (get_brig_function_name): Likewise.
797 * ipa-hsa.c (process_hsa_functions): Likewise.
798 (ipa_hsa_write_summary): Likewise.
799 * symbol-summary.h (symtab_duplication): Use ::get function/
802 2018-06-08 Martin Liska <mliska@suse.cz>
804 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
806 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
807 (hsa_register_kernel): Likewise.
808 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
809 * hsa-gen.c (hsa_get_host_function): Likewise.
810 (get_brig_function_name): Likewise.
811 (generate_hsa): Likewise.
812 (pass_gen_hsail::execute): Likewise.
813 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
814 (devirtualization_time_bonus): Likewise.
815 (ipcp_propagate_stage): Likewise.
816 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
817 (edge_set_predicate): Likewise.
818 (evaluate_conditions_for_known_args): Likewise.
819 (evaluate_properties_for_edge): Likewise.
820 (ipa_fn_summary::reset): Likewise.
821 (ipa_fn_summary_t::duplicate): Likewise.
822 (dump_ipa_call_summary): Likewise.
823 (ipa_dump_fn_summary): Likewise.
824 (analyze_function_body): Likewise.
825 (compute_fn_summary): Likewise.
826 (estimate_edge_devirt_benefit): Likewise.
827 (estimate_edge_size_and_time): Likewise.
828 (estimate_calls_size_and_time): Likewise.
829 (estimate_node_size_and_time): Likewise.
830 (inline_update_callee_summaries): Likewise.
831 (remap_edge_change_prob): Likewise.
832 (remap_edge_summaries): Likewise.
833 (ipa_merge_fn_summary_after_inlining): Likewise.
834 (ipa_update_overall_fn_summary): Likewise.
835 (read_ipa_call_summary): Likewise.
836 (inline_read_section): Likewise.
837 (write_ipa_call_summary): Likewise.
838 (ipa_fn_summary_write): Likewise.
839 (ipa_free_fn_summary): Likewise.
840 * ipa-hsa.c (process_hsa_functions): Likewise.
841 (ipa_hsa_write_summary): Likewise.
842 (ipa_hsa_read_section): Likewise.
843 * ipa-icf.c (sem_function::merge): Likewise.
844 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
845 (do_estimate_edge_time): Likewise.
846 (estimate_size_after_inlining): Likewise.
847 (estimate_growth): Likewise.
848 (growth_likely_positive): Likewise.
849 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
850 (inline_call): Likewise.
851 * ipa-inline.c (caller_growth_limits): Likewise.
852 (can_inline_edge_p): Likewise.
853 (can_inline_edge_by_limits_p): Likewise.
854 (compute_uninlined_call_time): Likewise.
855 (compute_inlined_call_time): Likewise.
856 (want_inline_small_function_p): Likewise.
857 (edge_badness): Likewise.
858 (update_caller_keys): Likewise.
859 (update_callee_keys): Likewise.
860 (recursive_inlining): Likewise.
861 (inline_small_functions): Likewise.
862 (inline_to_all_callers_1): Likewise.
863 (dump_overall_stats): Likewise.
864 (early_inline_small_functions): Likewise.
865 (early_inliner): Likewise.
866 * ipa-inline.h (estimate_edge_growth): Likewise.
867 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
868 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
869 * ipa-prop.h (IPA_NODE_REF): Likewise.
870 (IPA_EDGE_REF): Likewise.
871 * ipa-pure-const.c (malloc_candidate_p): Likewise.
872 (propagate_malloc): Likewise.
873 * ipa-split.c (execute_split_functions): Likewise.
874 * symbol-summary.h: Rename get to get_create.
876 (get_create): Likewise.
877 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
879 2018-06-08 Martin Liska <mliska@suse.cz>
881 * symbol-summary.h (release): Move definition out of class
883 (symtab_removal): Likewise.
884 (symtab_duplication): Likewise.
886 2018-06-08 Martin Liska <mliska@suse.cz>
888 * symbol-summary.h (function_summary): Move constructor
889 implementation out of class declaration.
891 (symtab_insertion): Likewise.
892 (symtab_removal): Likewise.
893 (symtab_duplication): Likewise.
896 2018-06-08 Martin Liska <mliska@suse.cz>
898 * Makefile.in: Remove support for MPX (macros, related functions,
899 fields in cgraph_node, ...).
900 * builtin-types.def (BT_BND): Likewise.
901 (BT_FN_BND_CONST_PTR): Likewise.
902 (BT_FN_CONST_PTR_BND): Likewise.
903 (BT_FN_VOID_PTR_BND): Likewise.
904 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
905 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
906 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
907 (expand_builtin_mempcpy_with_bounds): Likewise.
908 (expand_builtin_memset_with_bounds): Likewise.
909 (expand_builtin_memset_args): Likewise.
910 (std_expand_builtin_va_start): Likewise.
911 (expand_builtin): Likewise.
912 (expand_builtin_with_bounds): Likewise.
913 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
914 (DEF_LIB_BUILTIN_CHKP): Likewise.
915 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
916 (DEF_CHKP_BUILTIN): Likewise.
917 (BUILT_IN_MEMCPY): Likewise.
918 (BUILT_IN_MEMMOVE): Likewise.
919 (BUILT_IN_MEMPCPY): Likewise.
920 (BUILT_IN_MEMSET): Likewise.
921 (BUILT_IN_STPCPY): Likewise.
922 (BUILT_IN_STRCAT): Likewise.
923 (BUILT_IN_STRCHR): Likewise.
924 (BUILT_IN_STRCPY): Likewise.
925 (BUILT_IN_STRLEN): Likewise.
926 (BUILT_IN_MEMCPY_CHK): Likewise.
927 (BUILT_IN_MEMMOVE_CHK): Likewise.
928 (BUILT_IN_MEMPCPY_CHK): Likewise.
929 (BUILT_IN_MEMSET_CHK): Likewise.
930 (BUILT_IN_STPCPY_CHK): Likewise.
931 (BUILT_IN_STRCAT_CHK): Likewise.
932 (BUILT_IN_STRCPY_CHK): Likewise.
933 * calls.c (store_bounds): Likewise.
934 (emit_call_1): Likewise.
935 (special_function_p): Likewise.
936 (maybe_warn_nonstring_arg): Likewise.
937 (initialize_argument_information): Likewise.
938 (finalize_must_preallocate): Likewise.
939 (compute_argument_addresses): Likewise.
940 (expand_call): Likewise.
941 * cfgexpand.c (expand_call_stmt): Likewise.
942 (expand_return): Likewise.
943 (expand_gimple_stmt_1): Likewise.
944 (pass_expand::execute): Likewise.
945 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
946 (cgraph_node::remove): Likewise.
947 (cgraph_node::dump): Likewise.
948 (cgraph_node::verify_node): Likewise.
949 * cgraph.h (chkp_function_instrumented_p): Likewise.
950 (symtab_node::get_alias_target): Likewise.
951 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
952 (cgraph_local_p): Likewise.
953 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
954 (cgraph_edge::rebuild_references): Likewise.
955 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
956 (walk_polymorphic_call_targets): Likewise.
957 (cgraph_node::expand_thunk): Likewise.
958 (symbol_table::output_weakrefs): Likewise.
959 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
960 (ix86_handle_option): Likewise.
961 * config/i386/constraints.md: Likewise.
962 * config/i386/i386-builtin-types.def (BND): Likewise.
966 * config/i386/i386-builtin.def (BDESC_END): Likewise.
967 (BDESC_FIRST): Likewise.
969 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
970 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
971 * config/i386/i386.c (enum reg_class): Likewise.
972 (ix86_target_string): Likewise.
973 (ix86_option_override_internal): Likewise.
974 (ix86_conditional_register_usage): Likewise.
975 (ix86_valid_target_attribute_inner_p): Likewise.
976 (ix86_set_indirect_branch_type): Likewise.
977 (ix86_set_current_function): Likewise.
978 (ix86_function_arg_regno_p): Likewise.
979 (init_cumulative_args): Likewise.
980 (ix86_function_arg_advance): Likewise.
981 (ix86_function_arg): Likewise.
982 (ix86_pass_by_reference): Likewise.
983 (ix86_function_value_regno_p): Likewise.
984 (ix86_function_value_1): Likewise.
985 (ix86_function_value_bounds): Likewise.
986 (ix86_return_in_memory): Likewise.
987 (ix86_setup_incoming_vararg_bounds): Likewise.
988 (ix86_va_start): Likewise.
989 (indirect_thunk_need_prefix): Likewise.
990 (print_reg): Likewise.
991 (ix86_print_operand): Likewise.
992 (ix86_expand_call): Likewise.
993 (ix86_output_function_return): Likewise.
994 (reg_encoded_number): Likewise.
995 (BDESC_VERIFYS): Likewise.
996 (ix86_init_mpx_builtins): Likewise.
997 (ix86_init_builtins): Likewise.
998 (ix86_emit_cmove): Likewise.
999 (ix86_emit_move_max): Likewise.
1000 (ix86_expand_builtin): Likewise.
1001 (ix86_builtin_mpx_function): Likewise.
1002 (ix86_get_arg_address_for_bt): Likewise.
1003 (ix86_load_bounds): Likewise.
1004 (ix86_store_bounds): Likewise.
1005 (ix86_load_returned_bounds): Likewise.
1006 (ix86_store_returned_bounds): Likewise.
1007 (ix86_class_likely_spilled_p): Likewise.
1008 (ix86_hard_regno_mode_ok): Likewise.
1009 (x86_order_regs_for_local_alloc): Likewise.
1010 (ix86_mitigate_rop): Likewise.
1011 (ix86_bnd_prefixed_insn_p): Likewise.
1012 (ix86_mpx_bound_mode): Likewise.
1013 (ix86_make_bounds_constant): Likewise.
1014 (ix86_initialize_bounds): Likewise.
1015 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1016 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1017 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1018 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1019 (TARGET_CHKP_BOUND_MODE): Likewise.
1020 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1021 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1022 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1023 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1024 * config/i386/i386.h (TARGET_MPX): Likewise.
1025 (TARGET_MPX_P): Likewise.
1026 (VALID_BND_REG_MODE): Likewise.
1027 (FIRST_BND_REG): Likewise.
1028 (LAST_BND_REG): Likewise.
1029 (enum reg_class): Likewise.
1030 (BND_REG_P): Likewise.
1031 (BND_REGNO_P): Likewise.
1032 (BNDmode): Likewise.
1033 (ADJUST_INSN_LENGTH): Likewise.
1034 * config/i386/i386.md: Likewise.
1035 * config/i386/i386.opt: Likewise.
1036 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1037 (defined): Likewise.
1038 (LINK_MPX): Likewise.
1039 (MPX_SPEC): Likewise.
1040 (LIBMPX_SPEC): Likewise.
1041 (LIBMPXWRAPPERS_SPEC): Likewise.
1042 (CHKP_SPEC): Likewise.
1043 * config/i386/predicates.md: Likewise.
1044 * dbxout.c (dbxout_type): Likewise.
1045 * doc/extend.texi: Likewise.
1046 * doc/invoke.texi: Likewise.
1047 * doc/md.texi: Likewise.
1048 * doc/tm.texi: Likewise.
1049 * doc/tm.texi.in: Likewise.
1050 * dwarf2out.c (is_base_type): Likewise.
1051 (gen_formal_types_die): Likewise.
1052 (gen_subprogram_die): Likewise.
1053 (gen_type_die_with_usage): Likewise.
1054 (gen_decl_die): Likewise.
1055 (dwarf2out_late_global_decl): Likewise.
1056 * expr.c (expand_assignment): Likewise.
1057 (emit_storent_insn): Likewise.
1058 (store_expr_with_bounds): Likewise.
1059 (store_expr): Likewise.
1060 (expand_expr_real_1): Likewise.
1061 * expr.h (store_expr_with_bounds): Likewise.
1062 * function.c (use_register_for_decl): Likewise.
1063 (struct bounds_parm_data): Likewise.
1064 (assign_parms_augmented_arg_list): Likewise.
1065 (assign_parm_find_entry_rtl): Likewise.
1066 (assign_parm_is_stack_parm): Likewise.
1067 (assign_parm_load_bounds): Likewise.
1068 (assign_bounds): Likewise.
1069 (assign_parms): Likewise.
1070 (expand_function_start): Likewise.
1071 * gcc.c (CHKP_SPEC): Likewise.
1072 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1073 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1074 (wrestrict_dom_walker::check_call): Likewise.
1075 * gimple.c (gimple_build_call_from_tree): Likewise.
1076 * gimple.h (enum gf_mask): Likewise.
1077 (gimple_call_with_bounds_p): Likewise.
1078 (gimple_call_set_with_bounds): Likewise.
1079 * gimplify.c (gimplify_init_constructor): Likewise.
1080 * ipa-cp.c (initialize_node_lattices): Likewise.
1081 (propagate_constants_across_call): Likewise.
1082 (find_more_scalar_values_for_callers_subset): Likewise.
1083 * ipa-hsa.c (process_hsa_functions): Likewise.
1084 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1085 * ipa-icf.c (sem_function::merge): Likewise.
1086 * ipa-inline.c (early_inliner): Likewise.
1087 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1088 (warn_function_cold): Likewise.
1089 (propagate_pure_const): Likewise.
1090 * ipa-ref.h (enum GTY): Likewise.
1091 * ipa-split.c (find_retbnd): Likewise.
1092 (consider_split): Likewise.
1093 (split_function): Likewise.
1094 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1095 * ipa.c (walk_polymorphic_call_targets): Likewise.
1096 (symbol_table::remove_unreachable_nodes): Likewise.
1097 (process_references): Likewise.
1098 (cgraph_build_static_cdtor_1): Likewise.
1099 * lto-cgraph.c (lto_output_node): Likewise.
1100 (output_refs): Likewise.
1101 (compute_ltrans_boundary): Likewise.
1102 (input_overwrite_node): Likewise.
1103 (input_node): Likewise.
1104 (input_cgraph_1): Likewise.
1105 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1106 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1107 (class pass_chkp_instrumentation_passes): Likewise.
1108 (make_pass_chkp_instrumentation_passes): Likewise.
1109 * passes.def: Likewise.
1110 * rtl.h (struct GTY): Likewise.
1111 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1112 * stor-layout.c (layout_type): Likewise.
1113 * symtab.c: Likewise.
1114 * target.def: Likewise.
1115 * targhooks.c (default_chkp_bound_type): Likewise.
1116 (default_chkp_bound_mode): Likewise.
1117 (default_builtin_chkp_function): Likewise.
1118 (default_chkp_function_value_bounds): Likewise.
1119 (default_chkp_make_bounds_constant): Likewise.
1120 (default_chkp_initialize_bounds): Likewise.
1121 * targhooks.h (default_chkp_bound_type): Likewise.
1122 (default_chkp_bound_mode): Likewise.
1123 (default_builtin_chkp_function): Likewise.
1124 (default_chkp_function_value_bounds): Likewise.
1125 (default_chkp_make_bounds_constant): Likewise.
1126 (default_chkp_initialize_bounds): Likewise.
1127 * toplev.c (compile_file): Likewise.
1128 (process_options): Likewise.
1129 * tree-core.h (DEF_BUILTIN): Likewise.
1130 (DEF_BUILTIN_CHKP): Likewise.
1131 * tree-inline.c (declare_return_variable): Likewise.
1132 (remap_gimple_stmt): Likewise.
1133 (copy_bb): Likewise.
1134 (initialize_inlined_parameters): Likewise.
1135 (expand_call_inline): Likewise.
1136 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1137 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1138 (make_pass_ipa_chkp_produce_thunks): Likewise.
1139 (make_pass_chkp): Likewise.
1140 (make_pass_chkp_opt): Likewise.
1141 (make_pass_chkp_instrumentation_passes): Likewise.
1142 * tree-pretty-print.c (dump_generic_node): Likewise.
1143 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1144 * tree-ssa-dce.c (propagate_necessity): Likewise.
1145 (eliminate_unnecessary_stmts): Likewise.
1146 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1147 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1148 * tree-ssa-sccvn.h: Likewise.
1149 * tree-ssa-strlen.c (get_string_length): Likewise.
1150 (valid_builtin_call): Likewise.
1151 (adjust_last_stmt): Likewise.
1152 (handle_builtin_strchr): Likewise.
1153 (handle_builtin_strcpy): Likewise.
1154 (handle_builtin_stxncpy): Likewise.
1155 (handle_builtin_memcpy): Likewise.
1156 (handle_builtin_strcat): Likewise.
1157 (strlen_check_and_optimize_stmt): Likewise.
1158 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1159 * tree-streamer-in.c: Likewise.
1160 * tree-streamer.c (record_common_node): Likewise.
1161 * tree.c (tree_code_size): Likewise.
1162 (wide_int_to_tree_1): Likewise.
1163 (type_contains_placeholder_1): Likewise.
1164 (build_common_tree_nodes): Likewise.
1165 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1166 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1167 (POINTER_BOUNDS_P): Likewise.
1168 (BOUNDED_TYPE_P): Likewise.
1169 (BOUNDED_P): Likewise.
1170 (CALL_WITH_BOUNDS_P): Likewise.
1171 (pointer_bounds_type_node): Likewise.
1172 * value-prof.c (gimple_ic): Likewise.
1173 * var-tracking.c (vt_add_function_parameters): Likewise.
1174 * varasm.c (make_decl_rtl): Likewise.
1175 (assemble_start_function): Likewise.
1176 (output_constant): Likewise.
1177 (maybe_assemble_visibility): Likewise.
1178 * varpool.c (ctor_for_folding): Likewise.
1179 * chkp-builtins.def: Remove.
1180 * ipa-chkp.c: Remove.
1181 * ipa-chkp.h: Remove.
1182 * rtl-chkp.c: Remove.
1183 * rtl-chkp.h: Remove.
1184 * tree-chkp-opt.c: Remove.
1185 * tree-chkp.c: Remove.
1186 * tree-chkp.h: Remove.
1188 2018-06-07 Carl Love <cel@us.ibm.com>
1190 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1191 vextract_fp_from_shortl): Add BE support.
1193 2018-06-07 Paul Koning <ni1d@arrl.net>
1195 * compare-elim.c (try_merge_compare): Don't merge compare if
1196 address contains a side effect.
1197 (try_eliminate_compare): Likewise.
1199 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1201 * config.gcc: Support "tremont".
1202 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1203 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1205 * config/i386/i386.c (m_TREMONT): Define.
1206 (processor_target_table): Add "tremont".
1207 (PTA_TREMONT): Define.
1208 (ix86_lea_outperforms): Add TARGET_TREMONT.
1209 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1210 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1211 and M_INTEL_GOLDMONT_PLUS.
1212 (fold_builtin_cpu): Add "tremont".
1213 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1214 (ix86_option_override_internal): Add "tremont".
1215 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1216 (processor_type): Add PROCESSOR_TREMONT.
1217 * config/i386/x86-tune.def: Add m_TREMONT.
1218 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1220 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1222 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1223 symbol defined for msp430i* devices to be lower case.
1225 gcc/testsuite/gcc.target/msp430/
1226 * msp430i-device-symbol.c: New test.
1227 * msp430f-device-symbol.c: New test.
1228 * msp430.h: New test header file.
1230 2018-06-07 Richard Biener <rguenther@suse.de>
1232 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1233 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1234 Properly wrap signed arithmetic if overflow wraps.
1236 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1238 PR tree-optimization/69615
1239 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1240 of a cast from a same precision integral SSA_NAME in a bb dominated
1241 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1242 cast to utype if rhs2 has already a compatible type.
1244 2018-06-07 Richard Biener <rguenther@suse.de>
1246 PR tree-optimization/85935
1247 * graphite-scop-detection.c (find_params_in_bb): Analyze
1248 condition operands with respect to the correct loop. Assert
1249 the analysis doesn't fail.
1251 2018-06-04 Carl Love <cel@us.ibm.com>
1253 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1254 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1255 as it is slightly cheaper.
1256 (first_match_or_eos_index_<mode>):
1257 Calculate index using natural element order.
1258 (first_match_index_<mode>):
1259 Calculate index using natural element order.
1260 (first_match_or_eos_index_<mode>):
1261 Calculate index using natural order.
1262 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1263 for BE and LE modes.
1264 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1265 P9V_BUILTIN_VCLZLSBB_V16QI.
1266 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1269 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1271 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1272 indentation and line wrap for many prototypes. Add missing
1273 @smallexample directives around block of prototypes for vec_xl and
1276 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1278 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1279 track if we pass or return IEEE 128-bit floating point.
1280 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1281 C++ mangling that is compatible with GCC 8.1.
1282 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1283 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1284 floating point types.
1285 (rs6000_function_arg_advance_1): Likewise.
1286 (rs6000_mangle_type): Optionally generate mangled names that match
1287 what GCC 8.1 generated for IEEE 128-bit floating point types.
1288 (rs6000_globalize_decl_name): If we have an external function that
1289 passes or returns IEEE 128-bit types, generate a weak reference
1290 from the mangled name used in GCC 8.1 to the current mangled
1292 (rs6000_init_builtins): Make __ibm128 use the long double type if
1293 long double is IBM extended double. Make __float128 use the long
1294 double type if long double is IEEE 128-bit.
1297 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1298 macro for __ibm128 built-in functions.
1299 (PACK_IF): Add __ibm128 pack/unpack functions.
1300 (UNPACK_IF): Likewise.
1301 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1302 enable long double built-in functions if long double is IEEE
1303 128-bit floating point.
1304 (rs6000_invalid_builtin): Update long double built-in function
1306 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1307 functions, adjust the built-in function to use the long double
1308 built-in function if __ibm128 and long double are the same type.
1309 * doc/extend.texi (PowerPC builtins): Update documention for
1310 __builtin_{,un}pack_longdouble. Add documentation for
1311 __builtin_{,un}pack_ibm128.
1313 2018-06-06 Jim Wilson <jimw@sifive.com>
1315 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1316 (struct machine_function): New field interrupt_mode.
1317 (riscv_handle_type_attribute): New function. Add forward declaration.
1318 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1319 (riscv_expand_epilogue): Check interrupt_mode field.
1320 (riscv_set_current_function): Check interrupt attribute args and
1321 set interrupt_mode field.
1322 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1323 (riscv_sret, riscv_uret): New.
1324 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1325 new arguments to interrupt attribute.
1327 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1330 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1331 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1333 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1335 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1336 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1337 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1338 several redundant entries.
1340 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1342 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1343 type from "rtx" to "rtx_insn *".
1344 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1345 for local "call_insn", removing cast.
1346 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1348 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1350 PR tree-optimization/86066
1351 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1352 for BIT_INSERT_EXPR stores.
1354 2018-06-06 Richard Biener <rguenther@suse.de>
1356 PR tree-optimization/86062
1357 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1358 component refs ontop
1359 of to be offsetted base.
1361 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1363 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1364 to be static and remove check on interrupt attribute name.
1366 gcc/testsuite/gcc.target/msp430/
1367 * function-attributes-4.c: New test.
1368 * static-interrupts.c: New test.
1370 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1372 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1373 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1375 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1378 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1380 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1381 Remove second argument, change how error is called.
1382 (aarch64_layout_arg): Remove second argument from
1383 aarch64_err_no_fpadvsimd call.
1384 (aarch64_init_cumulative_args): Ditto.
1385 (aarch64_gimplify_va_arg_expr): Ditto.
1386 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1388 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1390 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1391 (*simple_return_indirect_internal<mode>): Rename from
1392 simple_return_indirect_internal. Use W mode iterator.
1393 (rstorssp): New expander.
1394 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1395 (clrssbsy): New expander.
1396 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1398 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1400 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1402 (cmse_check_pointed_object): Likewise.
1404 2018-06-05 Martin Liska <mliska@suse.cz>
1406 PR gcov-profile/47618
1407 * doc/invoke.texi: Document how -fprofile-dir format
1410 2018-06-05 Richard Biener <rguenther@suse.de>
1412 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1413 removal pretend DOM info isn't available so we do not update
1414 it and only remove edges, not dominated blocks. Actually free
1415 DOM info in case we removed something. Remove unreachable blocks.
1416 (mfb_keep_latches): Work with either DOM info or marked backedges.
1417 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1418 first. Mark backedges if DOM info isn't available.
1419 (Re-)compute DOM info after cleanup_control_flow_pre.
1421 2018-06-05 Richard Biener <rguenther@suse.de>
1423 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1424 (locus_discrim_hasher::hash): Adjust.
1425 (locus_discrim_hasher::equal): Likewise.
1426 (next_discriminator_for_locus): Work on line directly.
1427 (same_line_p): Pass in expanded locus1 as well.
1428 (assign_discriminators): Avoid redundant location expansions.
1430 2018-06-05 Richard Biener <rguenther@suse.de>
1432 PR tree-optimization/86046
1433 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1434 if required after clearing TREE_ADDRESSABLE.
1436 2018-06-05 Richard Biener <rguenther@suse.de>
1438 PR tree-optimization/86047
1439 * tree-ssa-loop.c (for_each_index): Glob handling of all
1440 decls and constants and really handle all of them.
1442 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1445 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1446 qualifier_void_pointer and qualifier_const_void_pointer.
1447 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1448 (arm_init_builtins): Handle the above.
1449 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1450 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1451 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1454 2018-06-05 Martin Liska <mliska@suse.cz>
1456 * auto-profile.c (read_autofdo_file): Do not use
1457 gcov_ctr_summary struct.
1458 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1459 * coverage.c (struct counts_entry): Likewise.
1460 (read_counts_file): Read just single summary entry.
1461 (get_coverage_counts): Use gcov_summary struct.
1462 * coverage.h (get_coverage_counts): Likewise.
1463 * gcov-dump.c (dump_working_sets): Likewise.
1464 (tag_summary): Dump just single summary.
1465 * gcov-io.c (gcov_write_summary): Write just histogram
1467 (gcov_read_summary): Read just single summary.
1468 (compute_working_sets): Use gcov_summary struct.
1469 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1470 of GCOV_COUNTERS_SUMMABLE.
1471 (GCOV_COUNTERS_SUMMABLE): Remove.
1472 (GCOV_FIRST_VALUE_COUNTER): Replace with
1473 GCOV_COUNTER_V_INTERVAL.
1474 (struct gcov_ctr_summary): Remove.
1475 (struct gcov_summary): Directly use fields of former
1477 (compute_working_sets): Use gcov_summary struct.
1478 * gcov.c (read_count_file): Do not use ctrs fields.
1479 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1481 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1483 * profile.c: Likewise.
1484 * profile.h: Likewise.
1486 2018-06-05 Martin Liska <mliska@suse.cz>
1488 PR gcov-profile/84846
1489 * gcov.c (output_lines): Print working directory only
1490 in intermediate format.
1492 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1494 * config/s390/s390-builtin-types.def: Add void function type.
1495 * config/s390/s390-builtins.def: Use the function type for the
1498 2018-06-04 Jim Wilson <jimw@sifive.com>
1500 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1502 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1503 and maybe_eh_return. Change regno to unsigned int. Use new args to
1504 handle EH_RETURN_DATA_REGNO registers properly.
1505 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1506 (riscv_expand_epilogue): Update comment. Change argument name and
1507 type. Update code to use new name and type. Pass new args to
1508 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1510 * config/riscv/riscv.md (NORMAL_RETURN): New.
1511 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1512 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1513 (eh_return): Call gen_eh_return_internal and emit barrier.
1514 (eh_return_internal): Call riscv_expand_epilogue.
1516 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1518 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1519 bit_insertion field and declare can_be_merged_into method.
1520 (merged_store_group::can_be_merged_into): New method.
1521 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1522 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1523 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1525 2018-06-04 Richard Biener <rguenther@suse.de>
1527 PR tree-optimization/85955
1528 * builtins.c (fold_builtin_sincos): Convert pointers to
1529 destination to appropriate type before dereferencing.
1531 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1533 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1535 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1537 * expr.c (expand_expr_real_1): Force the operand into memory if
1538 its TYPE_MODE is BLKmode and if there is no integer mode for
1539 the number of bits being extracted.
1541 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1545 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1546 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1548 2018-06-04 Richard Biener <rguenther@suse.de>
1550 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1551 (cleanup_tree_cfg_noloop): ... single caller. Do
1552 start_recording_case_labels later.
1554 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1556 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1557 to _IMMINTRIN_H_INCLUDED.
1558 * config/i386/pconfigintrin.h: Ditto.
1559 * config/i386/waitpkgintrin.h: Ditto.
1560 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1561 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1562 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1563 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1564 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1565 waitpkgintrin.h and cldemoteintrin.h.
1567 2018-06-04 Richard Biener <rguenther@suse.de>
1569 PR tree-optimization/86038
1570 * tracer.c (find_best_successor): Check probability for
1571 being initialized, bail out if not.
1573 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1576 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1577 of bits to ignore when comparing architectures.
1579 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1581 PR tree-optimization/69615
1582 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1583 maximum or minimum of the type, try to merge it also as if
1584 range1 is + [-, x - 1] or + [x + 1, -].
1587 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1589 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1591 PR tree-optimization/86034
1592 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1593 the unsigned bitfield type in a bit insertion sequence if it does not
1594 have a larger precision than the bitfield size.
1595 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1597 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1599 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1601 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1603 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1604 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1605 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1606 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1608 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1610 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1611 Disable -fdelete-null-pointer-checks for ELF toolchain.
1613 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1614 Kito Cheng <kito.cheng@gmail.com>
1616 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1617 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1618 (nds32*-*-*): Add glibc and uclibc conditions.
1619 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1620 (TARGET_EXCEPT_UNWIND_INFO): Define.
1621 * config/nds32/elf.h: New file.
1622 * config/nds32/linux.h: New file.
1623 * config/nds32/nds32-elf.opt: New file.
1624 * config/nds32/nds32-linux.opt: New file.
1625 * config/nds32/nds32-fp-as-gp.c
1626 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1627 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1629 (nds32_asm_file_end): Ditto.
1630 (nds32_print_operand): Ditto.
1631 (nds32_insert_attributes): Ditto.
1632 (nds32_init_libfuncs): New function.
1633 (TARGET_HAVE_TLS): Define.
1634 (TARGET_INIT_LIBFUNCS): Define.
1635 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1637 (TARGET_ELF): Apply different mcmodel setting.
1638 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1639 been migrated into elf.h and linux.h files.
1640 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1641 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1642 (mcmodel): The content has been migrated into nds32-elf.opt and
1643 nds32-linux.opt files.
1644 * config/nds32/t-elf: New file.
1645 * config/nds32/t-linux: New file.
1647 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1648 Shiva Chen <shiva0217@gmail.com>
1650 * config/nds32/constants.md (unspec_volatile_element): Add
1651 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1652 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1654 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1655 (make_pass_nds32_fp_as_gp): Declare.
1656 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1658 (nds32_asm_function_end_prologue): Remove unused asm output.
1659 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1660 (nds32_asm_file_start): Output necessary fp_as_gp information.
1661 (nds32_option_override): Adjust register usage.
1662 (nds32_expand_prologue): Consider fp_as_gp situation.
1663 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1664 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1667 (simple_return): Ditto.
1668 (omit_fp_begin): Output special directive for fp_as_gp.
1669 (omit_fp_end): Output special directive for fp_as_gp.
1670 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1671 mforbid-fp-as-gp): New options.
1673 2018-06-01 Mark Wielaard <mark@klomp.org>
1675 * dwarf2out.c (dwarf2out_finish): Remove generation of
1676 DW_AT_loclists_base.
1678 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1680 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1681 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1682 (struct merged_store_group): Add bit_insertion field.
1683 (dump_char_array): Use standard hexadecimal format.
1684 (merged_store_group::merged_store_group): Set bit_insertion to false.
1685 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1686 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1687 also print the mask in the dump file.
1688 (pass_store_merging::gate): Minor tweak.
1689 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1690 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1691 stores with INTEGER_CST stores.
1692 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1693 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1694 and use it throughout. Generate bit insertion sequences if need be.
1695 (pass_store_merging::process_store): Remove redundant condition.
1696 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1698 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1700 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1701 the 128-bit floating point types. Fix function comment.
1703 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1705 * config/aarch64/aarch64-simd.md
1706 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1708 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1711 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1713 PR tree-optimization/85989
1714 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1716 (backprop::intersect_uses): Check it when deciding whether this
1717 is a backedge reference.
1718 (backprop::process_block): Add each phi to m_visited_phis
1719 after visiting it, then clear it at the end.
1721 2018-06-01 Richard Biener <rguenther@suse.de>
1723 * tree-vectorizer.h (vect_dr_stmt): New function.
1724 (vect_get_load_cost): Adjust.
1725 (vect_get_store_cost): Likewise.
1726 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1727 Use vect_dr_stmt instead of DR_SMTT.
1728 (vect_record_base_alignments): Likewise.
1729 (vect_calculate_target_alignment): Likewise.
1730 (vect_compute_data_ref_alignment): Likewise and make static.
1731 (vect_update_misalignment_for_peel): Likewise.
1732 (vect_verify_datarefs_alignment): Likewise.
1733 (vector_alignment_reachable_p): Likewise.
1734 (vect_get_data_access_cost): Likewise. Pass down
1735 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1736 (vect_get_peeling_costs_all_drs): Likewise.
1737 (vect_peeling_hash_get_lowest_cost): Likewise.
1738 (vect_enhance_data_refs_alignment): Likewise.
1739 (vect_find_same_alignment_drs): Likewise.
1740 (vect_analyze_data_refs_alignment): Likewise.
1741 (vect_analyze_group_access_1): Likewise.
1742 (vect_analyze_group_access): Likewise.
1743 (vect_analyze_data_ref_access): Likewise.
1744 (vect_analyze_data_ref_accesses): Likewise.
1745 (vect_vfa_segment_size): Likewise.
1746 (vect_small_gap_p): Likewise.
1747 (vectorizable_with_step_bound_p): Likewise.
1748 (vect_prune_runtime_alias_test_list): Likewise.
1749 (vect_analyze_data_refs): Likewise.
1750 (vect_supportable_dr_alignment): Likewise.
1751 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1752 (vect_gen_prolog_loop_niters): Likewise.
1753 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1754 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1756 (vect_recog_mask_conversion_pattern): Likewise.
1757 (vect_try_gather_scatter_pattern): Likewise.
1758 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1759 to vect_get_store_cost.
1760 (vect_get_store_cost): Get stmt_info instead of DR.
1761 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1762 (vect_get_load_cost): Get stmt_info instead of DR.
1764 2018-06-01 Richard Biener <rguenther@suse.de>
1767 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1768 vars wrapped in MEM_REFs.
1770 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1772 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1773 Fix subreg tests so that we only return a choice between
1774 GENERAL_REGS and FP_REGS if the original classes included both.
1776 2018-06-01 Richard Biener <rguenther@suse.de>
1779 * tree-ssa-structalias.c (get_function_part_constraint):
1780 Handle NULL fi->decl.
1781 (find_func_aliases_for_call): Properly handle indirect
1782 fi from direct call.
1783 (find_func_clobbers): Likewise.
1784 (ipa_pta_execute): Likewise.
1785 (create_variable_info_for): For functions that are ifunc_resolver
1786 resolve to a varinfo that contains the result of the resolver call.
1787 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1790 2018-05-31 Michael Collison <michael.collison@arm.com>
1792 * config/aarch64/aarch64.md:
1793 (*fix_to_zero_extenddfdi2): New pattern.
1794 * gcc.target/aarch64/fix_extend1.c: New testcase.
1796 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1800 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1801 and BUILT_IN_STRNCMP_EQ.
1802 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1803 BUILT_IN_STRNCMP_EQ.
1804 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1805 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1806 (gimple_fold_builtin): Likewise.
1807 * tree-ssa-strlen.c (compute_string_length): New function.
1808 (determine_min_obsize): New function.
1809 (handle_builtin_string_cmp): New function to handle calls to
1810 string compare functions.
1811 (strlen_optimize_stmt): Add handling to builtin string compare
1813 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1814 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1815 * tree.c (build_common_builtin_nodes): Add new defines of
1816 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1818 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1821 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1822 functions with naked attribute.
1824 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1826 * config/i386/sse.md (avx_vec_concat<mode>):
1827 Substitute concat_tg_mode mode attribute with xtg_mode.
1828 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1829 (concat_tg_mode): Remove mode attribute.
1831 2018-05-31 Martin Sebor <msebor@redhat.com>
1834 * calls.c (alloc_max_size): Correct a logic error/typo.
1835 Treat excessive arguments as infinite. Warn for invalid arguments.
1836 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1838 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1841 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1842 and movx for Haswell.
1844 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1845 Cesar Philippidis <cesar@codesourcery.com>
1848 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1849 when emitting error on private/firstprivate reductions.
1850 * omp-low.c (lower_omp_target): Avoid reference-type processing
1851 on pointers for firstprivate clause.
1853 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1855 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1858 * config/aarch64/aarch64-simd.md
1859 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1860 (aarch64_ld1_x3_<mode>): Likewise
1861 (aarch64_st1x2<VALLDIF:mode>): Likewise
1862 (aarch64_st1_x2_<mode>): Likewise
1863 (aarch64_st1x3<VALLDIF:mode>): Likewise
1864 (aarch64_st1_x3_<mode>): Likewise
1865 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1866 (vld1_s8_x3): Likewise.
1867 (vld1_u16_x3): Likewise.
1868 (vld1_s16_x3): Likewise.
1869 (vld1_u32_x3): Likewise.
1870 (vld1_s32_x3): Likewise.
1871 (vld1_u64_x3): Likewise.
1872 (vld1_s64_x3): Likewise.
1873 (vld1_f16_x3): Likewise.
1874 (vld1_f32_x3): Likewise.
1875 (vld1_f64_x3): Likewise.
1876 (vld1_p8_x3): Likewise.
1877 (vld1_p16_x3): Likewise.
1878 (vld1_p64_x3): Likewise.
1879 (vld1q_u8_x3): Likewise.
1880 (vld1q_s8_x3): Likewise.
1881 (vld1q_u16_x3): Likewise.
1882 (vld1q_s16_x3): Likewise.
1883 (vld1q_u32_x3): Likewise.
1884 (vld1q_s32_x3): Likewise.
1885 (vld1q_u64_x3): Likewise.
1886 (vld1q_s64_x3): Likewise.
1887 (vld1q_f16_x3): Likewise.
1888 (vld1q_f32_x3): Likewise.
1889 (vld1q_f64_x3): Likewise.
1890 (vld1q_p8_x3): Likewise.
1891 (vld1q_p16_x3): Likewise.
1892 (vld1q_p64_x3): Likewise.
1893 (vst1_s64_x2): Likewise.
1894 (vst1_u64_x2): Likewise.
1895 (vst1_f64_x2): Likewise.
1896 (vst1_s8_x2): Likewise.
1897 (vst1_p8_x2): Likewise.
1898 (vst1_s16_x2): Likewise.
1899 (vst1_p16_x2): Likewise.
1900 (vst1_s32_x2): Likewise.
1901 (vst1_u8_x2): Likewise.
1902 (vst1_u16_x2): Likewise.
1903 (vst1_u32_x2): Likewise.
1904 (vst1_f16_x2): Likewise.
1905 (vst1_f32_x2): Likewise.
1906 (vst1_p64_x2): Likewise.
1907 (vst1q_s8_x2): Likewise.
1908 (vst1q_p8_x2): Likewise.
1909 (vst1q_s16_x2): Likewise.
1910 (vst1q_p16_x2): Likewise.
1911 (vst1q_s32_x2): Likewise.
1912 (vst1q_s64_x2): Likewise.
1913 (vst1q_u8_x2): Likewise.
1914 (vst1q_u16_x2): Likewise.
1915 (vst1q_u32_x2): Likewise.
1916 (vst1q_u64_x2): Likewise.
1917 (vst1q_f16_x2): Likewise.
1918 (vst1q_f32_x2): Likewise.
1919 (vst1q_f64_x2): Likewise.
1920 (vst1q_p64_x2): Likewise.
1921 (vst1_s64_x3): Likewise.
1922 (vst1_u64_x3): Likewise.
1923 (vst1_f64_x3): Likewise.
1924 (vst1_s8_x3): Likewise.
1925 (vst1_p8_x3): Likewise.
1926 (vst1_s16_x3): Likewise.
1927 (vst1_p16_x3): Likewise.
1928 (vst1_s32_x3): Likewise.
1929 (vst1_u8_x3): Likewise.
1930 (vst1_u16_x3): Likewise.
1931 (vst1_u32_x3): Likewise.
1932 (vst1_f16_x3): Likewise.
1933 (vst1_f32_x3): Likewise.
1934 (vst1_p64_x3): Likewise.
1935 (vst1q_s8_x3): Likewise.
1936 (vst1q_p8_x3): Likewise.
1937 (vst1q_s16_x3): Likewise.
1938 (vst1q_p16_x3): Likewise.
1939 (vst1q_s32_x3): Likewise.
1940 (vst1q_s64_x3): Likewise.
1941 (vst1q_u8_x3): Likewise.
1942 (vst1q_u16_x3): Likewise.
1943 (vst1q_u32_x3): Likewise.
1944 (vst1q_u64_x3): Likewise.
1945 (vst1q_f16_x3): Likewise.
1946 (vst1q_f32_x3): Likewise.
1947 (vst1q_f64_x3): Likewise.
1948 (vst1q_p64_x3): Likewise.
1950 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1952 * config/msp430/msp430.c (msp430_output_labelref): Prepend
1953 user_label_prefix to name.
1955 * tree-core.h: Update comment about the format of NAME string
1956 passed to handler in attribute_spec.
1958 * config/msp430/msp430.md: Remove erroneous subreg expression from
1959 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
1960 zero_extend{q,h}isi2.
1962 2018-05-30 Borislav Petkov <bp@suse.de>
1964 * doc/extend.texi: Document some architecture specific
1965 constraints and sort entries.
1967 2018-05-30 Martin Sebor <msebor@redhat.com>
1970 * builtins.c (expand_builtin_stpcpy_1): New function.
1971 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
1972 only if the former succeeds.
1974 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1976 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
1979 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1981 * doc/invoke.texi (-flinker-output): Document
1983 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
1985 * passes.c (ipa_write_summaries): Only modify statements if body
1987 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
1988 incrementally linking.
1989 (ipa_passes): Likewise.
1990 * lto-cgraph.c (lto_output_node): When incrementally linking do not
1991 pass down resolution info.
1992 * common.opt (flag_incremental_link): Update info.
1993 * gcc.c (plugin specs): Turn flinker-output=* to
1994 -plugin-opt=-linker-output-known
1995 * toplev.c (compile_file): Also cut compilation when doing incremental
1997 * flag-types. (enum lto_partition_model): Add
1998 LTO_LINKER_OUTPUT_NOLTOREL.
1999 (invoke.texi): Add -flinker-output docs.
2000 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2001 link same way as WPA; do not stream in dead initializers.
2003 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2004 darf2out_register_external_decl): Support incremental link.
2006 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2008 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2010 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2012 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2013 it down to simple_object_copy_lto_debug_sections.
2014 (run_gcc): Determine incremental LTO link time and configure
2015 lto1 into non-wpa mode, disable renaming of debug sections.
2017 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2019 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2020 descriptions of various incorrectly documented functions.
2022 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2025 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2026 address check not strict.
2028 2018-05-30 Richard Biener <rguenther@suse.de>
2030 PR tree-optimization/85964
2031 * tracer.c (better_p): Drop initialized count check, we only
2032 call the function with initialized counts now.
2033 (find_best_successor): Do find a best edge if one
2034 has uninitialized count.
2035 (find_best_predecessor): Likewise. Do BB frequency check only
2036 if count is initialized.
2038 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2040 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2041 (aarch64_ldrstr_offset_compare): New.
2042 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2043 load/store orderings.
2044 (aarch64_gen_adjusted_ldpstp): Likewise.
2046 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2048 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2049 Check for subset of GENERAL_REGS and FP_REGS.
2050 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2053 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2055 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2056 and wi::to_poly_offset. Add the current offset and then check
2057 whether the sum fits, rather than using an unchecked addition of
2058 a checked term. Check for a shwi rather than a uhwi.
2059 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2060 (store_constructor): Use poly_int_tree_p.
2061 (expand_expr_real_1): Likewise.
2062 * function.c (assign_temp): Likewise.
2063 * fold-const.c (const_binop): Use poly_int_tree_p and
2065 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2067 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2068 to_poly_offset for MEM offsets.
2069 * ipa-icf.c (sem_variable::equals): Likewise.
2070 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2071 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2072 wi::to_poly_offset for BIT_FIELD_REF offsets.
2073 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2075 * var-tracking.c (emit_note_insn_var_location): Use
2076 tree_to_poly_uint64.
2078 2018-05-29 Jim Wilson <jimw@sifive.com>
2080 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2082 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2085 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2086 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2088 (sse4_1_round<mode>2): Use nonimmediate_operand
2089 for operand 1 predicate.
2091 2018-05-29 Martin Sebor <msebor@redhat.com>
2092 Richard Biener <rguenther@suse.de>
2095 * calls.c (get_size_range): Call determine_value_range instead
2096 of get_value_range..
2097 * tree-vrp.h (determine_value_range): Declared new function.
2098 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2100 2018-05-29 Richard Biener <rguenther@suse.de>
2102 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2103 sure to use non-pattern stmts for get_earlier_stmt arguments.
2104 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2105 called on pattern stmts.
2106 (get_later_stmt): Likewise.
2108 2018-05-29 Martin Liska <mliska@suse.cz>
2110 PR gcov-profile/85759
2111 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2114 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2116 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2118 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2122 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2123 VEC_PACK_FLOAT_EXPR): New tree codes.
2124 * tree-pretty-print.c (op_code_prio): Handle
2125 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2126 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2127 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2128 * tree-inline.c (estimate_operator_cost): Likewise.
2129 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2130 * fold-const.c (const_binop): Likewise.
2131 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2132 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2133 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2134 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2135 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2136 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2137 * expr.c (expand_expr_real_2): Likewise.
2138 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2139 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2140 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2142 * optabs.c (expand_widen_pattern_expr): For
2143 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2144 sign from result type rather than operand's type.
2145 (expand_binop_directly): For vec_packu_float_optab and
2146 vec_packs_float_optab allow result type to be different from operand's
2148 * optabs-tree.c (optab_for_tree_code): Handle
2149 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2150 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2151 * tree-vect-generic.c (expand_vector_operations_1): Handle
2152 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2153 VEC_PACK_FLOAT_EXPR.
2154 * tree-vect-stmts.c (supportable_widening_operation): Handle
2156 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2157 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2158 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2159 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2160 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2162 (vec_pack<floatprefix>_float_<mode>): New expander.
2163 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2165 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2166 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2167 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2168 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2169 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2171 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2172 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2173 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2174 VEC_PACK_FLOAT_EXPR): Document.
2176 2018-05-29 Richard Biener <rguenther@suse.de>
2178 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2180 (stmt_vec_info_vec): Make pointer.
2181 (init_stmt_vec_info_vec): Remove.
2182 (free_stmt_vec_info_vec): Likewise.
2183 (set_stmt_vec_info_vec): New function.
2184 (free_stmt_vec_infos): Likewise.
2185 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2186 (set_vinfo_for_stmt): Likewise.
2187 (get_earlier_stmt): Likewise.
2188 (get_later_stmt): Likewise.
2189 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2190 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2191 (vec_info::~vec_info): Free stmt_vec_infos.
2192 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2193 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2194 (pass_slp_vectorize::execute): Likewise.
2195 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2196 (free_stmt_vec_info_vec): Likewise.
2197 (set_stmt_vec_info_vec): New function.
2198 (free_stmt_vec_infos): Likewise.
2199 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2200 the global stmt_vec_info_vec.
2201 * tree-parloops.c (gather_scalar_reductions): Use
2202 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2205 2018-05-29 Richard Biener <rguenther@suse.de>
2207 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2209 2018-05-29 Martin Liska <mliska@suse.cz>
2210 David Malcolm <dmalcolm@redhat.com>
2212 * vec.c (test_reverse): New.
2213 (vec_c_tests): Add new test.
2214 * vec.h (vl_ptr>::reverse): New function.
2216 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2218 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2220 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2223 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2225 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2227 2018-05-28 Richard Biener <rguenther@suse.de>
2229 PR tree-optimization/85933
2230 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2231 look at stmts marked as vectorizable.
2233 2018-05-28 Richard Biener <rguenther@suse.de>
2235 PR tree-optimization/85934
2236 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2237 vector boolean check before scalar optimization.
2239 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2241 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2244 2018-05-28 Mark Wielaard <mark@klomp.org>
2246 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2247 if it is an expression containing a minus sign.
2249 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2251 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2253 2018-05-27 Paul Koning <ni1d@arrl.net>
2255 * config/pdp11/pdp11.md (truncsihi2): Remove.
2257 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2258 Chung-Ju Wu <jasonwucj@gmail.com>
2260 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2262 (unaligned_store_dw): Ditto.
2263 * config/nds32/nds32-memory-manipulation.c
2264 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2265 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2266 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2267 (emit_setmem_word_loop): Rename to ...
2268 (emit_setmem_doubleword_loop): ... this.
2269 (nds32_gen_dup_4_byte_to_word_value): New function.
2270 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2271 (nds32_expand_setmem_loop): Refine implementation.
2272 (nds32_expand_setmem_loop_v3m): Ditto.
2273 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2276 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2278 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2280 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2282 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2283 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2284 machine->attr_no_prologue_p.
2285 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2286 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2287 (nds32_expand_epilogue): Consider attr_naked_p.
2288 (nds32_expand_epilogue_v3pop): Likewise.
2289 (nds32_can_use_return_insn): Likewise.
2290 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2291 attr_no_prologue_p fields.
2292 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2294 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2297 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2299 * config/i386/sse.md
2300 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2302 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2304 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2306 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2308 (*<floatsuffix>floatv2div2sf2): Rename to ...
2309 (*float<floatunssuffix>v2div2sf2): ... this.
2310 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2311 (float<floatunssuffix>v2div2sf2_mask): ... this.
2312 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2313 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2314 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2316 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2318 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2320 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2322 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2324 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2326 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2327 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2328 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2329 gen_ufix_truncv8dfv8si2.
2330 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2331 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2332 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2333 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2334 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2335 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2336 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2337 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2339 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2343 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2345 2018-05-25 Jim Wilson <jimw@sifive.com>
2347 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2348 * config/riscv/riscv.c (struct machine_function): Add
2349 interrupt_handler_p and attribute_checked_p fields.
2350 (riscv_attribute_table): Add interrupt.
2351 (riscv_interrupt_type_p): New.
2352 (riscv_save_reg_p): Save extra regs for interrupt handler.
2353 (riscv_use_save_libcall): Return false for interrupt handler.
2354 (riscv_first_stack_step): Add forward declaration.
2355 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2356 for interrupt handler with large frame. Use it for saved reg list.
2357 (riscv_expand_prologue): Move flag_stack_usage_info support to
2358 eliminate duplication.
2359 (riscv_expand_epilogue): Generate mret for interrupt handler.
2360 (riscv_epilogue_uses): New.
2361 (riscv_can_use_return_insn): Return false for interrupt handler.
2362 (riscv_function_ok_for_sibcall): Likewise.
2363 (riscv_set_current_function): Add interrupt handler support.
2364 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2365 * config/riscv/riscv.md (UNSPECV_MRET): New.
2367 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2369 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2371 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2373 PR tree-optimization/85712
2374 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2375 this candidate has already been replaced in-situ by a copy.
2377 2018-05-25 Jason Merrill <jason@redhat.com>
2379 PR c++/80485 - inline function non-zero address.
2380 * symtab.c (nonzero_address): Check DECL_COMDAT.
2382 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2385 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2386 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2387 extension of SImode operation. Use const123_operand predicate.
2388 (*saddsi_1): Remove.
2389 (*saddl_se_1): Ditto.
2391 (*ssubl_se_1): Ditto.
2392 * config/alpha/predicates.md (const123_operand): New predicate.
2393 * config/alpha/constraints.md (P): Use IN_RANGE.
2395 2018-05-25 Richard Biener <rguenther@suse.de>
2397 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2399 (ref_maybe_used_by_stmt_p): Likewise.
2400 (stmt_may_clobber_ref_p): Likewise.
2401 (stmt_may_clobber_ref_p_1): Likewise.
2402 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2404 (ref_maybe_used_by_stmt_p): Likewise.
2405 (stmt_may_clobber_ref_p): Likewise.
2406 (stmt_may_clobber_ref_p_1): Likewise.
2407 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2408 the alias oracle to disambiguate DRs with stmts DR analysis
2410 (vect_analyze_data_refs): Do not give up on not analyzable
2411 DRs for BB vectorization. Remove code truncating the dataref
2414 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2417 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2418 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2419 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2421 2018-05-25 Richard Biener <rguenther@suse.de>
2423 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2424 function, combining stmt data ref gathering and fatal analysis
2426 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2427 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2428 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2429 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2430 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2432 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2434 PR tree-optimization/85720
2435 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2436 SCC if all partitions are builtins.
2437 (version_loop_by_alias_check): New parameter. Generate cancelable
2438 runtime alias check if all partitions are builtins.
2439 (distribute_loop): Update call to above function.
2441 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2443 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2444 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2445 (parm_default_def_partition_arg): Ditto.
2446 (set_parm_default_def_partition): Ditto.
2447 (get_parm_default_def_partitions): Ditto and make it static.
2448 (get_undefined_value_partitions): Ditto and make it static.
2449 (remove_ssa_form): Refactor call to init_var_map here.
2450 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2451 computation for loop region.
2452 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2453 (register_default_def): Delete.
2454 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2455 (parm_default_def_partition_arg): Ditto.
2456 (set_parm_default_def_partition): Ditto.
2457 (get_parm_default_def_partitions): Ditto and make it static.
2458 (get_undefined_value_partitions): Ditto and make it static.
2459 (coalesce_with_default, coalesce_with_default): Update comment.
2460 (create_coalesce_list_for_region): New func factored out from
2461 create_outofssa_var_map.
2462 (populate_coalesce_list_for_outofssa): New func factored out from
2463 create_outofssa_var_map and coalesce_ssa_name.
2464 (create_outofssa_var_map): Delete.
2465 (coalesce_ssa_name): Refactor to support live range computation.
2466 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2467 (get_parm_default_def_partitions): Delete.
2468 (get_undefined_value_partitions): Ditto.
2469 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2470 computation for loop region.
2471 (new_tree_live_info, loe_visit_block): Ditto.
2472 (live_worklist, set_var_live_on_entry): Ditto.
2473 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2474 * tree-ssa-live.h (struct _var_map): New fields.
2475 (init_var_map): Change decl.
2476 (region_contains_p): New.
2478 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2480 * tree-ssa-live.h (live_merge_and_clear): Delete.
2482 2018-05-25 Richard Biener <rguenther@suse.de>
2485 * tree-dump.c (dequeue_and_dump): Remove access to removed
2486 operand 2 of a SWITCH_EXPR.
2488 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2490 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2491 multiplication and division.
2492 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2493 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2494 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2495 (cond_udiv_optab, cond_umod_optab): New optabs.
2496 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2497 (IFN_COND_RDIV): New internal functions.
2498 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2499 TRUNC_MOD_EXPR and RDIV_EXPR.
2500 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2501 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2503 (SVE_INT_BINARY): Include mult.
2504 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2505 (optab, sve_int_op): Handle mult.
2506 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2508 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2509 for SVE_INT_BINARY_SD.
2511 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2513 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2514 (optab, sve_int_op): Handle div and udiv.
2515 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2516 for SVE_INT_BINARY_SD.
2517 (*<optab><mode>3): New insn for the same.
2519 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2521 * tree-vect-patterns.c: Include predict.h.
2522 (vect_recog_divmod_pattern): Restrict check for division support
2523 to when optimizing for size.
2525 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2527 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2528 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2529 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2530 (gimple_match_op::set_op): Likewise.
2531 (gimple_resimplify4): Declare.
2532 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2533 (expr::gen_transform): Likewise.
2534 (decision_tree::gen): Generate a simplification routine for 4 operands.
2535 * gimple-match-head.c (gimple_simplify): Add an overload for
2536 4 operands. In the top-level function, handle up to 4 call
2537 arguments and call gimple_resimplify4.
2538 (gimple_resimplify4): New function.
2539 (build_call_internal): Pass a fourth operand.
2540 (maybe_push_to_seq): Likewise.
2541 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2542 Fold VEC_COND_EXPRs of an operation and a default value into
2543 an IFN_COND_* function if possible.
2544 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2546 (SVE_COND_FP_BINARY): Include them.
2547 (optab, sve_fp_op): Handle them.
2548 (SVE_INT_BINARY_REV): New code iterator.
2549 (SVE_COND_FP_BINARY_REV): New int iterator.
2550 (commutative): New int attribute.
2551 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2553 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2555 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2556 (*cond_<optab><mode>): New patterns for reversed operands.
2558 2018-05-25 Richard Biener <rguenther@suse.de>
2560 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2561 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2562 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2563 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2564 * tree-vect-data-refs.c (everywhere): Adjust users.
2565 * tree-vect-loop.c (everywhere): Likewise.
2566 * tree-vect-slp.c (everywhere): Likewise.
2567 * tree-vect-stmts.c (everywhere): Likewise.
2568 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2570 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2572 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2574 (gcc_cv_as_section_exclude): ... this.
2575 Try Solaris as #exclude syntax.
2576 * configure: Regenerate.
2577 * config.in: Regenerate.
2578 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2580 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2581 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2583 * varasm.c (default_elf_asm_named_section): Don't check if
2584 HAVE_GAS_SECTION_EXCLUDE is defined.
2586 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2588 * doc/md.texi: Update the documentation of the cond_* optabs
2589 to mention the new final operand. Fix GET_MODE_NUNITS call.
2590 Describe the scalar case too.
2591 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2592 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2594 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2595 (get_conditional_internal_fn): Update comment.
2596 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2597 accumulator value as a final argument to conditional functions.
2598 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2599 a define_expand and add an "else" operand. Assert for now that
2600 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2601 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2602 (*cond_<optab><mode>): New patterns.
2603 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2604 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2605 (UNSPEC_COND_EOR): Delete.
2606 (optab): Remove associated mappings.
2607 (SVE_INT_BINARY): New code iterator.
2608 (sve_int_op): Remove int attribute and add "minus" to the code
2610 (SVE_COND_INT_OP): Delete.
2611 (SVE_COND_FP_OP): Rename to...
2612 (SVE_COND_FP_BINARY): ...this.
2614 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2616 * optabs.c (can_reuse_operands_p): New function.
2617 (maybe_legitimize_operands): Try to reuse the results for
2620 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2622 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2623 Add {q} suffix to insn mnemonic.
2625 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2627 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2628 (msp430_warn_func_return): New.
2630 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2632 * fold-const.c (tree_nonzero_bits): New function.
2633 * fold-const.h (tree_nonzero_bits): Likewise.
2634 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2635 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2637 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2641 * varasm.c (assemble_alias): Check ifunc_resolver only on
2644 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2647 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2648 when memory input operand is handled.
2650 2018-05-24 Luis Machado <luis.machado@linaro.org>
2652 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2654 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2656 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2658 * match.pd: Delay FMA folds until after vectorization.
2660 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2663 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2664 address check not strict.
2666 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2668 * gimple-match.h (gimple_match_op): New class.
2669 (mprts_hook): Replace parameters with a gimple_match_op *.
2670 (maybe_build_generic_op): Likewise.
2671 (gimple_simplified_result_is_gimple_val): Replace parameters with
2672 a const gimple_match_op *.
2673 (gimple_simplify): Replace code_helper * and tree * parameters with
2674 a gimple_match_op * parameter.
2675 (gimple_resimplify1): Replace code_helper *, tree and tree *
2676 parameters with a gimple_match_op * parameter.
2677 (gimple_resimplify2): Likewise.
2678 (gimple_resimplify3): Likewise.
2679 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2680 parameters with a gimple_match_op * parameter.
2681 * gimple-match-head.c (gimple_simplify): Change prototypes of
2682 auto-generated functions to take a gimple_match_op * instead of
2683 separate code_helper * and tree * parameters. Make the same
2684 change in the top-level overload and update calls to the
2685 gimple_resimplify routines. Update calls to the auto-generated
2686 functions and to maybe_push_res_to_seq in the publicly-facing
2687 operation-specific gimple_simplify overloads.
2688 (gimple_match_op::MAX_NUM_OPS): Define.
2689 (gimple_resimplify1): Replace rcode and ops with a single res_op
2690 parameter. Update call to gimple_simplify.
2691 (gimple_resimplify2): Likewise.
2692 (gimple_resimplify3): Likewise.
2693 (mprts_hook): Replace parameters with a gimple_match_op *.
2694 (maybe_build_generic_op): Likewise.
2695 (build_call_internal): Replace type, nargs and ops with
2696 a gimple_match_op *.
2697 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2698 with a single gimple_match_op *. Update calls to mprts_hook,
2699 build_call_internal and gimple_simplified_result_is_gimple_val.
2700 Factor out code that is common to the tree_code and combined_fn cases.
2701 * genmatch.c (expr::gen_transform): Replace tem_code and
2702 tem_ops with a gimple_match_op called tem_op. Update calls
2703 to the gimple_resimplify functions and maybe_push_res_to_seq.
2704 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2705 res_ops. Update call to the gimple_resimplify functions.
2706 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2707 (decision_tree::gen): Make the functions take a gimple_match_op *
2708 called res_op instead of separate res_code and res_ops parameters.
2709 Update call accordingly.
2710 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2711 and ops with a single res_op parameter. Update calls to
2712 maybe_build_generic_op and maybe_push_res_to_seq.
2713 (fold_stmt_1): Update calls to gimple_simplify and
2714 replace_stmt_with_simplification.
2715 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2716 and gimple_simplified_result_is_gimple_val.
2717 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2719 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2720 with a gimple_match_op *.
2721 (vn_nary_build_or_lookup): Likewise. Update call to
2722 vn_nary_build_or_lookup_1.
2723 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2724 gimple_match_op *. Update calls to the gimple_resimplify routines
2725 and to gimple_simplified_result_is_gimple_val.
2726 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2727 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2728 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2729 (visit_nary_op): Likewise.
2730 (visit_reference_op_load): Likewise.
2732 2018-05-23 Luis Machado <luis.machado@linaro.org>
2734 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2735 modifier for printing the step amount.
2737 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2740 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2743 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2745 * doc/sourcebuild.texi (Endianness): New subsubsection.
2747 2018-05-23 Luis Machado <luis.machado@linaro.org>
2749 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2750 <prefetch_dynamic_strides>: New const bool field.
2751 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2752 prefetch_dynamic_strides.
2753 (exynosm1_prefetch_tune): Likewise.
2754 (thunderxt88_prefetch_tune): Likewise.
2755 (thunderx_prefetch_tune): Likewise.
2756 (thunderx2t99_prefetch_tune): Likewise.
2757 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2759 (aarch64_override_options_internal): Update to set
2760 PARAM_PREFETCH_DYNAMIC_STRIDES.
2761 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2762 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2763 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2764 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2765 prefetch-dynamic-strides setting.
2767 2018-05-23 Luis Machado <luis.machado@linaro.org>
2769 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2770 <minimum_stride>: New const int field.
2771 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2772 minimum_stride field defaulting to -1.
2773 (exynosm1_prefetch_tune): Likewise.
2774 (thunderxt88_prefetch_tune): Likewise.
2775 (thunderx_prefetch_tune): Likewise.
2776 (thunderx2t99_prefetch_tune): Likewise.
2777 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2778 <default_opt_level>: Set to 3.
2779 (aarch64_override_options_internal): Update to set
2780 PARAM_PREFETCH_MINIMUM_STRIDE.
2781 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2782 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2783 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2784 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2785 stride is constant and is below the minimum stride threshold.
2787 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2789 * config/arm/arm-cpus.in (mode26): Delete.
2790 (armv4): Delete mode26 reference.
2791 * config/arm/arm.c (arm_configure_build_target): Delete use of
2794 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2796 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2798 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2799 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2801 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2802 (floatunsdidf2): Ditto.
2804 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2806 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2807 (fixuns_trunc<mode>si2_avx512f): Ditto.
2808 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2809 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2810 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2812 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2814 PR rtl-optimization/79985
2815 * df-scan.c (df_insn_refs_collect): Remove special case for
2816 global registers and asm statements.
2818 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2820 * extend.texi (Global Register Variables): Rewrite the bullet list.
2821 Note that the register is available for allocation. Note that access
2822 via inline asm must use constraints. Add note about async-signal
2823 handlers. Remove paragraph about automagic register selection.
2825 2018-05-23 Richard Biener <rguenther@suse.de>
2827 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2828 of fixed offset from memset VN.
2830 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2832 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2834 (alloc_cand_and_find_basis): Initialize first_interp field.
2835 (slsr_process_mul): Modify first_interp field.
2836 (slsr_process_add): Likewise.
2837 (slsr_process_cast): Modify first_interp field for each new
2839 (slsr_process_copy): Likewise.
2840 (dump_candidate): Dump first_interp field.
2841 (replace_mult_candidate): Process all interpretations, not just
2843 (replace_rhs_if_not_dup): Likewise.
2844 (replace_one_candidate): Likewise.
2846 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2848 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2850 (aarch64_needs_frame_chain): New function.
2851 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2853 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2856 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2857 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2858 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2859 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2860 as true for strict-align.
2861 (aarch64_can_inline_p): Perform checks even when callee has no
2862 attributes to check for strict alignment.
2863 * doc/extend.texi (AArch64 Function Attributes): Document
2865 * doc/invoke.texi: (AArch64 Options): Likewise.
2867 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2869 PR tree-optimization/85853
2870 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2871 the handling of the root of the node to...
2872 (vect_slp_analyze_node_operations_1): ...this new function,
2873 and run the whole thing with the child nodes' def types
2874 set according to their SLP node's def type.
2876 2018-05-23 Richard Biener <rguenther@suse.de>
2879 * tree-data-ref.c (create_runtime_alias_checks): Defer
2880 and ignore overflow warnings.
2882 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2884 PR tree-optimization/85822
2885 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2888 2018-05-23 Richard Biener <rguenther@suse.de>
2890 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2891 memset constants via native_interpret_expr.
2893 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2896 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2898 (cgraph_node::create_alias): Likewise.
2899 (cgraph_node::get_availability): Check ifunc_resolver instead
2900 of looking up ifunc attribute.
2901 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2902 * varasm.c (do_assemble_alias): Likewise.
2903 (assemble_alias): Likewise.
2904 (default_binds_local_p_3): Likewise.
2905 * cgraph.h (cgraph_node): Add ifunc_resolver.
2906 (cgraph_node::only_called_directly_or_aliased_p): Return false
2908 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2910 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2911 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2912 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2913 instead of looking up ifunc attribute.
2915 2018-05-22 Luis Machado <luis.machado@linaro.org>
2917 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2919 2018-05-22 Martin Sebor <msebor@redhat.com>
2922 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
2923 only when expasion succeeds.
2924 (expand_builtin_strcmp): Same.
2925 (expand_builtin_strncmp): Same.
2927 2018-05-22 Martin Sebor <msebor@redhat.com>
2929 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
2931 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2932 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2934 * config/aarch64/aarch64-ldpstp.md: Replace uses of
2935 aarch64_mem_pair_operand with memory_operand and delete operand swapping
2937 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2938 Add check for legitimate_address.
2939 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
2940 (aarch64_swap_ldrstr_operands): New.
2941 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
2944 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
2945 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2947 * config/aarch64/aarch64.md: New patterns to generate stp
2949 (store_pair_sw, store_pair_dw): New patterns to generate stp for
2950 single words and double words.
2951 (load_pair_sw, load_pair_dw): Likewise.
2952 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
2954 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
2956 * config/aarch64/aarch64-ldpstp.md: Modify peephole
2957 for different mode ldpstp and add peephole for merged zero stores.
2959 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
2961 (aarch64_gen_store_pair): Rename calls to match new patterns.
2962 (aarch64_gen_load_pair): Rename calls to match new patterns.
2963 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
2964 (load_pair<DREG:mode><DREG2:mode>): ... This.
2965 (store_pair<mode>): Rename to...
2966 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
2967 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
2969 (V_INT_EQUIV): Handle SImode.
2970 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
2973 2018-05-22 Martin Sebor <msebor@redhat.com>
2976 * calls.c (maybe_warn_nonstring_arg): Use string length to set
2977 or ajust the presumed bound on an operation to avoid unnecessary
2980 2018-05-22 Martin Sebor <msebor@redhat.com>
2982 PR tree-optimization/85826
2983 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
2984 assuming that a DECL necesarily has a constant size.
2986 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
2989 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
2991 2018-05-22 Richard Biener <rguenther@suse.de>
2993 PR tree-optimization/85834
2994 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
2995 non-constant and non-zero memset arguments.
2997 2018-05-22 Martin Liska <mliska@suse.cz>
3000 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3002 2018-05-22 Richard Biener <rguenther@suse.de>
3004 PR tree-optimization/85863
3005 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3006 comparisons when vectype is specified.
3007 (vectorizable_condition): Do not specify vectype for
3008 vect_is_simple_cond when SLP vectorizing.
3010 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3013 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3014 define __ibm128 as long double.
3015 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3016 as a distinct type when IEEE 128-bit support is enabled.
3017 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3018 128-bit types to use the correct functions.
3019 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3020 convert between 128-bit floating point types that have different
3021 modes but the same representation, instead of using gen_lowpart to
3023 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3025 (IFKF_reg): New attributes to give the register constraints for
3027 (extend<mode>tf2_internal): New insns to mark an explicit
3028 conversion between 128-bit floating point types that have a
3029 different mode but share the same representation.
3031 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3033 PR tree-optimization/85814
3034 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3035 a null return from get_strinfo when unsharing the next
3036 strinfo in the chain.
3038 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3041 * varasm.c (weak_finish): Clean up weak_decls.
3043 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3045 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3046 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3047 UNSPEC_UADALP values.
3048 * config/aarch64/iterators.md (ABAL): New int iterator.
3051 (sur): Add mappings for the above.
3052 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3054 (aarch64_<sur>abal<mode>_4): Likewise.
3055 (aarch64_<sur>adalp<mode>_3): Likewise.
3056 (<sur>sadv16qi): New define_expand.
3058 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3060 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3061 (*movdf_internal): Ditto.
3062 (*rcpsf2_sse): Ditto.
3063 (*rsqrtsf2_sse): Ditto.
3064 (*sqrt<mode>2_sse): Ditto.
3066 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3068 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3070 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3071 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3072 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3073 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3075 * config/aarch64/arm_neon.h: Likewise.
3076 * config/aarch64/iterators.md (VQ_I): New.
3078 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3080 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3082 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3085 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3087 * config/nds32/constraints.md (S): New constraint.
3088 * config/nds32/nds32.md (call_internal): Use constraint S.
3089 (call_value_internal): Likewise.
3090 (sibcall_internal): Likewise.
3091 (sibcall_value_internal): Likewise.
3093 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3094 Chung-Ju Wu <jasonwucj@gmail.com>
3096 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3099 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3100 Chung-Ju Wu <jasonwucj@gmail.com>
3102 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3103 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3104 (nds32_rtx_costs_impl): Simplify.
3105 (nds32_address_cost_impl): Simplify.
3106 (nds32_init_rtx_costs): New function.
3107 (nds32_rtx_costs_speed_prefer): Likewise.
3108 (nds32_rtx_costs_size_prefer): Likewise.
3109 (nds32_address_cost_speed_prefer): Likewise.
3110 (nds32_address_cost_speed_fwprop): Likewise.
3111 (nds32_address_cost_size_prefer): Likewise.
3112 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3113 * config/nds32/nds32.c (nds32_option_override): Use
3114 nds32_init_rtx_costs function.
3116 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3118 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3119 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3120 (TARGET_PIPELINE_N8): Likewise.
3121 (TARGET_PIPELINE_N10): Likewise.
3122 (TARGET_PIPELINE_N13): Likewise.
3123 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3125 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3127 * config/nds32/nds32-fpu.md: Update copyright year.
3129 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3131 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3133 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3135 * config/nds32/nds32.c
3136 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3137 * config/nds32/nds32.opt (minline-asm-r15): New option.
3139 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3141 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3143 * config/nds32/nds32.md (abssi2): New pattern.
3145 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3147 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3148 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3149 Merge insn pattern from sse_cvtsi2ss<round_name> and
3150 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3151 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3152 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3153 using SWI48 mode iterator.
3154 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3155 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3156 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3157 pattern from sse_cvttss2si<round_saeonly_name>
3158 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3159 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3160 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3161 using SWI48 mode iterator.
3162 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3163 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3164 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3165 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3166 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3167 using SWI48 mode iterator.
3168 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3169 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3170 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3171 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3172 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3173 SWI48 mode iterator.
3174 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3175 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3176 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3177 pattern from sse_cvttsd2si<round_saeonly_name>
3178 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3180 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3182 * config/nds32/nds32-md-auxiliary.c
3183 (nds32_valid_smw_lwm_base_p): Refine.
3184 (nds32_output_smw_single_word): Refine.
3185 (nds32_output_smw_double_word): New.
3186 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3188 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3191 (nds32_output_stack_pop): Refine.
3192 (nds32_expand_unaligned_load): Refine.
3193 (nds32_expand_unaligned_store): Refine.
3195 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3196 Chung-Ju Wu <jasonwucj@gmail.com>
3198 * config/nds32/constants.md: Add TP_REGNUM constant.
3199 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3200 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3202 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3203 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3204 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3205 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3206 and PIC code generation.
3207 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3209 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3211 * config/nds32/nds32.md: Support TLS and PIC.
3212 * config/nds32/nds32.c: Support TLS and PIC.
3213 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3214 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3217 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3219 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3220 mode with E_ prefix.
3222 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3223 Chung-Ju Wu <jasonwucj@gmail.com>
3225 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3226 * config/nds32/nds32-md-auxiliary.c
3227 (symbolic_reference_mentioned_p): New.
3228 (nds32_legitimize_ict_address): New.
3229 (nds32_expand_ict_move): New.
3230 (nds32_indirect_call_referenced_p): New.
3231 (nds32_symbol_binds_local_p): Delete.
3232 (nds32_long_call_p): Modify.
3233 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3234 * config/nds32/nds32-protos.h
3235 (symbolic_reference_mentioned_p): Declare.
3236 (nds32_legitimize_ict_address): Declare.
3237 (nds32_expand_ict_move): Declare.
3238 (nds32_indirect_call_referenced_p): Declare.
3239 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3240 (nds32_relax_group): Use nds32_ict_const_p as condition.
3241 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3242 (nds32_asm_file_start): Output ict_model directive in asm code.
3243 (nds32_legitimate_address_p): Consider indirect call.
3244 (nds32_print_operand): Consider indirect call.
3245 (nds32_print_operand_address): Consider indirect call.
3246 (nds32_insert_attributes): Handle "indirect_call" attribute.
3247 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3248 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3249 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3250 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3251 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3252 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3253 (TARGET_ICT_MODEL_SMALL): Define.
3254 (TARGET_ICT_MODEL_LARGE): Define.
3255 * config/nds32/nds32.md (movsi): Consider ict model.
3256 (call, call_value): Consider ict model.
3257 (sibcall, sibcall_value): Consider ict model.
3258 * config/nds32/nds32.opt (mict-model): New option.
3259 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3262 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3263 Monk Chiang <sh.chiang04@gmail.com>
3264 Jim Wilson <jimw@sifive.com>
3266 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3267 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3268 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3269 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3270 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3271 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3272 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3273 compute save_libcall_adjustment properly.
3274 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3275 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3276 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3277 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3278 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3279 (ABI_SPEC): Handle mabi=ilp32e.
3280 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3281 (RVE): Add RVE mask.
3282 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3283 <-march>: Add rv32e as an example.
3285 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3288 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3289 (intra_create_variable_infos): Handle C++ constructors.
3291 2018-05-18 Martin Liska <mliska@suse.cz>
3293 * passes.def: Remove a redundant pass.
3295 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3298 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3300 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3302 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3304 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3306 (armv2, armv2a, armv3, armv3m): Delete architectures.
3307 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3308 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3309 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3311 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3312 (*mulsidi3adddi): Likewise.
3313 (mulsidi3): Likewise.
3314 (*mulsidi3_nov6): Likewise.
3315 (umulsidi3): Likewise.
3316 (umulsidi3_nov6): Likewise.
3317 (umaddsidi4): Likewise.
3318 (*umulsidi3adddi): Likewise.
3319 (smulsi3_highpart): Likewise.
3320 (*smulsi3_highpart_nov6): Likewise.
3321 (umulsi3_highpart): Likewise.
3322 (*umulsi3_highpart_nov6): Likewise.
3323 * config/arm/arm.h (arm_arch3m): Delete.
3324 * config/arm/arm.c (arm_arch3m): Delete.
3325 (arm_option_override_internal): Update armv3-related comment.
3326 (arm_configure_build_target): Delete use of isa_bit_mode32.
3327 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3328 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3329 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3331 (mulusa3): Likewise.
3332 * config/arm/arm-protos.h (arm_arch3m): Delete.
3333 * config/arm/arm-tables.opt: Regenerate.
3334 * config/arm/arm-tune.md: Likewise.
3335 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3336 deleted architectures.
3338 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3340 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3341 (armv5t, armv5te): New features.
3342 (ARMv5, ARMv5e): Delete fgroups.
3343 (ARMv5t, ARMv5te): Adjust for above changes.
3345 (armv5, armv5e): Delete arches.
3346 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3348 (*call_reg_arm): Likewise.
3349 (*call_value_reg_armv5): Likewise.
3350 (*call_value_reg_arm): Likewise.
3351 (*call_symbol): Likewise.
3352 (*call_value_symbol): Likewise.
3353 (*sibcall_insn): Likewise.
3354 (*sibcall_value_insn): Likewise.
3356 (prefetch): Likewise.
3357 (define_split and define_peephole2 dependent on arm_arch5):
3359 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3361 (TARGET_ARM_QBIT): Likewise.
3362 (TARGET_DSP_MULTIPLY): Likewise.
3363 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3364 (arm_arch5, arm_arch5e): Delete.
3365 (arm_arch5t, arm_arch5te): Declare.
3366 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3367 (arm_arch5t): Declare.
3368 (arm_option_reconfigure_globals): Update for the above.
3369 (arm_options_perform_arch_sanity_checks): Update comment, replace
3370 use of arm_arch5 with arm_arch5t.
3371 (use_return_insn): Likewise.
3372 (arm_emit_call_insn): Likewise.
3373 (output_return_instruction): Likewise.
3374 (arm_final_prescan_insn): Likewise.
3375 (arm_coproc_builtin_available): Likewise.
3376 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3377 arm_arch5e with arm_arch5t and arm_arch5te.
3378 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3379 (arm_arch5t, arm_arch5te): Declare.
3380 * config/arm/arm-tables.opt: Regenerate.
3381 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3382 * config/arm/t-multilib: Likewise.
3383 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3384 instead of arm_arch5.
3385 (*call_reg_thumb1): Likewise.
3386 (*call_value_reg_thumb1_v5): Likewise.
3387 (*call_value_reg_thumb1): Likewise.
3388 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3390 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3392 2018-05-18 Martin Liska <mliska@suse.cz>
3394 PR gcov-profile/84846
3395 * doc/gcov.texi: Document -t option of gcov tool.
3397 2018-05-18 Martin Liska <mliska@suse.cz>
3399 PR gcov-profile/84846
3400 * gcov.c (print_usage): Add new -t option.
3401 (process_args): Handle the option.
3402 (generate_results): Use stdout as output when requested by
3405 2018-05-18 Martin Liska <mliska@suse.cz>
3407 PR gcov-profile/84846
3408 * coverage.c (coverage_init): Write PWD to .gcno file.
3409 * doc/gcov.texi: Document how working directory is printed.
3410 * gcov-dump.c (dump_gcov_file): Print PWD.
3411 * gcov.c (output_intermediate_file): Likewise.
3412 (read_graph_file): Read PWD string.
3413 (output_lines): Print PWD.
3415 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3418 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3419 for retval and return false if all args to phi are zero.
3421 2018-05-18 Richard Biener <rguenther@suse.de>
3423 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3425 (evrp_dom_walker::before_dom_children): Call it.
3427 2018-05-18 Richard Biener <rguenther@suse.de>
3429 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3430 results when processing array refs with variable index.
3432 2018-05-18 Toon Moene <toon@moene.org>
3434 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3435 directly after that of -floop-interchange. Indicate that both
3436 options are enabled by default when specifying -O3.
3438 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3440 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3441 iterator. Delete separate integer-mode vec_set<mode> expander.
3442 (aarch64_simd_vec_setv2di): Delete.
3443 (vec_setv2di): Delete.
3444 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3445 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3446 the "w, r" alternative.
3448 2018-05-18 Martin Liska <mliska@suse.cz>
3450 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3451 * tree-pass.h (make_pass_lower_switch_O0): New function.
3452 * tree-switch-conversion.c (node_has_low_bound): Remove.
3453 (node_has_high_bound): Likewise.
3454 (node_is_bounded): Likewise.
3455 (class pass_lower_switch): Make it a template type and create
3457 (pass_lower_switch::execute): Add template argument.
3458 (make_pass_lower_switch): New function.
3459 (make_pass_lower_switch_O0): New function.
3460 (do_jump_if_equal): Remove.
3461 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3462 all the hard work to tree optimization passes.
3464 2018-05-18 Martin Liska <mliska@suse.cz>
3466 * dbgcnt.c (limit_low): Renamed from limit.
3467 (limit_high): New variable.
3468 (dbg_cnt_is_enabled): Check for upper limit.
3469 (dbg_cnt): Adjust dumping.
3470 (dbg_cnt_set_limit_by_index): Add new argument for high
3472 (dbg_cnt_set_limit_by_name): Likewise.
3473 (dbg_cnt_process_single_pair): Parse new format.
3474 (dbg_cnt_process_opt): Use strtok.
3475 (dbg_cnt_list_all_counters): Remove 'value' and add
3477 * doc/invoke.texi: Document changes.
3479 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3481 * doc/sourcebuild.texi (scalar_all_fma): Document.
3482 * tree.def (FMA_EXPR): Delete.
3483 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3484 * internal-fn.c (ternary_direct): New macro.
3485 (expand_ternary_optab_fn): Likewise.
3486 (direct_ternary_optab_supported_p): Likewise.
3487 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3488 * builtins.c (fold_builtin_fma): Delete.
3489 (fold_builtin_3): Don't call it.
3490 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3491 * expr.c (expand_expr_real_2): Likewise.
3492 * fold-const.c (operand_equal_p): Likewise.
3493 (fold_ternary_loc): Likewise.
3494 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3495 * gimple.c (DEFTREECODE): Likewise.
3496 * gimplify.c (gimplify_expr): Likewise.
3497 * optabs-tree.c (optab_for_tree_code): Likewise.
3498 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3499 * tree-eh.c (operation_could_trap_p): Likewise.
3500 (stmt_could_throw_1_p): Likewise.
3501 * tree-inline.c (estimate_operator_cost): Likewise.
3502 * tree-pretty-print.c (dump_generic_node): Likewise.
3503 (op_code_prio): Likewise.
3504 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3505 * tree-ssa-operands.c (get_expr_operands): Likewise.
3506 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3507 * fold-const-call.h (fold_fma): Delete.
3508 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3509 CFN_FNMA and CFN_FNMS.
3511 * genmatch.c (combined_fn): New enum.
3512 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3513 (commutative_op): New function.
3514 (commutate): Use it. Handle more than 2 operands.
3515 (dt_operand::gen_gimple_expr): Use commutative_op.
3516 (parser::parse_expr): Allow :c to be used with non-binary
3517 operators if the commutative operand is known.
3518 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3519 CFN_FMS, CFN_FNMA and CFN_FNMS.
3520 (backprop::process_assign_use): Remove FMA_EXPR handling.
3521 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3522 (gen_hsa_fma): New function.
3523 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3524 IFN_FNMA and IFN_FNMS.
3525 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3526 * gimple-fold.h (follow_all_ssa_edges): Declare.
3527 * gimple-fold.c (follow_all_ssa_edges): New function.
3528 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3529 gimple_build interface and use follow_all_ssa_edges to fold the result.
3530 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3531 instead of checking for optabs directly.
3532 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3533 rather than FMA_EXPRs.
3534 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3535 call to IFN_FMA instead of an FMA_EXPR.
3537 2018-05-17 Jim Wilson <jimw@sifive.com>
3539 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3540 sign extended promoted subreg, and the range does not have the sign bit
3541 set, then do a sign extend.
3543 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3544 test, check for sign extended subreg and/or constant operands, and
3545 do a sign extend in that case.
3547 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3549 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3550 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3552 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3553 Change logics_shift_reg to logics_shift_imm.
3554 (thunderx2t99_fp_loadpair_basic): Delete.
3555 (thunderx2t99_fp_storepair_basic): Delete.
3556 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3557 (thunderx2t99_asimd_polynomial): Delete.
3558 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3559 and neon_fp_mul_d_scalar_q.
3560 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3561 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3562 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3563 (thunderx2t99_asimd_lut): Add missing tbl types.
3564 (thunderx2t99_asimd_ext): Delete.
3565 (thunderx2t99_asimd_load1_1_mult): Delete.
3566 (thunderx2t99_asimd_load1_2_mult): Delete.
3567 (thunderx2t99_asimd_load1_ldp): New.
3568 (thunderx2t99_asimd_load1): New.
3569 (thunderx2t99_asimd_load2): Add missing *load2* types.
3570 (thunderx2t99_asimd_load3): New.
3571 (thunderx2t99_asimd_load4): New.
3572 (thunderx2t99_asimd_store1_1_mult): Delete.
3573 (thunderx2t99_asimd_store1_2_mult): Delete.
3574 (thunderx2t99_asimd_store2_mult): Delete.
3575 (thunderx2t99_asimd_store2_onelane): Delete.
3576 (thunderx2t99_asimd_store_stp): New.
3577 (thunderx2t99_asimd_store1): New.
3578 (thunderx2t99_asimd_store2): New.
3579 (thunderx2t99_asimd_store3): New.
3580 (thunderx2t99_asimd_store4): New.
3582 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3584 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3585 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3587 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3588 Segher Boessenkool <segher@kernel.crashing.org>
3591 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3594 2018-05-17 Richard Biener <rguenther@suse.de>
3596 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3597 for pruning loop and prune defs feeding only already visited PHIs.
3599 2018-05-17 Richard Biener <rguenther@suse.de>
3601 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3603 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3604 Richard Biener <rguenther@suse.de>
3606 PR tree-optimization/85793
3607 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3608 for VMAT_ELEMENTWISE.
3610 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3612 * internal-fn.h (lookup_internal_fn): Declare
3613 * internal-fn.c (lookup_internal_fn): New function.
3614 * gimple.c (gimple_build_call_from_tree): Handle calls to
3616 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3617 internal function names.
3618 * tree-pretty-print.c (dump_generic_node): Likewise.
3619 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3621 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3623 * gimple-fold.h (gimple_build): Make the function forms take
3624 combined_fn rather than built_in_function.
3625 (gimple_simplify): Likewise.
3626 * gimple-match-head.c (gimple_simplify): Likewise.
3627 * gimple-fold.c (gimple_build): Likewise.
3628 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3629 rather than gimple_build_call_internal.
3630 (get_initial_defs_for_reduction): Likewise.
3631 (vect_create_epilog_for_reduction): Likewise.
3632 (vectorizable_live_operation): Likewise.
3634 2018-05-17 Martin Liska <mliska@suse.cz>
3636 * gimple-ssa-sprintf.c (format_directive): Do not use
3637 space in between 'G_' and '('.
3639 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3642 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3643 even if the mask is not all ones.
3646 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3648 (ix86_gimple_fold_builtin): Likewise.
3651 * config/i386/i386.c: Include tree-vector-builder.h.
3652 (ix86_vector_shift_count): New function.
3653 (ix86_fold_builtin): Fold shift builtins by scalar count.
3654 (ix86_gimple_fold_builtin): Likewise.
3656 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3657 _mm512_setzero): New intrinsics.
3659 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3660 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3662 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3663 code generation for cases where splatting a value is not useful.
3664 * simplify-rtx.c (simplify_ternary_operation): Simplify
3665 vec_merge across a vec_duplicate and a paradoxical subreg forming
3666 a vector mode to a vec_concat.
3668 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3670 * config.gcc: Support "goldmont-plus".
3671 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3673 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3674 PROCESSOR_GOLDMONT_PLUS.
3675 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3676 (processor_target_table): Add "goldmont-plus".
3677 (PTA_GOLDMONT_PLUS): Define.
3678 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3679 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3680 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3681 (fold_builtin_cpu): Add "goldmont-plus".
3682 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3683 (ix86_option_override_internal): Add "goldmont-plus".
3684 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3685 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3686 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3687 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3689 2018-05-17 Richard Biener <rguenther@suse.de>
3691 PR tree-optimization/85757
3692 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3693 remove defs that only feed that PHI from further processing.
3695 2018-05-16 Jim Wilson <jimw@sifive.com>
3697 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3699 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3701 2018-05-16 Mark Wielaard <mark@klomp.org>
3703 * dwarf2out.c (count_index_strings): New function.
3704 (output_indirect_strings): Call count_index_strings and generate
3705 header for dwarf_version >= 5.
3707 2018-05-16 Mark Wielaard <mark@klomp.org>
3709 * dwarf2out.c (dwarf_FORM): New function.
3710 (set_indirect_string): Use dwarf_FORM.
3711 (reset_indirect_string): Likewise.
3712 (size_of_die): Likewise.
3713 (value_format): Likewise.
3714 (output_die): Likewise.
3715 (add_skeleton_AT_string): Likewise.
3716 (output_macinfo_op): Likewise.
3717 (index_string): Likewise.
3718 (output_index_string_offset): Likewise.
3719 (output_index_string): Likewise.
3720 (count_index_strings): Likewise.
3722 2018-05-16 Carl Love <cel@us.ibm.com>
3724 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3725 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3727 2018-05-16 Martin Jambor <mjambor@suse.cz>
3729 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3730 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3732 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3734 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3735 (fnma<mode>4): Likewise.
3736 (fms<mode>4): Likewise.
3737 (fnms<mode>4): Likewise.
3738 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3739 (aarch64_fnma<mode>4): Likewise.
3740 (aarch64_fms<mode>4): Likewise.
3741 (aarch64_fnms<mode>4): Likewise.
3742 (aarch64_fnmadd<mode>4): Likewise.
3744 2018-05-16 Jason Merrill <jason@redhat.com>
3746 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3748 2018-05-16 Richard Biener <rguenther@suse.de>
3750 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3751 (dump_stmt_cost): Declare.
3752 (add_stmt_cost): Dump cost we add.
3753 (add_stmt_costs): New function.
3754 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3756 (vect_analyze_stmt): Adjust prototype.
3757 (vectorizable_condition): Likewise.
3758 (vectorizable_live_operation): Likewise.
3759 (vectorizable_reduction): Likewise.
3760 (vectorizable_induction): Likewise.
3761 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3762 cost vector to pass to vectorizable_ and record afterwards.
3763 (vect_model_reduction_cost): Take cost vector argument and adjust.
3764 (vect_model_induction_cost): Likewise.
3765 (vectorizable_reduction): Likewise.
3766 (vectorizable_induction): Likewise.
3767 (vectorizable_live_operation): Likewise.
3768 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3769 SLP_TREE_NUMBER_OF_VEC_STMTS.
3770 (vect_analyze_slp_cost_1): Remove.
3771 (vect_analyze_slp_cost): Likewise.
3772 (vect_slp_analyze_node_operations): Take visited args and
3773 a target cost vector. Avoid processing already visited stmt sets.
3774 (vect_slp_analyze_operations): Use a local cost vector to gather
3775 costs and register those of non-discarded instances.
3776 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3777 (vect_schedule_slp_instance): Remove copying of
3778 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3780 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3781 adding cost. Record cost entry location.
3782 (vect_prologue_cost_for_slp_op): Function to compute cost of
3783 a constant or invariant generated for SLP vect in the prologue,
3784 split out from vect_analyze_slp_cost_1.
3785 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3786 (vect_model_promotion_demotion_cost): Likewise.
3787 (vect_model_store_cost): Likewise, make static.
3788 (vect_model_load_cost): Likewise.
3789 (vectorizable_bswap): Add cost vector arg and adjust.
3790 (vectorizable_call): Likewise.
3791 (vectorizable_simd_clone_call): Likewise.
3792 (vectorizable_conversion): Likewise.
3793 (vectorizable_assignment): Likewise.
3794 (vectorizable_shift): Likewise.
3795 (vectorizable_operation): Likewise.
3796 (vectorizable_store): Likewise.
3797 (vectorizable_load): Likewise.
3798 (vectorizable_condition): Likewise.
3799 (vectorizable_comparison): Likewise.
3800 (can_vectorize_live_stmts): Likewise.
3801 (vect_analyze_stmt): Likewise.
3802 (vect_transform_stmt): Adjust calls to vectorizable_*.
3803 * tree-vectorizer.c: Include gimple-pretty-print.h.
3804 (dump_stmt_cost): New function.
3806 2018-05-16 Richard Biener <rguenther@suse.de>
3808 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3809 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3810 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3811 (check_name): New callback.
3812 (dse_classify_store): Track cycles via a visited bitmap of PHI
3813 defs and simplify handling of in-loop and across loop dead stores
3814 and properly fail for loop-variant refs. Handle byte-tracking with
3815 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3818 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3820 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3821 (vect_get_mask_type_for_stmt): Likewise.
3822 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3824 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3825 to determine the statement's vector type and the vector type that
3826 should be used for calculating nunits. Deal with cases in which
3827 the type has to be deferred.
3828 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3829 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3830 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3831 (vect_determine_vf_for_stmt): New functions, split out from...
3832 (vect_determine_vectorization_factor): ...here.
3833 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3834 (vect_get_mask_type_for_stmt): New functions, split out from
3835 vect_determine_vectorization_factor.
3837 2018-05-16 Richard Biener <rguenther@suse.de>
3839 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3840 verify the [VEC_]COND_EXPR embedded comparison.
3842 2018-05-15 Martin Sebor <msebor@redhat.com>
3844 PR tree-optimization/85753
3845 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3846 RECORD_TYPE in addition to ARRAY_TYPE.
3848 2018-05-15 Martin Sebor <msebor@redhat.com>
3851 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3853 2018-05-15 Richard Biener <rguenther@suse.de>
3855 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3856 add by_clobber_p one. Change algorithm to collect all defs
3857 representing uses we need to walk and try reducing them to
3858 a single one before failing.
3859 (dse_dom_walker::dse_optimize_stmt): Adjust.
3861 2018-05-13 Mark Wielaard <mark@klomp.org>
3863 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3864 (size_of_loc_descr): Likewise.
3865 (output_loc_operands): Likewise.
3866 (output_loc_operands_raw): Likewise.
3867 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3868 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3869 (hash_loc_operands): Likewise.
3870 (compare_loc_operands): Likewise.
3872 2018-05-14 Mark Wielaard <mark@klomp.org>
3874 * dwarf2out.c (count_index_addrs): New function.
3875 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3877 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3879 PR tree-optimization/83648
3880 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3881 return value as malloc candidate.
3883 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3886 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3887 param as true in call to suggest_attribute.
3889 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3891 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3892 -mreadonly-in-sdata.
3894 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3896 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3898 (aarch64_crypto_aesd_fused): Likewise.
3900 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3902 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3903 (movsi_aarch64): Likewise.
3904 (load_pairsi): Likewise.
3905 (load_pairdi): Likewise.
3906 (store_pairsi): Likewise.
3907 (store_pairdi): Likewise.
3908 (load_pairsf): Likewise.
3909 (load_pairdf): Likewise.
3910 (store_pairsf): Likewise.
3911 (store_pairdf): Likewise.
3912 (zero_extend): Likewise.
3913 (trunc): Swap alternatives.
3914 (fcvt_target): Add '?' to prefer w over r.
3916 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3919 * config/i386/i386.md: Disallow non-commutative arithmetics in
3920 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3921 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
3922 in the peephole2 before it.
3924 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
3926 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
3927 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
3928 (ix86_handle_option): Handle -mcldemote.
3929 * config.gcc: New header.
3930 * config/i386/cldemoteintrin.h: New file.
3931 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
3932 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3934 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3935 OPTION_MASK_ISA_CLDEMOTE.
3936 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
3937 (ix86_valid_target_attribute_inner_p): Ditto.
3938 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
3939 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
3940 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
3941 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
3942 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
3944 * config/i386/i386.opt: Add -mcldemote.
3945 * config/i386/x86intrin.h: New header.
3946 * doc/invoke.texi: Add -mcldemote.
3948 2018-05-14 Richard Biener <rguenther@suse.de>
3950 * doc/match-and-simplify.texi: Adjust :s documentation.
3952 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
3954 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
3955 intended memcpy size.
3956 (REORDER_45): Likewise.
3958 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
3960 * sort.cc: New file.
3961 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
3962 * vec.c (qsort_chk): Use gcc_qsort.
3963 * Makefile.in (OBJS-libcommon): Add sort.o.
3964 (build/sort.o): New target. Use it...
3965 (BUILD_RTL): ... here, and...
3966 (build/gencfn-macros): ... here, and...
3967 (build/genmatch): ... here.
3969 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3970 Chung-Ju Wu <jasonwucj@gmail.com>
3972 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
3973 * config/nds32/nds32-graywolf.md: New file.
3974 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
3975 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
3977 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
3978 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
3979 * config/nds32/nds32.md (pipeline_model): Add graywolf.
3980 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
3981 * config/nds32/pipelines.md: Include n15 settings.
3983 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3984 Chung-Ju Wu <jasonwucj@gmail.com>
3986 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
3987 * config/nds32/nds32-n13.md: New file.
3988 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
3989 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
3991 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
3992 * config/nds32/nds32.md (pipeline_model): Add n13.
3993 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
3994 * config/nds32/pipelines.md: Include n13 settings.
3996 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
3997 Chung-Ju Wu <jasonwucj@gmail.com>
3999 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4000 * config/nds32/nds32-n10.md: New file.
4001 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4002 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4004 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4005 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4006 * config/nds32/nds32.md (pipeline_model): Add n10.
4007 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4008 * config/nds32/pipelines.md: Include n10 settings.
4010 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4011 Kito Cheng <kito.cheng@gmail.com>
4012 Chung-Ju Wu <jasonwucj@gmail.com>
4014 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4015 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4016 Add enum values for DSP extension instructions.
4017 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4019 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4020 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4022 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4023 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4024 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4025 * config/nds32/nds32-intrinsic.md: Likewise.
4026 * config/nds32/nds32_intrinsic.h: Likewise.
4027 * config/nds32/nds32-md-auxiliary.c: Likewise.
4028 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4029 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4030 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4031 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4032 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4033 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4034 TYPE_DMAC in switch statement.
4035 * config/nds32/nds32.c: New checking and implementation for DSP
4036 extension instructions.
4037 * config/nds32/nds32.h: Likewise.
4038 * config/nds32/nds32.md: Likewise.
4039 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4040 * config/nds32/predicates.md: Implement new predicates for DSP
4043 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4045 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4046 Reformat alternatives and attributes so it is easier to identify
4047 which constraints/attributes go with which instruction.
4048 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4049 (mov<mode>_softfloat32, FMOVE64): Likewise.
4050 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4051 (mov<mode>_softfloat64, FMOVE64): Likewise.
4053 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4055 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4057 (Basic PowerPC Built-in Functions): The new name of the
4058 subsection previously known as "PowerPC Built-in Functions".
4059 (Basic PowerPC Built-in Functions Available on all Configurations):
4061 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4062 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4063 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4064 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4066 2018-05-11 Martin Jambor <mjambor@suse.cz>
4069 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4072 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4075 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4077 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4079 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4080 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4081 (ix86_handle_option): Handle -mwaitpkg.
4082 * config.gcc: New header.
4083 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4084 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4085 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4087 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4088 OPTION_MASK_ISA_WAITPKG.
4089 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4090 (ix86_option_override_internal): Add PTA_WAITPKG.
4091 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4092 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4093 IX86_BUILTIN_TPAUSE.
4094 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4095 __builtin_ia32_umwait and __builtin_ia32_tpause.
4096 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4097 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4098 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4099 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4100 UNSPECV_TPAUSE): New.
4101 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4102 * config/i386/i386.opt: Add -mwaitpkg.
4103 * config/i386/waitpkgintrin.h: New file.
4104 * config/i386/x86intrin.h: New header.
4105 * doc/invoke.texi: Add -mwaitpkg.
4107 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4110 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4112 (cortex-m0): Use armv6s-m isa.
4113 (cortex-m0plus): Likewise.
4114 (cortex-m1): Likewise.
4115 (cortex-m0.small-multiply): Likewise.
4116 (cortex-m0plus.small-multiply): Likewise.
4117 (cortex-m1.small-multiply): Likewise.
4119 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4120 Jakub Jelinek <jakub@redhat.com>
4122 PR tree-optimization/85692
4123 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4124 source permute as well.
4126 2018-05-11 Martin Liska <mliska@suse.cz>
4129 * doc/extend.texi: Document LLVM style format for no_sanitize
4132 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4134 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4135 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4136 (mode_supports_vsx_dform_quad): Likewise.
4137 (mode_supports_vmx_dform): Move these functions to be next to the
4138 other mode_supports functions.
4139 (mode_supports_dq_form): Likewise.
4140 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4141 mode_supports_dq_form.
4142 (reg_offset_addressing_ok_p): Likewise.
4143 (offsettable_ok_by_alignment): Likewise.
4144 (rs6000_legitimate_offset_address_p): Likewise.
4145 (legitimate_lo_sum_address_p): Likewise.
4146 (rs6000_legitimize_address): Likewise.
4147 (rs6000_legitimize_reload_address): Likewise.
4148 (rs6000_secondary_reload_inner): Likewise.
4149 (rs6000_preferred_reload_class): Likewise.
4150 (rs6000_output_move_128bit): Likewise.
4152 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4154 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4155 Generate SImode target register for null target.
4156 <case IX86_BUILTIN_XGETBV>: Ditto.
4157 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4158 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4160 2018-05-10 Carl Love <cel@us.ibm.com>
4162 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4163 dcbtt and dcbtstt if operands[2] is 0.
4165 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4168 * config/i386/sse.md (usadv64qi): New expander.
4170 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4172 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4173 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4174 -maltivec=be support.
4175 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4176 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4177 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4178 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4179 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4180 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4181 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4182 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4183 altivec_vsumsws): Adjust.
4184 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4185 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4186 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4188 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4189 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4190 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4191 (altivec_lve<VI_char>x): Delete expand.
4192 (*altivec_lve<VI_char>x_internal): Rename to...
4193 (altivec_lve<VI_char>x): ... this.
4194 (altivec_lvxl_<mode>): Delete expand.
4195 (*altivec_lvxl_<mode>_internal): Rename to ...
4196 (altivec_lvxl_<mode>): ... this.
4197 (altivec_stvxl_<mode>): Delete expand.
4198 (*altivec_stvxl_<mode>_internal): Rename to ...
4199 (altivec_stvxl_<mode>): ... this.
4200 (altivec_stve<VI_char>x): Delete expand.
4201 (*altivec_stve<VI_char>x_internal): Rename to ...
4202 (altivec_stve<VI_char>x): ... this.
4203 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4204 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4205 reduc_plus_scal_<mode>): Adjust.
4206 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4208 (rs6000_cpu_cpp_builtins): Adjust.
4209 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4210 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4211 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4212 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4213 -maltivec=be support.
4214 (rs6000_split_vec_extract_var): Adjust.
4215 (rs6000_split_v4si_init): Adjust.
4216 (swap_selector_for_mode): Delete.
4217 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4218 altivec_expand_stvex_be): Delete.
4219 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4220 -maltivec=be support.
4221 (rs6000_gimple_fold_builtin): Ditto.
4222 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4224 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4225 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4226 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4227 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4228 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4229 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4230 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4231 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4232 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4233 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4234 anonymous split): Adjust.
4235 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4236 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4238 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4240 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4241 when --with-gxx-include-dir is also specified.
4242 * configure: Regenerate.
4244 2018-05-09 Jim Wilson <jimw@sifive.com>
4247 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4248 * config/riscv/t-withmultilib: New.
4249 * config/riscv/withmultilib.h: New.
4250 * doc/install.texi: Document RISC-V --with-multilib-list support.
4252 2018-05-09 Richard Biener <rguenther@suse.de>
4254 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4256 (vect_bb_vectorization_profitable_p): Adjust. Compute
4257 actual scalar cost using the cost vector and the add_stmt_cost
4260 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4262 PR rtl-optimization/85645
4263 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4264 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4266 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4268 PR rtl-optimization/85645
4269 * shrink-wrap.c (spread_components): Return a boolean saying if
4270 anything was changed.
4271 (try_shrink_wrapping_separate): Iterate spread_components until
4272 nothing changes anymore.
4274 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4276 PR rtl-optimization/85645
4277 * regrename.c (build_def_use): Also kill the chains that include the
4278 destination of a REG_CFA_REGISTER note.
4280 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4282 PR rtl-optimization/85645
4283 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4284 insn that has a REG_CFA_REGISTER note.
4286 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4288 * cfgexpand.c (expand_clobber): New function.
4289 (expand_gimple_stmt_1): Use it.
4290 * tree-vect-stmts.c (vect_clobber_variable): New function,
4292 (vectorizable_simd_clone_call): ...here.
4293 (vectorizable_store): Emit a clobber either side of an
4294 IFN_STORE_LANES sequence.
4295 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4297 2018-05-09 Tom de Vries <tom@codesourcery.com>
4300 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4301 (define_insn "trap_if_false"): Add exit after trap.
4303 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4305 PR rtl-optimization/85638
4306 * bb-reorder.c: Include common/common-target.h.
4307 (create_forwarder_block): New function extracted from...
4308 (fix_up_crossing_landing_pad): ...here. Rename into...
4309 (dw2_fix_up_crossing_landing_pad): ...this.
4310 (sjlj_fix_up_crossing_landing_pad): New function.
4311 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4312 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4313 from both partitions and exit the loop after one iteration.
4315 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4318 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4320 (Basic PowerPC Built-in Functions): The new name of the
4321 subsection previously known as "PowerPC Built-in Functions".
4322 (Basic PowerPC Built-in Functions Available on all Configurations):
4324 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4326 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4328 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4330 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4333 2018-05-08 Jim Wilson <jimw@sifive.com>
4335 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4336 (LD_EMUL_SUFFIX): New.
4337 (LINK_SPEC): Use it.
4339 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4341 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4343 (Basic PowerPC Built-in Functions): The new name of the
4344 subsection previously known as "PowerPC Built-in Functions".
4345 (Basic PowerPC Built-in Functions Available on all Configurations):
4347 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4349 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4351 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4353 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4356 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4359 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4360 after cmpelim optimization.
4362 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4364 * config.gcc: Support "goldmont".
4365 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4366 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4368 * config/i386/i386.c (m_GOLDMONT): Define.
4369 (processor_target_table): Add "goldmont".
4370 (PTA_GOLDMONT): Define.
4371 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4372 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4373 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4374 (fold_builtin_cpu): Add "goldmont".
4375 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4376 (ix86_option_override_internal): Add "goldmont".
4377 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4378 (processor_type): Add PROCESSOR_GOLDMONT.
4379 * config/i386/i386.md: Add CPU "glm".
4380 * config/i386/glm.md: New file.
4381 * config/i386/x86-tune.def: Add m_GOLDMONT.
4382 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4384 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4387 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4389 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4390 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4391 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4394 * config/i386/i386.c (ix86_fold_builtin): Handle
4395 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4398 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4399 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4401 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4404 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4405 (check_arch): Likewise.
4406 (check_fpu): Return the result rather than printing it.
4407 (end arch): Fix operator precedence.
4408 (end cpu): Likewise.
4409 (END): Print the result from check_fpu.
4411 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4412 Alan Hayward <alan.hayward@arm.com>
4413 David Sherwood <david.sherwood@arm.com>
4415 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4416 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4417 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4418 (*fcmuo<mode>_and): New patterns.
4420 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4422 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4423 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4424 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4425 (cmp_op, sve_imm_con): New code attributes.
4426 (SVE_COND_INT_CMP, imm_con): Delete.
4427 (cmp_op): Remove above unspecs from int attribute.
4428 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4430 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4431 comparison-specific unspecs.
4432 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4433 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4434 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4435 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4436 (*vec_fcm<cmp_op><mode>): Rename to...
4437 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4438 (*vec_fcmuo<mode>): Rename to...
4439 (*fcmuo<mode>): ...this and adjust likewise.
4440 (*pred_fcm<cmp_op><mode>): New pattern.
4441 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4442 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4444 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4446 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4447 (aarch64_emit_sve_predicated_cond): New function.
4448 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4449 (aarch64_emit_unspec_cond_or): Replace with...
4450 (aarch64_emit_sve_or_conds): ...this new function. Use
4451 aarch64_emit_sve_ptrue_op for the individual comparisons and
4452 aarch64_emit_binop to OR them together.
4453 (aarch64_emit_inverted_unspec_cond): Replace with...
4454 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4455 aarch64_emit_sve_ptrue_op for the comparison and
4456 aarch64_emit_unop to invert the result.
4457 (aarch64_expand_sve_vec_cmp_float): Update after the above
4458 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4460 2018-05-07 Nathan Sidwell <nathan@acm.org>
4462 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4463 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4464 (Backwards Compatibility): Likewise.
4466 2018-05-07 Luis Machado <luis.machado@linaro.org>
4470 2018-05-07 Luis Machado <luis.machado@linaro.org>
4472 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4473 <prefetch_dynamic_strides>: New const bool field.
4474 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4475 prefetch_dynamic_strides.
4476 (exynosm1_prefetch_tune): Likewise.
4477 (thunderxt88_prefetch_tune): Likewise.
4478 (thunderx_prefetch_tune): Likewise.
4479 (thunderx2t99_prefetch_tune): Likewise.
4480 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4482 (aarch64_override_options_internal): Update to set
4483 PARAM_PREFETCH_DYNAMIC_STRIDES.
4484 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4485 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4486 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4487 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4488 prefetch-dynamic-strides setting.
4490 2018-05-07 Luis Machado <luis.machado@linaro.org>
4492 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4493 <minimum_stride>: New const int field.
4494 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4495 minimum_stride field.
4496 (exynosm1_prefetch_tune): Likewise.
4497 (thunderxt88_prefetch_tune): Likewise.
4498 (thunderx_prefetch_tune): Likewise.
4499 (thunderx2t99_prefetch_tune): Likewise.
4500 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4501 (aarch64_override_options_internal): Update to set
4502 PARAM_PREFETCH_MINIMUM_STRIDE.
4503 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4504 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4505 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4506 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4507 stride is constant and is below the minimum stride threshold.
4509 2018-05-07 Luis Machado <luis.machado@linaro.org>
4511 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4514 2018-05-07 Luis Machado <luis.machado@linaro.org>
4516 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4517 <prefetch_dynamic_strides>: New const bool field.
4518 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4519 prefetch_dynamic_strides.
4520 (exynosm1_prefetch_tune): Likewise.
4521 (thunderxt88_prefetch_tune): Likewise.
4522 (thunderx_prefetch_tune): Likewise.
4523 (thunderx2t99_prefetch_tune): Likewise.
4524 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4526 (aarch64_override_options_internal): Update to set
4527 PARAM_PREFETCH_DYNAMIC_STRIDES.
4528 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4529 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4530 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4531 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4532 prefetch-dynamic-strides setting.
4534 2018-05-07 Luis Machado <luis.machado@linaro.org>
4536 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4537 <minimum_stride>: New const int field.
4538 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4539 minimum_stride field.
4540 (exynosm1_prefetch_tune): Likewise.
4541 (thunderxt88_prefetch_tune): Likewise.
4542 (thunderx_prefetch_tune): Likewise.
4543 (thunderx2t99_prefetch_tune): Likewise.
4544 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4545 (aarch64_override_options_internal): Update to set
4546 PARAM_PREFETCH_MINIMUM_STRIDE.
4547 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4548 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4549 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4550 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4551 stride is constant and is below the minimum stride threshold.
4553 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4556 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4557 the type is addressable. Don't force op into register if it has
4560 2018-05-05 Roland McGrath <mcgrathr@google.com>
4563 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4564 any section for which we don't know a specific type it should have,
4565 regardless of name. Previously this was done only for the exact
4566 names ".init_array", ".fini_array", and ".preinit_array".
4567 (default_elf_asm_named_section): Add comment about
4568 relationship with default_section_type_flags and SECTION_NOTYPE.
4569 (get_section): Don't consider it a type conflict if one side has
4570 SECTION_NOTYPE and the other doesn't, as long as neither has the
4571 SECTION_BSS et al used in the default_section_type_flags logic.
4573 2018-05-05 Tom de Vries <tom@codesourcery.com>
4576 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4577 (workaround_barsyncs): New function.
4578 (nvptx_reorg): Use workaround_barsyncs.
4579 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4580 (define_expand "nvptx_membar_cta"): New define_expand.
4581 (define_insn "*nvptx_membar_cta"): New insn.
4583 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4585 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4586 To improve optimization opportunities.
4587 * builtin-types.def: The new needed builtin types for the above.
4589 2018-05-04 Richard Biener <rguenther@suse.de>
4591 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4592 * gimple-ssa-store-merging.c
4593 (imm_store_chain_info::output_merged_store): Remove redundant create,
4594 release split_store vector contents on failure.
4595 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4596 scalar stmt vector on cache hit.
4598 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4600 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4602 * config.gcc (powerpc-xilinx-eabi*): Remove.
4603 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4605 (fusion_addis_mem_combo_load): Ditto.
4606 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4608 (rs6000_cpu_cpp_builtins): Ditto.
4609 * config/rs6000/rs6000-linux.c
4610 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4611 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4612 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4614 (rs6000_setup_reg_addr_masks): Ditto.
4615 (rs6000_init_hard_regno_mode_ok): Ditto.
4616 (rs6000_option_override_internal): Ditto.
4617 (legitimate_lo_sum_address_p): Ditto.
4618 (rs6000_legitimize_address): Ditto.
4619 (rs6000_legitimize_reload_address): Ditto.
4620 (rs6000_legitimate_address_p): Ditto.
4621 (abi_v4_pass_in_fpr): Ditto.
4622 (setup_incoming_varargs): Ditto.
4623 (rs6000_gimplify_va_arg): Ditto.
4624 (rs6000_split_multireg_move): Ditto.
4625 (rs6000_savres_strategy): Ditto.
4626 (rs6000_emit_prologue_components): Ditto.
4627 (rs6000_emit_epilogue_components): Ditto.
4628 (rs6000_emit_prologue): Ditto.
4629 (rs6000_emit_epilogue): Ditto.
4630 (rs6000_elf_file_end): Ditto.
4631 (rs6000_function_value): Ditto.
4632 (rs6000_libcall_value): Ditto.
4633 * config/rs6000/rs6000.h: Ditto.
4634 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4635 (TARGET_MINMAX): ... this. New.
4636 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4637 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4638 (*movsi_internal1_single): Delete.
4639 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4640 mfpu=, mxilinx-fpu): Delete.
4641 * config/rs6000/singlefp.h: Delete.
4642 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4643 * config/rs6000/t-rs6000: Ditto.
4644 * config/rs6000/t-xilinx: Delete.
4645 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4646 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4647 (VStype_simple): Delete.
4648 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4649 * config/rs6000/xfpu.h: Delete.
4650 * config/rs6000/xfpu.md: Delete.
4651 * config/rs6000/xilinx.h: Delete.
4652 * config/rs6000/xilinx.opt: Delete.
4653 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4654 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4656 2018-05-04 Tom de Vries <tom@codesourcery.com>
4659 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4662 2018-05-04 Richard Biener <rguenther@suse.de>
4665 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4666 (expand_complex_div_wide): Likewise.
4667 (expand_complex_operations_1): Likewise.
4668 (expand_complex_libcall): Preserve EH info of the original stmt.
4669 (tree_lower_complex): Handle removed blocks.
4670 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4671 on complex multiplication and division libcall builtins.
4673 2018-05-04 Richard Biener <rguenther@suse.de>
4676 * fold-const.c (negate_expr_p): Restrict negation of operand
4677 zero of a division to when we know that can happen without
4679 (fold_negate_expr_1): Likewise.
4681 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4684 * real.h (real_nextafter): Declare.
4685 * real.c (real_nextafter): New function.
4686 * fold-const-call.c (fold_const_nextafter): New function.
4687 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4688 CASE_CFN_NEXTTOWARD.
4689 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4690 even when arg1_mode is different from arg0_mode.
4692 2018-05-03 Nathan Sidwell <nathan@acm.org>
4694 * doc/extend.texi (Deprecated Features): Remove
4696 (Backwards Compatibility): Likewise.
4697 * doc/invoke.texi (C++ Language Options): Likewise.
4698 (C++ Dialect Options): Likewise.
4700 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4703 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4704 _mm512_mask_mullox_epi64): New intrinsics.
4706 2018-05-03 Tom de Vries <tom@codesourcery.com>
4709 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4710 dump files): Add offload-tree.
4712 2018-05-03 Richard Biener <rguenther@suse.de>
4714 PR tree-optimization/85615
4715 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4716 to loops not nested in BBs loop father to avoid creating multi-entry
4719 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4721 PR tree-optimization/70291
4722 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4723 arguments. Change return type to tree. Emit libcall as a new
4724 statement rather than replacing existing one when inplace_p is true.
4725 (expand_complex_multiplication_components): New function.
4726 (expand_complex_multiplication): Expand floating-point complex
4727 multiplication using the above.
4728 (expand_complex_division): Rename inner_type parameter to type.
4729 Update expand_complex_libcall call-site.
4730 (expand_complex_operations_1): Update expand_complex_multiplication
4731 and expand_complex_division call-sites.
4733 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4736 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4737 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4738 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4739 the highest significant bit of the shift count mask is clear. In
4740 check whether and[sq]i3 is needed verify that all significant bits
4741 of the shift count other than the highest are set.
4743 2018-05-02 Tom de Vries <tom@codesourcery.com>
4746 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4747 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4748 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4749 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4750 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4751 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4752 __builtin_goacc_parlevel_size.
4754 2018-05-02 Richard Biener <rguenther@suse.de>
4756 PR tree-optimization/85597
4757 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4758 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4761 2018-05-02 Tom de Vries <tom@codesourcery.com>
4764 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4765 dump files): Add ltrans-tree.
4767 2018-05-02 Tom de Vries <tom@codesourcery.com>
4770 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4771 dump files): Add wpa-ipa.
4773 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4775 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4776 powerpc*-*-linux*paired* target.
4777 * config/rs6000/750cl.h: Delete.
4778 * config/rs6000/paired.h: Delete.
4779 * config/rs6000/paired.md: Delete.
4780 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4782 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4783 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4784 comment. Remove paired float support.
4785 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4786 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4788 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4789 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4791 * config/rs6000/rs6000.c: Remove paired float support.
4792 (paired_expand_vector_init, paired_expand_vector_move,
4793 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4794 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4795 paired_expand_builtin, paired_expand_predicate_builtin,
4796 paired_init_builtins): Delete.
4797 * config/rs6000/rs6000.h: Remove paired float support.
4798 * config/rs6000/rs6000.md: Remove paired float support.
4799 (move_from_CR_ov_bit): Delete.
4800 * config/rs6000/rs6000.opt (mpaired): Delete.
4801 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4802 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4804 2018-05-02 Richard Biener <rguenther@suse.de>
4807 * gimplify.c (gimplify_save_expr): When in SSA form allow
4808 SAVE_EXPRs to compute to SSA vars.
4810 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4813 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4814 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4815 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4816 clobber operands[2], instead use a new pseudo. Formatting fixes.
4818 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4820 PR tree-optimization/85586
4821 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4822 exit early for statements in the same group if the accesses are
4825 2018-05-02 Tom de Vries <tom@codesourcery.com>
4828 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4831 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4833 PR tree-optimization/85143
4834 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4836 2018-05-01 Tom de Vries <tom@codesourcery.com>
4839 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4840 not found" error message.
4842 2018-05-01 Tom de Vries <tom@codesourcery.com>
4845 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4846 * vec.c (test_ordered_remove_if): New function.
4847 (vec_c_tests): Call test_ordered_remove_if.
4848 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4849 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4850 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4851 VEC_ORDERED_REMOVE_IF.
4853 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4855 PR tree-optimization/82665
4856 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4857 pointer subtraction where arguments come from a memchr call.
4859 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4861 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4862 --push-state --as-needed and --pop-state instead of --as-needed and
4863 --no-as-needed if ld supports it.
4864 * configure: Regenerated.
4867 * doc/install.texi2html: Replace _002d with - and _002a with * in
4868 generated html files using sed.
4870 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4873 * gcc-rich-location.c (blank_line_before_p): New function.
4874 (use_new_line): New function.
4875 (gcc_rich_location::add_fixit_insert_formatted): New function.
4876 * gcc-rich-location.h
4877 (gcc_rich_location::add_fixit_insert_formatted): New function.
4879 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4881 * selftest.c (assert_streq): Rename "expected" and "actual" to
4882 "val1" and "val2". Extend NULL-handling to cover both inputs
4883 symmetrically, while still requiring both to be non-NULL for a pass.
4884 * selftest.h (assert_streq): Rename "expected" and "actual" to
4886 (ASSERT_EQ): Likewise.
4887 (ASSERT_EQ_AT): Likewise.
4888 (ASSERT_KNOWN_EQ): Likewise.
4889 (ASSERT_KNOWN_EQ_AT): Likewise.
4890 (ASSERT_NE): Likewise.
4891 (ASSERT_MAYBE_NE): Likewise.
4892 (ASSERT_MAYBE_NE_AT): Likewise.
4893 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4894 the assertion to pass.
4895 (ASSERT_STREQ_AT): Likewise.
4897 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4899 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4900 interaction with -pie.
4902 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4904 * selftest.h: Fix alphabetization of per-source-file selftest
4907 2018-04-30 Jason Merrill <jason@redhat.com>
4909 PR c++/61982 - dead stores to destroyed objects.
4910 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4913 2018-04-30 Jason Merrill <jason@redhat.com>
4915 * tree.c (build_clobber): New.
4916 * tree.h: Declare it.
4917 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4919 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4921 * diagnostic-show-locus.c (layout::layout): Update for
4922 location_get_source_line returning a char_span.
4923 (struct char_span): Move to input.h.
4924 (struct correction): Update for fields in char_span becoming
4926 (struct source_line): Update for location_get_source_line
4927 returning a char_span.
4928 (layout::print_line): Likewise.
4929 * edit-context.c (edited_file::print_content): Likewise.
4930 (edited_file::print_diff_hunk): Likewise.
4931 (edited_file::print_run_of_changed_lines): Likewise.
4932 (edited_file::get_num_lines): Likewise.
4933 (edited_line::edited_line): Likewise.
4934 * final.c (asm_show_source): Likewise.
4935 * input.c (location_get_source_line): Convert return type
4936 from const char * to char_span, losing the final "line_len"
4938 (dump_location_info): Update for the above.
4939 (get_substring_ranges_for_loc): Likewise. Use a char_span
4940 when handling the literal within the line.
4941 (test_reading_source_line): Update for location_get_source_line
4942 returning a char_span.
4943 * input.h (class char_span): Move here from
4944 diagnostic-show-locus.c, converting from a struct to a class.
4945 Make data members private.
4946 (char_span::operator bool): New.
4947 (char_span::length): New.
4948 (char_span::get_buffer): New.
4949 (char_span::operator[]): New.
4950 (char_span::subspan): Make const.
4951 (char_span::xstrdup): New.
4952 (location_get_source_line): Convert return type from const char *
4953 to char_span, losing the final "line_size" param.
4955 2018-04-30 Jan Hubicka <jh@suse.cz>
4957 * lto-wrapper.c (ltrans_priorities): New static var.
4958 (cmp_priority): New.
4959 (run_gcc): Read priorities and if doing parallel build order
4960 the Makefile by them.
4962 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4964 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
4966 2018-04-30 Richard Biener <rguenther@suse.de>
4968 * tree-cfg.c (verify_address): Remove base argument, add
4969 flag whether to check TREE_ADDRESSABLE and do that.
4970 (verify_expr): Remove.
4971 (verify_types_in_gimple_reference): Add pieces from verify_expr.
4972 (verify_gimple_assign_single): Likewise.
4973 (verify_gimple_switch): Likewise.
4974 (verify_expr_location_1): Dereference tp once. Add (disabled)
4975 piece from verify_expr.
4976 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
4978 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4980 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
4982 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
4984 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
4985 (small_data_pattern): Likewise.
4986 (arc_rewrite_small_data): Likewise.
4987 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
4988 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
4989 (get_symbol_alignment): New function.
4990 (legitimate_small_data_address_p): Likewise.
4991 (legitimate_scaled_address): Update, call
4992 legitimate_small_data_address_p.
4993 (output_sdata): New static variable.
4994 (arc_print_operand): Update how we handle small data operands.
4995 (arc_print_operand_address): Likewise.
4996 (arc_legitimate_address_p): Update, use
4997 legitimate_small_data_address_p.
4998 (arc_rewrite_small_data_p): Remove.
4999 (arc_rewrite_small_data_1): Likewise.
5000 (arc_rewrite_small_data): Likewise.
5001 (small_data_pattern): Likewise.
5002 (compact_sda_memory_operand): Update to use
5003 legitimate_small_data_address_p and get_symbol_alignment.
5004 (prepare_move_operands): Don't rewite sdata pattern.
5005 (prepare_extend_operands): Remove.
5006 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5008 (zero_extendqisi2): Likewise.
5009 (zero_extendhisi2): Likewise.
5010 (extendqihi2): Likewise.
5011 (extendqisi2): Likewise.
5012 (extendhisi2): Likewise.
5016 * config/arc/constraints.md (Usd): Change it to memory constraint.
5018 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5020 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5021 as source of std instructions.
5022 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5023 allow 6-bit constants as source for store instructions.
5024 (movdi_insn): Update instruction pattern to allow 6-bit constants
5025 as source for store instructions.
5027 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5029 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5031 2018-04-30 Nathan Sidwell <nathan@acm.org>
5032 Sandra Loosemore <sandra@codesourcery.com>
5034 * dumpfile.c (dump_open): Allow '-' for stdout.
5035 * doc/invoke.texi (Developer Options): Document dump filename
5036 determination early. Document stdin/stdout selection.
5038 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5040 Microblaze Target: PIC data text relative
5042 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5043 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5045 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5046 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5047 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5048 New addressing mode for data-text relative position indepenedent code.
5049 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5050 'ADDRESS_SYMBOLIC_TXT_REL'.
5051 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5052 (microblaze_legitimate_pic_operand): Exclude function calls from
5053 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5054 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5056 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5057 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5058 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5059 for 'address + offset'.
5060 (microblaze_expand_prologue): Add new function prologue call for
5062 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5063 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5064 table in case of TARGET_PIC_DATA_TEXT_REL.
5065 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5066 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5067 Add new macros 'UNSPEC_TEXT',
5068 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5069 + exclude function calls from 'UNSPEC_PLT' in case of data text
5071 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5072 new target hook for generating address diff vector tables in case of
5074 * doc/tm.texi : Regenerate.
5075 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5076 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5077 of addr diff vector generation.
5078 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5079 target hook definition.
5080 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5081 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5082 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5083 option for data text relative.
5085 2018-04-30 Richard Biener <rguenther@suse.de>
5087 * tree-chrec.h (evolution_function_is_constant_p): Remove
5089 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5091 2018-04-30 Richard Biener <rguenther@suse.de>
5094 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5096 2018-04-30 Richard Biener <rguenther@suse.de>
5098 PR tree-optimization/28364
5099 PR tree-optimization/85275
5100 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5101 copying first exit test.
5103 2018-04-28 Mark Wielaard <mark@klomp.org>
5105 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5107 (dwarf_AT): Handle DW_AT_addr_base.
5108 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5110 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5113 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5114 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5115 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5116 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5118 2018-04-28 Richard Biener <rguenther@suse.de>
5120 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5121 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5122 to reflect use. Only add interesting stmts.
5124 2018-04-27 Martin Jambor <mjambor@suse.cz>
5127 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5128 the jump function allows for passing through aggregate values.
5130 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5132 * input.h (in_system_header_at): Convert from macro to inline
5134 (from_macro_expansion_at): Likewise.
5135 (from_macro_definition_at): Likewise.
5137 2018-04-27 Jeff Law <law@redhat.com>
5139 * config.gcc: Mark tile* targets as deprecated/obsolete.
5141 2018-04-27 Richard Biener <rguenther@suse.de>
5143 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5146 2018-04-27 Richard Biener <rguenther@suse.de>
5148 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5150 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5152 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5153 with Yd constraint. Set "preferred_for_speed" attribute from
5154 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5156 (*movdi_internal): Ditto.
5157 (movti_interunit splitters): Remove
5158 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5159 (movdi_interunit splitters): Ditto.
5160 * config/i386/constraints.md (Ye): Remove.
5161 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5163 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5166 * config/aarch64/constraints.md (Usg): Limit to 31.
5169 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5171 PR tree-optimization/85529
5172 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5173 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5174 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5175 zero extension or masking of the MSB bit.
5176 (optimize_range_tests): Add FIRST_BB argument, pass it through
5177 to optimize_range_tests_var_bound.
5178 (maybe_optimize_range_tests, reassociate_bb): Adjust
5179 optimize_range_tests callers.
5181 2018-04-26 Richard Biener <rguenther@suse.de>
5182 Jakub Jelinek <jakub@redhat.com>
5184 * cgraph.h (symbol_table): Just declare debug method here.
5185 * symtab.c (symbol_table::debug): Define.
5187 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5189 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5191 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5193 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5194 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5195 (*movdi_internal): Substitute Yi and Yj constraint with x
5196 and Ym and Yn constraint with y constraint. Update "isa"
5197 attribute and set "preferred_for_speed" attribute from
5198 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5199 (*movsi_internal): Ditto.
5200 (*movdf_internal): Ditto.
5201 (*movsf_internal): Ditto.
5202 (*zero_extendsidi2): Ditto.
5203 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5204 (sse2_loadld): Ditto.
5205 (*vec_extract<ssevecmodelower>_0): Ditto.
5206 (*vec_extractv4si_0_zext_sse4): Ditto.
5207 (vec_concatv2di): Ditto.
5208 (*vec_dup<mode>): Ditto.
5209 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5210 * config/i386/constraints.md (Yi): Remove.
5215 2018-04-26 Nathan Sidwell <nathan@acm.org>
5217 * dumpfile.c (dump_open): New.
5218 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5219 (dump_finish): Detect stdio/stderr by value not name.
5221 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5223 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5225 2018-04-26 Tom de Vries <tom@codesourcery.com>
5228 * config/nvptx/nvptx.c (verify_neutering_jumps)
5229 (verify_neutering_labels): New function
5230 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5232 2018-04-26 Tom de Vries <tom@codesourcery.com>
5235 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5236 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5239 2018-04-26 Richard Biener <rguenther@suse.de>
5240 Tom de Vries <tom@codesourcery.com>
5243 * lto-streamer-out.c (output_function): Fixup loops if required to match
5244 discovery done in the reader.
5246 2018-04-26 Richard Biener <rguenther@suse.de>
5248 PR tree-optimization/85116
5249 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5250 have a loop exit from the single latch predecessor. Remove
5251 case of header with just condition.
5252 (ch_base::copy_headers): Exclude infinite loops from any
5254 (pass_ch::execute): Record exits.
5256 2018-04-26 Richard Biener <rguenther@suse.de>
5258 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5259 prologue cost vector and pass it to vect_get_load_cost.
5260 (vect_get_peeling_costs_all_drs): Likewise.
5261 (vect_peeling_hash_get_lowest_cost): Likewise.
5262 (vect_enhance_data_refs_alignment): Likewise.
5264 2018-04-26 Richard Biener <rguenther@suse.de>
5267 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5268 checking of integer<->pointer conversions.
5269 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5270 sign-/zero-extending pointer types.
5271 (expand_omp_for_static_chunk): Likewise.
5273 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5274 Jean Lee <xiaoyur347@gmail.com>
5276 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5277 (TARGET_ASAN_SHADOW_OFFSET): Define.
5278 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5279 true for -fsanitize=address.
5281 2018-04-25 Mark Wielaard <mark@klomp.org>
5283 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5286 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5288 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5289 than "alu", remove explicit "memory" and "imm_disp" attributes.
5290 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5293 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5294 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5295 gen_lowpart_no_emit.
5297 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5300 * config/i386/i386.c (ix86_expand_builtin): Change memory
5301 operand to XI, extend p0 to Pmode.
5302 * config/i386/i386.md: Change unspec volatile and operand
5303 1 mode to XI, change operand 0 mode to P.
5305 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5307 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5308 GET_MODE_MASK before any checking.
5309 (nds32_can_use_bset_p): Likewise.
5310 (nds32_can_use_btgl_p): Likewise.
5312 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5314 * config/nds32/nds32-doubleword.md: New define_split pattern for
5315 illegal register number.
5317 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5319 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5321 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5323 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5325 2018-04-25 Richard Biener <rguenther@suse.de>
5327 * lto-streamer.h (LTO_major_version): Bump to 8.
5329 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5331 * BASE-VER: Set to 9.0.0.
5333 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5335 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5336 in __abskf2 and __powikf2.
5338 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5341 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5342 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5343 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5344 Use the above on operand 2. Reindent.
5345 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5347 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5350 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5352 * config/i386/i386.opt (mcet): Removed.
5353 * doc/install.texi: Remove -mcet documentation.
5354 * doc/invoke.texi: Likewise.
5356 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5359 * doc/install.texi: Remove -mcet from bootstrap-cet.
5361 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5364 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5365 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5369 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5370 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5371 containing a CONST_VECTOR.
5373 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5375 * doc/install.texi: Update newlib dependency for nvptx.
5377 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5380 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5381 instead of INTVAL when shifting x left.
5383 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5385 PR tree-optimization/85478
5386 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5387 vect_grouped_store_supported for single element vectors.
5389 2018-04-24 Richard Biener <rguenther@suse.de>
5392 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5393 load cost increase to the case of non-constant step.
5395 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5398 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5399 destination if any_malformed_asm.
5401 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5404 * expr.c (store_field): In the bitfield case, if the value comes from
5405 a function call and is returned in registers by means of a PARALLEL,
5406 do not change the mode of the temporary unless BLKmode and VOIDmode.
5408 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5410 PR rtl-optimization/85423
5411 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5412 dependencies to debug insns when the previous insn is non-debug.
5414 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5416 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5417 enums into a single definition.
5418 (fls): Fix predicates and printing.
5421 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5423 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5424 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5425 and short u6 immediate.
5426 (check_if_valid_sleep_operand): Remove.
5427 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5429 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5431 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5432 flag_always_save_lp condition.
5433 * config/nds32/nds32.opt (malways-save-lp): New option.
5435 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5437 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5438 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5439 * config/nds32/nds32.h
5440 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5441 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5443 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5445 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5446 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5448 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5449 Chung-Ju Wu <jasonwucj@gmail.com>
5451 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5453 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5454 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5456 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5458 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5460 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5462 * config/nds32/nds32-protos.h (nds32_data_alignment,
5463 nds32_local_alignment): Declare.
5464 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5465 nds32_local_alignment): New functions.
5466 (TARGET_CONSTANT_ALIGNMENT): Define.
5467 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5469 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5471 * config/nds32/nds32.c
5472 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5473 (TARGET_MODES_TIEABLE_P): Likewise.
5475 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5477 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5480 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5481 Chung-Ju Wu <jasonwucj@gmail.com>
5483 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5484 for unaligned access.
5485 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5487 * config/nds32/nds32-intrinsic.md: Likewise.
5488 * config/nds32/nds32_intrinsic.h: Likewise.
5489 * config/nds32/nds32.h (nds32_builtins): Likewise.
5490 * config/nds32/nds32.opt (munaligned-access): New option.
5491 * config/nds32/nds32.c (nds32_asm_file_start): Display
5492 flag_unaligned_access status.
5494 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5496 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5497 -mno-relax is present.
5498 * config/riscv/linux.h (LINK_SPEC): Ditto.
5500 2018-04-20 Martin Sebor <msebor@redhat.com>
5503 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5505 (gimple_fold_builtin_stxcpy_chk): Same.
5506 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5508 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5511 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5512 __powikf2 when long double is IEEE 128-bit.
5514 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5516 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5517 step to make sure stack always aligned.
5519 2018-04-20 Carl Love <cel@us.ibm.com>
5522 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5523 size check for arg0.
5525 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5526 Tom de Vries <tom@codesourcery.com>
5529 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5530 Emit insns for calls too.
5531 (nvptx_find_par): Always look for worker-level predecessor insn.
5532 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5534 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5535 (nvptx_process_pars): Propagate frames for calls.
5537 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5540 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5542 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5543 (ix86_handle_option): Don't handle OPT_mibt.
5544 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5546 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5548 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5549 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5550 (ix86_target_macros): Define __CET__ with flag_cf_protection
5551 for -fcf-protection.
5552 * config/i386/i386.c (isa2_opts): Remove -mibt.
5553 * config/i386/i386.h (TARGET_IBT): Removed.
5554 (TARGET_IBT_P): Likewise.
5555 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5556 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5557 * config/i386/i386.opt (mcet): Update help message.
5560 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5561 -mcet as an alias for -mshstk.
5563 2018-04-20 Richard Biener <rguenther@suse.de>
5566 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5567 complexity by forcing a single use of the multiply operand.
5569 2018-04-20 Martin Jambor <mjambor@suse.cz>
5572 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5573 recursion dependency to only apply to non-clones.
5575 2018-04-20 Martin Jambor <mjambor@suse.cz>
5578 * ipa-cp.c (create_specialized_node): Check that clones of
5579 self-recursive edges exist during IPA-CP.
5581 2018-04-19 Toon Moene <toon@moene.org>
5583 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5586 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5588 PR tree-optimization/85467
5589 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5590 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5591 VECTOR_CST element to type.
5593 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5596 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5597 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5598 (builtin_longjmp): Likewise.
5599 (save_stack_nonlocal): New pattern.
5600 (restore_stack_nonlocal): Likewise.
5602 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5605 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5606 Replace ASM_OUTPUT_LABEL with fprintf.
5608 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5611 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5612 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5613 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5614 define __IBT__ and __SHSTK__ for -fcf-protection.
5615 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5617 (ix86_trampoline_init): Likewise.
5618 (x86_output_mi_thunk): Likewise.
5619 (ix86_notrack_prefixed_insn_p): Likewise.
5620 (ix86_option_override_internal): Don't disallow -fcf-protection.
5621 * config/i386/i386.md (rdssp<mode>): Also enable for
5623 (incssp<mode>): Likewise.
5624 (nop_endbr): Likewise.
5625 * config/i386/i386.opt (mcet): Change help message to built-in
5629 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5630 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5631 enable CET built-in functions.
5633 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5635 * common/config/i386/i386-common.c
5636 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5637 OPTION_MASK_ISA_MOVDIRI_UNSET,
5638 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5639 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5640 * config.gcc (movdirintrin.h): New header.
5641 * config/i386/cpuid.h (bit_MOVDIRI,
5642 bit_MOVDIR64B): New bits.
5643 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5645 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5646 (VOID, PVOID, PCVOID)): New function types.
5647 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5648 __builtin_ia32_directstoreu_u64,
5649 __builtin_ia32_movdir64b): New builtins.
5650 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5651 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5653 (ix86_valid_target_attribute_inner_p): Ditto.
5654 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5655 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5656 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5657 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5658 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5659 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5660 (movdiri<mode>, movdir64b_<mode>): New.
5661 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5662 * config/i386/immintrin.h: Include movdirintrin.h.
5663 * config/i386/movdirintrin.h: New file.
5664 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5666 2018-04-19 Richard Biener <rguenther@suse.de>
5669 * cfg.c (clear_bb_flags): When loop state says we have
5670 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5672 2018-04-19 Richard Biener <rguenther@suse.de>
5674 PR tree-optimization/84737
5675 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5676 copying restrict info.
5677 (vect_setup_realignment): Use it.
5678 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5679 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5680 the first DR to all generated stores.
5681 (vectorizable_load): Likewise for loads.
5683 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5685 PR tree-optimization/85446
5686 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5687 the integral and pointer types to have the same precision.
5689 * doc/install.texi: Document --disable-cet being the default and
5692 2018-04-18 Martin Liska <mliska@suse.cz>
5694 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5697 2018-04-18 Martin Liska <mliska@suse.cz>
5700 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5703 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5704 arguments if they are comparable.
5706 2018-04-18 Martin Liska <mliska@suse.cz>
5709 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5712 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5715 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5718 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5719 ENDBR after calling __morestack.
5721 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5724 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5725 by using gcc_base_ver to generate a gcc_driver_version, and use
5726 it when generating GCC_DRIVER_NAME.
5727 * configure: Regenerate.
5729 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5732 * config.gcc: Obsolete powerpc*-*-*spe*.
5734 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5737 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5738 (stabstr_D): Change type of unum from unsigned int to
5739 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5742 2018-04-17 Jim Wilson <jimw@sifive.com>
5745 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5746 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5747 Set arg_pointer_offset after using pretend_args_size.
5749 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5751 PR rtl-optimization/85431
5752 * dse.c (record_store): Ignore zero width stores.
5755 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5756 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5757 __builtin_stack_restore rather than after it.
5758 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5759 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5760 argument instead of virtual_dynamic_stack_rtx.
5762 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5764 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5766 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5767 Add note to error message to explain internal mapping of overloaded
5768 built-in function name to non-overloaded built-in function name.
5769 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5772 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5775 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5776 where the inputs overlap with the output.
5778 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5780 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5781 (=v, v) alternative and explicit "memory" attribute.
5782 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5783 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5785 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5786 "sselog1" type instead of "sselog".
5787 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5788 "sselog". Remove explicit "memory" attribute.
5789 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5790 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5792 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5793 "sselog1" type instead of "sselog". Remove explicit "memory"
5795 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5796 use "sselog1" type instead of "sselog". Remove explicit "memory"
5798 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5799 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5801 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5802 "sselog1" type instead of "sselog". Remove explicit "memory"
5804 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5805 use "sselog1" type instead of "sselog". Remove explicit "memory"
5809 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5812 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5815 2018-04-17 Martin Jambor <mjambor@suse.cz>
5818 * ipa-cp.c (create_specialized_node): Call
5819 expand_all_artificial_thunks if necessary.
5821 2018-04-17 Martin Liska <mliska@suse.cz>
5824 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5825 in message, remote space in between '_G' and '('.
5827 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5830 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5831 avx512f_vmcmp<mode>3<round_saeonly_name>,
5832 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5833 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5834 avx512f_rndscale<mode><round_saeonly_name>,
5835 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5836 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5837 Use %<iptr>2 instead of %2 for -masm=intel.
5838 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5839 avx512f_vcvttss2usi<round_saeonly_name>,
5840 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5842 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5843 avx512f_vcvttsd2usi<round_saeonly_name>,
5844 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5845 Use %q1 instead of %1 for -masm=intel.
5846 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5847 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5848 of %3 for -masm=intel.
5849 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5851 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5853 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5855 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5856 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5858 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5860 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5861 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5862 %g1 and one with %0 and %1.
5863 (avx512er_vmrcp28<mode><round_saeonly_name>,
5864 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5866 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5867 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5868 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5869 of %0 and %{%4%} for -masm=intel.
5870 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5871 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5872 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5873 order of %0 and %{%5%}%{z%} for -masm=intel.
5875 2018-04-17 Jan Hubicka <jh@suse.cz>
5878 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5880 2018-04-17 Martin Liska <mliska@suse.cz>
5883 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5884 for target_clone error message. Make default implementation
5885 clone to be a local declaration.
5886 (separate_attrs): Add new argument and check for an empty
5888 (expand_target_clones): Handle it.
5889 (ipa_target_clone): Make redirection just for target_clones
5892 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5893 Tom de Vries <tom@codesourcery.com>
5896 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5897 tiled basic blocks without omp continue statements.
5899 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5902 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5903 vec_extract expression as having side effects to make sure it gets
5906 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5909 * config/i386/i386.c (get_builtin_code_for_version): Check
5912 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5915 * gcc/config.gcc: Support "skylake".
5916 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5918 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5919 (processor_target_table): Add "skylake".
5920 (ix86_option_override_internal): Add "skylake".
5921 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
5922 PROCESSOR_CANNONLAKE.
5923 (get_builtin_code_for_version): Fix priority for
5924 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
5925 PROCESSOR_SKYLAKE-AVX512.
5926 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
5927 (processor_type): Add PROCESSOR_SKYLAKE.
5929 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
5930 Jason Merrill <jason@redhat.com>
5933 * convert.c (convert_to_integer_1): Use direct recursion for
5934 enumeral types and types with a precision less than the number
5935 of bits in their mode.
5937 2018-04-16 Julia Koval <julia.koval@intel.com>
5940 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
5941 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
5943 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
5946 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5947 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
5948 and -mno-direct-move.
5950 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
5953 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
5954 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
5955 Ensure negative shifts result in {0}.
5957 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
5959 PR rtl-optimization/79916
5960 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
5961 regs (if any) to define how to gnerate SD moves when LRA is in
5964 2018-04-13 Jakub Jelinek <jakub@redhat.com>
5966 PR rtl-optimization/85393
5967 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
5968 * except.c (expand_dw2_landing_pad_for_region): Make static.
5969 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
5970 a label and unconditional jump to old_bb, rather than
5971 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
5974 PR rtl-optimization/85376
5975 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
5976 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
5977 instead of a specific value.
5979 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5980 Bin Cheng <bin.cheng@arm.com>
5982 PR tree-optimization/82965
5983 PR tree-optimization/83991
5984 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
5985 by_profile_only parameter.
5986 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
5987 information if the loop was predicted to iterate too many times.
5988 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
5990 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
5993 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
5996 2018-04-13 Martin Liska <mliska@suse.cz>
5997 Jakub Jelinek <jakub@redhat.com>
6000 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6001 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6002 * builtins.c (expand_builtin_memory_copy_args): Use
6003 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6004 handle dest_addr == pc_rtx.
6006 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6009 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6010 asked to not generate direct moves.
6011 (fix_trunc<mode>si2_stfiwx): Similar.
6012 (fix_trunc<mode>si2_internal): Similar.
6014 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6017 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6018 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6019 lookup if dest in some wider mode is known to be const0_rtx and
6020 if so, record permanent equivalence for it to be ZERO_EXTEND of
6021 the narrower mode destination.
6023 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6025 * lto-streamer-out.c (output_function): Revert 259346.
6026 * omp-expand.c (expand_oacc_for): Likewise.
6028 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6030 PR rtl-optimization/85354
6031 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6032 * sel-sched.c (sel_global_init): ... here.
6034 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6037 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6038 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6039 mode for PE-COFF targets.
6040 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6041 (i386_pe_asm_lto_end): Likewise.
6042 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6043 (TARGET_ASM_LTO_END): Likewise.
6044 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6045 (i386_pe_asm_lto_start): New function.
6046 (i386_pe_asm_lto_end): Likewise.
6048 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6049 Richard Biener <rguenther@suse.de>
6052 * lto-streamer-out.c (output_function): Fix CFG loop state before
6054 * omp-expand.c (expand_oacc_for): Handle calls to internal
6055 functions like regular functions.
6057 2018-04-12 Richard Biener <rguenther@suse.de>
6060 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6061 for the early LTO debug to properly generate references to it
6062 during DIE emission. Do not re-use that for the skeleton for
6064 (dwarf2out_early_finish): Likewise.
6066 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6069 * config/i386/sse.md
6070 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6071 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6072 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6073 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6074 and output is a reg, avoid creating invalid lowpart subreg, but
6075 instead split into a 512-bit move. Don't split if not AVX512VL,
6076 input is xmm16+ reg and output is a mem.
6077 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6078 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6079 xmm16+ reg and output is a mem.
6081 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6083 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6084 also for flag_dwarf2_cfi_asm.
6086 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6088 PR rtl-optimization/85342
6089 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6090 a bool scalar var inside of the loop instead. Don't try to update
6091 recog_data.operand after failed apply_change_group.
6093 2018-04-12 Tom de Vries <tom@codesourcery.com>
6096 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6097 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6098 array with flexible array member as array without given dimension.
6099 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6100 argument for undefined param to true.
6102 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6105 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6106 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6107 from PowerPC section.
6108 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6109 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6110 help text that is too long.
6111 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6112 help text that is too long.
6113 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6114 help text that is too long.
6116 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6118 * config/alpha/alpha.md (stack_probe_internal): Rename
6119 from "probe_stack". Update all callers.
6121 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6123 PR rtl-optimization/84566
6124 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6125 sched_macro_fuse_insns.
6127 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6130 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6131 (compute_block_dependences): ... from here.
6133 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6135 PR tree-optimization/85331
6136 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6137 from int to HOST_WIDE_INT.
6139 2018-04-11 Martin Jambor <mjambor@suse.cz>
6142 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6143 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6144 not the same as the source val.
6145 (cgraph_edge_brings_value_p): New parameter.
6146 (gather_edges_for_value): Pass destination value to
6147 cgraph_edge_brings_value_p.
6148 (perhaps_add_new_callers): Likewise.
6149 (get_info_about_necessary_edges): Likewise and exclude values brought
6150 only by self-recursive edges.
6151 (create_specialized_node): Redirect only clones of self-calling edges.
6152 (+self_recursive_pass_through_p): New function.
6153 (find_more_scalar_values_for_callers_subset): Use it.
6154 (find_aggregate_values_for_callers_subset): Likewise.
6155 (known_aggs_to_agg_replacement_list): Removed.
6156 (decide_whether_version_node): Re-calculate known constants for all
6157 remaining context clones.
6159 2018-04-11 Richard Biener <rguenther@suse.de>
6162 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6163 from early DWARF output.
6164 (dwarf2out_early_finish): Output line info unconditionally into
6165 early DWARF and add reference to it.
6167 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6170 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6171 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6172 other than V2DFmode using iptr mode attribute.
6173 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6175 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6177 PR rtl-optimization/84659
6178 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6180 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6183 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6185 (output_loc_list): Pass address of a dummy size variable even in the
6186 locview handling loop.
6187 (index_location_lists): Add comment on why skip_loc_list_entry can't
6190 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6193 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6196 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6199 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6200 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6201 and -mstring-compare-inline-limit.
6203 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6206 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6207 for stack clash protection in a register whenever we need it to be in
6210 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6212 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6213 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6215 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6218 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6220 (mlong-double-): Ditto.
6221 * config/rs6000/sysv4.opt (msdata=): Ditto.
6222 (mtls-size=): Ditto.
6224 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6226 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6227 erroneous entries for
6228 "vector int vec_ldl (int, long int *)", and
6229 "vector unsigned int vec_ldl (int, unsigned long int *)".
6230 Add comments and entries for
6231 "vector bool char vec_ldl (int, bool char *)",
6232 "vector bool short vec_ldl (int, bool short *)",
6233 "vector bool int vec_ldl (int, bool int *)",
6234 "vector bool long long vec_ldl (int, bool long long *)",
6235 "vector pixel vec_ldl (int, pixel *)",
6236 "vector long long vec_ldl (int, long long *)",
6237 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6238 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6239 type tree bool_long_long_type_node and correct definition of
6240 bool_V2DI_type_node to make reference to this new type tree.
6241 (rs6000_mangle_type): Replace erroneous reference to
6242 bool_long_type_node with bool_long_long_type_node.
6243 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6244 comments to emphasize sign distinctions for char and int types and
6245 replace RS6000_BTI_bool_long constant with
6246 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6247 use of RS6000_BTI_pixel.
6248 (bool_long_type_node): Remove this macro definition.
6249 (bool_long_long_type_node): New macro definition
6251 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6253 PR rtl-optimization/85300
6254 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6255 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6256 simplify_unary_operation fails.
6258 2018-04-10 Martin Liska <mliska@suse.cz>
6260 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6261 cgraph_edge and ipa_ref.
6263 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6267 * config/i386/sse.md
6268 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6269 computation of the VEC_MERGE selector from mask.
6270 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6271 Fix decoding of the VEC_MERGE selector into mask.
6273 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6275 PR tree-optimization/85286
6276 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6278 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6280 * final.c (final_1): Set insn_last_address as well as
6281 insn_current_address.
6283 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6286 * explow.c (emit_stack_probe): Call validize_mem on memory location
6287 before passing it to gen_probe_stack. Create address operand and
6288 legitimize it for the probe_stack_address case.
6290 2018-04-09 Jan Hubicka <jh@suse.cz>
6293 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6294 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6295 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6297 (free_lang_data): Rebuild type inheritance graph.
6299 2018-04-09 Martin Sebor <msebor@redhat.com>
6301 * invoke.texi (-finline-small-functions): Mention other optimization
6303 (-findirect-inlining, -fpartial-inlining): Same.
6304 (-finline-functions-called-once): Same.
6305 (-freorder-blocks-and-partition): Same.
6307 2018-04-09 Jan Hubicka <jh@suse.cz>
6310 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6311 jumps; choose last target that matches the criteria (i.e.
6312 no partition changes for non-crossing jumps).
6313 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6314 support for redirecting crossing jumps to non-crossing.
6316 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6318 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6319 also for naked functions.
6321 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6323 * config/arc/arc.md (add_shift): New pattern.
6324 (add_shift2): Likewise.
6325 (sub_shift): Likewise.
6326 (sub_shift_cmp0_noout): Likewise.
6327 (compare_si_ashiftsi): Likewise.
6328 (xbfu_cmp0_noout): New combine pattern.
6329 (xbfu_cmp0"): Likewise.
6330 (movsi_set_cc_insn): Place the predicable variant first.
6331 (commutative_binary_cmp0_noout): Remove clobber.
6332 (commutative_binary_cmp0): New pattern.
6333 (noncommutative_binary_cmp0): Likewise.
6334 (noncommutative_binary_cmp0_noout): Likewise.
6335 (noncommutative_binary_comparison_result_used): Removed.
6336 (rsub_cmp0): New pattern.
6337 (rsub_cmp0_noout): Likewise.
6338 (extzvsi): Changed, keep only meaningful variants.
6339 (SQH, SEZ): New iterators.
6340 (SQH_postfix): New mode attribute.
6341 (SEZ_prefix): New code attribute.
6342 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6343 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6344 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6346 (noncommutative_operator): Check the availability of barrel
6349 2018-04-09 Richard Biener <rguenther@suse.de>
6351 PR tree-optimization/85284
6352 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6353 Only use the niter constraining form of simple_iv when the exit
6356 2018-04-09 Tom de Vries <tom@codesourcery.com>
6359 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6360 (define_expand "*memory_barrier"): New define_expand.
6361 (define_insn "memory_barrier"): New insn.
6363 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6365 PR rtl-optimization/80463
6366 PR rtl-optimization/83972
6367 PR rtl-optimization/83480
6369 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6370 correct producer for the insn.
6371 (tidy_control_flow): Fixup seqnos in case of debug insns.
6373 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6375 PR rtl-optimization/83913
6377 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6378 different sched-times when merging exprs.
6380 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6382 PR rtl-optimization/83962
6384 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6385 tidy_fallthru_edge and tidy_control_flow.
6387 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6389 PR rtl-optimization/83530
6391 * sel-sched.c (force_next_insn): New global variable.
6392 (remove_insn_for_debug): When force_next_insn is true, also leave only
6393 next insn in the ready list.
6394 (sel_sched_region): When the region wasn't scheduled, make another pass
6395 over it with force_next_insn set to 1.
6397 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6399 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6401 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6402 for interrupt control.
6403 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6404 functions for interrupt control.
6405 * config/nds32/nds32-intrinsic.md: Likewise.
6406 * config/nds32/nds32_intrinsic.h: Likewise.
6407 * config/nds32/nds32.h (nds32_builtins): Likewise.
6409 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6411 * config/nds32/nds32.c (nds32_init_machine_status,
6412 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6413 strict_aligned_p field.
6414 (nds32_expand_to_rtl_hook): New function.
6415 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6416 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6418 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6419 Chung-Ju Wu <jasonwucj@gmail.com>
6421 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6422 * config/nds32/nds32-n7.md: New file.
6423 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6424 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6426 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6427 * config/nds32/nds32.md (pipeline_model): Add n7.
6428 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6429 * config/nds32/pipelines.md: Include n7 settings.
6431 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6432 Chung-Ju Wu <jasonwucj@gmail.com>
6434 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6435 * config/nds32/nds32-e8.md: New file.
6436 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6437 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6439 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6440 * config/nds32/nds32.md (pipeline_model): Add e8.
6441 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6442 * config/nds32/pipelines.md: Include e8 settings.
6444 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6445 Chung-Ju Wu <jasonwucj@gmail.com>
6447 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6448 * config/nds32/nds32-n8.md: New file.
6449 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6450 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6452 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6453 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6454 * config/nds32/nds32.md (pipeline_model): Add n8.
6455 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6456 * config/nds32/pipelines.md: Include n8 settings.
6458 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6459 Chung-Ju Wu <jasonwucj@gmail.com>
6461 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6462 * config/nds32/nds32-n9-2r1w.md: New file.
6463 * config/nds32/nds32-n9-3r2w.md: New file.
6464 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6465 nds32_register_ports): New or modify for cpu n9.
6466 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6468 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6469 * config/nds32/nds32-utils.c: New file.
6470 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6471 TARGET_MUL_SLOW): Define.
6472 * config/nds32/nds32.md (pipeline_model): New attribute.
6473 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6474 New options that support cpu n9.
6475 * config/nds32/pipelines.md: Include n9 settings.
6476 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6478 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6480 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6481 information if necessary.
6482 (output_cond_branch_compare_zero): Likewise.
6483 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6484 (nds32_target_alignment): Refine for alignment.
6485 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6486 (FUNCTION_BOUNDARY): Modify.
6487 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6489 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6491 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6493 * config/nds32/constants.md (unspec_volatile_element): Add values for
6494 TLB operation and data prefetch.
6495 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6496 functions for TLB operation and data prefetch.
6497 * config/nds32/nds32-intrinsic.md: Likewise.
6498 * config/nds32/nds32_intrinsic.h: Likewise.
6499 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6500 (nds32_print_operand): Likewise.
6501 * config/nds32/nds32.h (nds32_builtins): Likewise.
6503 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6504 Andrew Pinski <pinsika@gcc.gnu.org>
6507 * match.pd: Use constant_boolean_node of correct type instead of
6508 boolean_true_node or boolean_false_node for simplifying
6509 pointer comparisons to zero.
6511 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6513 PR tree-optimization/80021
6514 * tree.c (verify_type_variant): Make error call in verify_variant_match
6515 translatable and remove final full stop.
6517 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6519 * config/nds32/constants.md (unspec_volatile_element): Add
6520 UNSPEC_VOLATILE_EH_RETURN.
6521 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6522 nds32_output_stack_pop): Support dwarf exception handling process.
6523 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6524 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6525 exception handling process.
6526 (nds32_compute_stack_frame): Likewise.
6527 (nds32_return_addr_rtx): Likewise.
6528 (nds32_initial_elimination_offset): Likewise.
6529 (nds32_expand_prologue): Likewise.
6530 (nds32_expand_epilogue): Likewise.
6531 (nds32_dynamic_chain_address): New function.
6532 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6534 (DYNAMIC_CHAIN_ADDRESS): Define.
6535 (EH_RETURN_DATA_REGNO): Define.
6536 (EH_RETURN_STACKADJ_RTX): Define.
6537 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6538 patterns for dwarf exception handling.
6540 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6542 * config/nds32/nds32.h: Clean up obsolete macros.
6544 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6546 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6547 Add enum values for particular instructions.
6548 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6549 particular intrinsic functions.
6550 * config/nds32/nds32-intrinsic.md: Likewise.
6551 * config/nds32/nds32_intrinsic.h: Likewise.
6552 * config/nds32/nds32.h (nds32_builtins): Likewise.
6553 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6554 (btst, ave): New patterns for particular instructions.
6556 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6558 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6559 Add enum values for atomic load/store and memory sync.
6560 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6562 * config/nds32/nds32-intrinsic.md: Likewise.
6563 * config/nds32/nds32_intrinsic.h: Likewise.
6564 * config/nds32/nds32.h (nds32_builtins): Likewise.
6566 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6568 PR tree-optimization/85257
6569 * fold-const.c (native_encode_vector): If not all elts could fit
6570 and off is -1, return 0 rather than offset.
6571 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6572 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6573 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6574 adjust buffer in native_interpret_expr call.
6576 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6578 * config/nds32/constants.md (unspec_volatile_element): Add cache
6579 control enum values.
6580 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6581 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6582 * config/nds32/nds32.c (nds32_cctl_names): New.
6583 (nds32_print_operand): Handle cache control register names.
6584 * config/nds32/nds32.h (nds32_builtins): New enum values.
6585 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6587 * config/nds32/nds32.md (type): Add mmu.
6588 * config/nds32/pipelines.md (simple_insn): Add mmu.
6590 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6592 * config/nds32/nds32.md (type): Remove call.
6593 * config/nds32/pipelines.md (simple_insn): Likewise.
6595 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6597 * config/nds32/constants.md (unspec_volatile_element): Add
6598 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6599 UNSPEC_VOLATILE_FMFCFG.
6600 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6601 description for fmfcfg and fmfcsr.
6602 (bdesc_1arg): Add fmtcsr.
6603 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6604 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6605 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6606 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6607 unspec_fmfcfg): New patterns.
6608 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6609 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6610 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6611 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6612 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6613 __nds32__fmfcfg): Define.
6615 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6617 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6618 intrinsic register names.
6619 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6620 intrinsic register enum values and macros.
6622 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6624 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6625 for load/store addressing form.
6626 (nds32_print_operand_address): Likewise.
6628 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6631 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6632 based on LABEL_REF. Remove useless assertion.
6633 (pic_address_needs_scratch): Fix formatting.
6634 (sparc_legitimize_pic_address): Minor tweaks.
6635 (sparc_delegitimize_address): Adjust assertion accordingly.
6636 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6637 into symbolic_operand.
6638 (movsi_high_pic_label_ref): Likewise.
6639 (movsi_lo_sum_pic_label_ref): Likewise.
6640 (movdi_pic_label_ref): Likewise.
6641 (movdi_high_pic_label_ref): Likewise.
6642 (movdi_lo_sum_pic_label_ref): Likewise.
6644 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6646 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6647 custom LIB_SPEC setup.
6649 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6650 Kito Cheng <kito.cheng@gmail.com>
6652 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6653 * config/riscv/freebsd.h: New.
6655 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6657 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6658 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6661 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6662 Kito Cheng <kito.cheng@gmail.com>
6664 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6665 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6666 * config/nds32/nds32-protos.h (nds32_output_call,
6667 nds32_output_return): Declare.
6668 * config/nds32/nds32.md: Refine all the call and return patterns.
6670 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6673 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6674 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6676 PR rtl-optimization/84872
6677 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6678 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6681 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6683 * expr.c (copy_blkmode_to_reg): Revert 254862.
6684 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6686 2018-04-06 Richard Biener <rguenther@suse.de>
6689 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6690 after seeing a component reference with an adjacent field. Treat
6691 refs to arrays at struct end of external decls similar to
6692 refs to unconstrained commons.
6694 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6697 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6698 look through SAVE_EXPRs with non-side-effects argument. Adjust
6700 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6703 2018-04-06 Richard Biener <rguenther@suse.de>
6706 * alias.c (find_base_term): New wrapper around find_base_term
6707 unwinding CSELIB_VAL_PTR changes.
6708 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6711 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6713 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6715 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6716 constant definitions.
6717 ("nop"): lr 0,0 -> nopr r0
6718 ("nop_lr0", "nop_lr1"): New insn definitions.
6720 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6722 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6723 NDS32_V3PUSH_AVAILABLE_P macro.
6725 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6726 Chung-Ju Wu <jasonwucj@gmail.com>
6728 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6729 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6730 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6731 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6732 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6733 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6734 * config/nds32/constraints.md: New constraints and checking for hard
6735 float configuration.
6736 * config/nds32/iterators.md: New mode iterator and attribute for hard
6737 float configuration.
6738 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6740 * config/nds32/nds32-fpu.md: New file.
6741 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6742 deal with hard float code generation.
6743 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6745 (abi_type, float_reg_number): New enum type.
6746 * config/nds32/nds32-predicates.c: New predicates for hard float.
6747 * config/nds32/nds32-protos.h: Declare functions for hard float.
6748 * config/nds32/nds32.c: Implementation for hard float configuration.
6749 * config/nds32/nds32.h: Definitions for hard float configuration.
6750 * config/nds32/nds32.md: Include hard float machine description and
6751 modify patterns for hard float configuration.
6752 * config/nds32/nds32.opt: New options for hard float configuration.
6753 * config/nds32/predicates.md: New predicates for hard float
6756 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6758 * common/config/nds32/nds32-common.c
6759 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6761 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6764 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6765 CONSTRUCTOR_ELT (ctor, ...)->value.
6767 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6770 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6772 2018-04-05 Tom de Vries <tom@codesourcery.com>
6775 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6778 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6779 Kito Cheng <kito.cheng@gmail.com>
6781 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6782 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6783 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6784 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6786 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6787 Kito Cheng <kito.cheng@gmail.com>
6789 * config/nds32/constraints.md (Ufe): New memory constraint.
6790 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6791 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6792 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6794 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6795 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6797 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6799 * config/nds32/nds32.md: Use optimize_size in the condition for
6800 alu-shift instructions.
6802 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6804 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6806 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6808 * config/nds32/nds32.md (negsi2): Refine pattern.
6810 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6811 Chung-Ju Wu <jasonwucj@gmail.com>
6813 * config/nds32/iterators.md (shift_rotate): New code iterator.
6814 (shift): New code attribute.
6815 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6816 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6817 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6818 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6819 bit-wise operations.
6820 (andsi3, *andsi3): Ditto.
6821 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6822 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6823 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6824 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6825 nds32_ior_operand, nds32_xor_operand): New predicates.
6827 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6829 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6830 (addsi3, subsi3): ... this.
6832 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6834 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6836 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6838 * config/nds32/nds32.md: Adjust indention.
6840 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6842 * config/nds32/nds32.md (feature): New attribute.
6844 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6846 * config/nds32/nds32.md (subtype): New attribute.
6848 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6851 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6852 expansion to perform a bitwise AND of the argument followed by a
6853 boolean negation of the result.
6855 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6857 PR rtl-optimization/84878
6858 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6859 the basic block. Assert the use reference is not artificial and that
6860 it has an associated insn.
6862 2018-04-04 Michael Matz <matz@suse.de>
6864 * builtins.c (compute_objsize): Pass correct operand
6865 to array_at_struct_end_p.
6867 2018-04-04 Richard Biener <rguenther@suse.de>
6870 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6871 from contexts for DINFO_LEVEL_TERSE and below.
6873 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6875 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6876 resiter_operand condition.
6877 * config/nds32/nds32.md (*move<mode>): Ditto.
6879 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6880 Monk Chiang <sh.chiang04@gmail.com>
6882 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6884 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6886 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6888 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6889 Kito Cheng <kito.cheng@gmail.com>
6891 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6892 nds32_cond_code_str, output_cond_branch,
6893 output_cond_branch_compare_zero, nds32_expand_cbranch,
6894 nds32_expand_cstore, nds32_expand_movcc,
6895 nds32_output_cbranchsi4_equality_zero,
6896 nds32_output_cbranchsi4_equality_reg,
6897 nds32_output_cbranchsi4_equality_reg_or_const_int,
6898 nds32_output_cbranchsi4_greater_less_zero: New functions.
6899 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6900 nds32_expand_cstore, nds32_expand_movcc,
6901 nds32_output_cbranchsi4_equality_zero,
6902 nds32_output_cbranchsi4_equality_reg,
6903 nds32_output_cbranchsi4_equality_reg_or_const_int,
6904 nds32_output_cbranchsi4_greater_less_zero): Declare.
6905 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6906 nds32_rimm11s_operand): New predicates.
6907 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6908 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6911 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6913 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6914 * config/nds32/nds32.md: Ditto.
6915 * config/nds32/pipelines.md: Ditto.
6917 2018-04-04 Richard Biener <rguenther@suse.de>
6919 PR tree-optimization/85168
6920 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6921 propagating abnormals.
6923 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6925 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
6927 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6928 Kito Cheng <kito.cheng@gmail.com>
6930 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
6931 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
6932 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
6933 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
6934 * config/nds32/nds32.md (sibcall_internal): New.
6935 (sibcall_register): Remove.
6936 (sibcall_immediate): Remove.
6937 (sibcall_value_internal): New.
6938 (sibcall_value_register): Remove.
6939 (sibcall_value_immediate): Remove.
6940 * config/nds32/predicates.md (nds32_general_register_operand): New.
6941 (nds32_call_address_operand): New.
6943 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6945 PR rtl-optimization/85167
6946 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
6947 bb_defs if *split_p, instead preinitialize it to NULL.
6949 PR tree-optimization/85156
6950 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
6951 evaluating the argument multiple times.
6953 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
6955 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
6957 (_mm_cvtpd_ps): Likewise.
6958 (_mm_cvttpd_epi32): Likewise.
6959 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
6960 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
6961 vector, pixel, and bool following altivec.h include.
6963 2018-04-03 Martin Sebor <msebor@redhat.com>
6965 * doc/extend.texi (Common Function Attributes): Clarify.
6966 (const attribute): Likewise.
6967 (pure attribute): Likewise.
6969 2018-04-03 Jakub Jelinek <jakub@redhat.com>
6972 * config/i386/i386.c (ix86_expand_vector_set): Use
6973 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
6975 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
6977 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
6978 instructions when changing rounding bits to preserve precision bits
6979 in the x87 control word.
6981 2018-04-03 Martin Liska <mliska@suse.cz>
6983 PR tree-optimization/82491
6984 * rtl.h (strip_offset_and_add): Replace += suboffset with
6985 poly_uint64 () + suboffset.
6987 2018-03-29 Martin Liska <mliska@suse.cz>
6988 Martin Jambor <mjambor@suse.cz>
6991 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
6992 param_type is not an integral or pointer type.
6994 2018-04-03 Richard Biener <rguenther@suse.de>
6996 * sese.h (recompute_all_dominators): Remove.
6998 2018-04-02 Martin Sebor <msebor@redhat.com>
7000 * doc/invoke.texi (-Wrestrict): Fix typos.
7002 2018-04-02 Jim Wilson <jimw@sifive.com>
7004 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7005 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7006 (<optab>di3, <optab>si3_extend): Likewise.
7007 (<optab>si3_mask, <optab>si3_mask_1): New.
7008 (<optab>di3_mask, <optab>di3_mask_1): New.
7009 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7010 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7011 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7013 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7015 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7018 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7020 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7021 (nds32_canonicalize_comparison): New function.
7023 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7024 Kito Cheng <kito.cheng@gmail.com>
7025 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7027 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7028 * config/nds32/constants.md (unspec_volatile_element): Add
7029 UNSPEC_VOLATILE_RELAX_GROUP.
7030 * config/nds32/nds32-relax-opt.c: New file.
7031 * config/nds32/nds32-predicates.c
7032 (nds32_symbol_load_store_p): New function.
7033 * config/nds32/nds32-protos.h
7034 (nds32_symbol_load_store_p): Declare function.
7035 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7036 * config/nds32/nds32.c
7037 (nds32_register_pass): New function to register pass.
7038 (nds32_register_passes): New function to register passes.
7039 * config/nds32/nds32.md (relax_group): New pattern.
7040 * config/nds32/nds32.opt (mrelax-hint): New option.
7041 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7043 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7045 * config/nds32/t-nds32: Modify files dependency.
7047 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7049 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7050 (PROFILE_HOOK): Define its implementation.
7052 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7054 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7055 type and 32-bit size.
7057 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7060 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7061 (V_128_256): New mode iterator.
7062 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7063 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7064 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7066 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7067 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7069 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7072 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7073 NaN inputs correctly.
7075 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7078 * config/rs6000/vsx.md (??r): New mode attribute.
7079 (*vsx_mov<mode>_64bit): Use it.
7080 (*vsx_mov<mode>_32bit): Likewise.
7082 2018-03-30 Martin Sebor <msebor@redhat.com>
7084 PR tree-optimization/84818
7085 * builtins.c (check_access): Use warning_n.
7087 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7090 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7092 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7095 2018-03-30 Julia Koval <julia.koval@intel.com>
7098 * x86-tune.def (movx, partial_reg_dependency): Enable for
7101 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7104 * lra-constraints.c (process_alt_operands): Move setting
7105 this_alternative_matches below.
7107 2018-03-29 Martin Liska <mliska@suse.cz>
7110 * doc/invoke.texi: Document how LTO works with debug info.
7111 Describe auto-load support of binutils. Mention 'x86-64'
7112 as valid option value of -march option.
7114 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7116 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7119 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7120 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7121 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7124 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7127 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7128 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7129 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7130 for RS6000_BTM_POWERPC64.
7131 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7132 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7133 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7138 2018-03-28 Carl Love <cel@us.ibm.com>
7141 2017-09-27 Carl Love <cel@us.ibm.com>
7143 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7144 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7145 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7148 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7150 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7151 instead of __vector bool.
7152 (_mm_max_pu8): Likewise.
7153 (_mm_min_pi16): Likewise.
7155 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7158 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7159 (DIVWEUO): Likewise.
7161 (DIVDEUO): Likewise.
7162 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7163 DIVWEUO and DIVDEUO.
7164 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7165 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7166 (div_extend): Likewise.
7167 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7169 (__builtin_divweuo): Likewise.
7170 (__builtin_divdeo): Likewise.
7171 (__builtin_divdeuo): Likewise.
7173 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7176 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7177 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7179 PR tree-optimization/82004
7180 * gimple-match-head.c (optimize_pow_to_exp): New function.
7181 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7182 Don't fold to exp if optimize_pow_to_exp is false.
7184 2018-03-28 Martin Liska <mliska@suse.cz>
7187 * calls.c (initialize_argument_information): Fix trailing space.
7188 * common.opt: Fix typo and provide better explanation for
7189 -fsanitize-coverage option.
7190 * config/i386/i386.opt: Fix typo.
7192 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7193 Martin Liska <mliska@suse.cz>
7196 * gimplify.c (asan_poison_variable): Don't do the check for
7197 gimplify_omp_ctxp here.
7198 (gimplify_decl_expr): Do it here.
7199 (gimplify_target_expr): Likewise.
7201 2018-03-28 Martin Liska <mliska@suse.cz>
7204 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7205 chkp_type_bounds_count if MPX is not enabled.
7207 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7209 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7211 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7214 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7215 function to create the function decl for complex long double
7216 multiply and divide for -mabi=ieeelongdouble.
7217 (init_float128_ieee): Call it.
7219 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7222 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7223 -fcf-protection=branch -mibt.
7224 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7226 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7229 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7230 arm_word_relocations.
7232 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7235 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7236 extern array declarations.
7238 2018-03-27 Richard Biener <rguenther@suse.de>
7241 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7242 explicit single_use checks.
7244 2018-03-27 Richard Biener <rguenther@suse.de>
7246 PR tree-optimization/85082
7247 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7250 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7252 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7253 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7254 Turn on fasynchronous-unwind-tables and funwind-tables.
7256 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7259 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7260 (*bmi_blsr_<mode>_ccz): Ditto.
7262 2018-03-26 Tom de Vries <tom@codesourcery.com>
7264 PR tree-optimization/85063
7265 * omp-general.c (offloading_function_p): New function. Factor out
7267 * omp-offload.c (pass_omp_target_link::gate): ... here.
7268 * omp-general.h (offloading_function_p): Declare.
7269 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7270 with attribute omp declare target for offloading functions.
7272 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7274 PR tree-optimization/84005
7275 * tree-data-ref.h (get_base_for_alignment): Declare.
7276 * tree-data-ref.c (get_base_for_alignment_1): New function.
7277 (get_base_for_alignment): Likewise.
7278 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7279 get_base_for_alignment to find a suitable base object, instead
7280 of always using drb->base_address.
7282 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7285 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7286 known size by default.
7288 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7291 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7292 and non BLKmode operands.
7294 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7297 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7298 Clean up attributes.
7300 2018-03-23 Richard Biener <rguenther@suse.de>
7303 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7304 we are going to emit early debug for LTO.
7306 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7309 * function.c (match_asm_constraints_1): Don't optimize if input
7310 doesn't satisfy general_operand predicate for output's mode.
7313 * alias.c (write_dependence_p): Don't require for x_canonicalized
7314 non-VOIDmode if x has VOIDmode.
7317 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7318 just don't try to optimize it rather than assert it never happens.
7320 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7322 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7323 macro expansions for definition of ST_INTERNAL_<mode> and
7324 LD_INTERNAL_<mode> builtins.
7325 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7327 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7329 (altivec_expand_st_builtin): Likewise.
7330 (altivec_expand_builtin): Remove calls to deleted functions.
7331 (rs6000_address_for_altivec): Delete this function.
7332 * config/rs6000/vector.md: Remove expands for
7333 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7335 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7338 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7339 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7340 re-computing once computed.
7341 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7342 (arm_init_machine_status): Initialize
7343 machine->static_chain_stack_bytes.
7345 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7348 * doc/extend.texi: Add four new prototypes for vec_ld.
7349 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7350 definitions for more logical presentation.
7351 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7352 entries for V1TI variants of __builtin_altivec_ld builtin.
7353 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7354 handling of V1TI variant of LVX icode pattern.
7355 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7356 (rs6000_gimple_fold_builtin): Likewise.
7357 (altivec_init_builtins): Add code to define
7358 __builtin_altivec_lvx_v1ti function.
7360 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7363 * function.c (match_asm_constraints_1): Don't do the optimization
7364 if input isn't a REG, SUBREG, MEM or constant.
7366 2018-03-22 Tom de Vries <tom@codesourcery.com>
7368 PR tree-optimization/84956
7369 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7370 bb_has_abnormal_pred.
7372 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7375 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7376 DECL_INITIAL (decl) to decl at the end.
7377 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7380 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7382 * doc/extend.texi (__builtin_tgmath): Document when complex
7383 integer types are treated as _Complex _Float64.
7385 2018-03-21 Tom de Vries <tom@codesourcery.com>
7387 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7389 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7391 PR tree-optimization/84960
7392 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7393 if it is ENTRY block, move them into single succ of ENTRY in that case.
7395 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7397 PR tree-optimization/84811
7398 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7399 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7400 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7401 (known_subrange_p): Update accordingly. Cast each value involved
7402 in the size comparison, rather than casting the result of the
7405 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7407 PR tree-optimization/84982
7408 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7409 by flipping the least significant bit rather than all bits from
7410 bitpos to bitpos + bitsize - 1.
7412 2018-03-21 Nathan Sidwell <nathan@acm.org>
7414 * doc/extend.texi (Deprecated Features): Remove mention of
7415 long-deleted deprecations.
7417 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7420 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7421 * configure: Regenerate.
7423 2018-03-21 Tom de Vries <tom@codesourcery.com>
7425 PR tree-optimization/83126
7426 * tree-parloops.c (num_phis): New function.
7427 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7429 2018-03-21 Nathan Sidwell <nathan@acm.org>
7431 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7432 mention anon-struct/union members and trailing attributes.
7434 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7436 PR tree-optimization/84969
7437 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7438 builtin memset partitions if they set different rhs values.
7440 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7442 PR rtl-optimization/84989
7443 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7444 VEC_DUPLICATE with scalar result mode.
7446 2018-03-21 Martin Liska <mliska@suse.cz>
7449 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7450 not intended return statement.
7452 2018-03-21 Martin Liska <mliska@suse.cz>
7455 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7456 (chkp_find_bound_slots_1): Limit number of iterations.
7458 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7461 * Minor grammar fixes for x86 options.
7463 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7466 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7467 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7469 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7472 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7473 (altivec_lvx_<mode>_1op): Likewise.
7474 (altivec_stvx_<mode>_2op): Likewise.
7475 (altivec_stvx_<mode>_1op): Likewise.
7476 (altivec_lvx_<VM2:mode>): New define_expand.
7477 (altivec_stvx_<VM2:mode>): Likewise.
7478 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7479 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7480 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7481 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7482 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7483 (rs6000_gen_lvx): Likewise.
7484 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7485 (altivec_expand_stv_builtin): Likewise.
7486 (altivec_expand_builtin): Likewise.
7487 * config/rs6000/vector.md: Likewise.
7489 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7492 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7495 2018-03-20 Richard Biener <rguenther@suse.de>
7498 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7499 sign-conversions as zero, fall back to standard scalar_stmt
7502 2018-03-20 Martin Liska <mliska@suse.cz>
7505 * predict.c (rebuild_frequencies): Handle case when we have
7506 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7508 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7511 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7512 flag_section_anchors.
7513 * varasm.c (use_blocks_for_decl_p): Remove hack for
7514 dw2_force_const_mem.
7517 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7519 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7520 be created, use lowpart_subreg of operands[0] rather than operands[0]
7522 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7523 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7524 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7525 and n constraint instead of aarch64_shift_imm_di and Usd.
7526 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7527 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7529 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7532 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7533 to favor GPR over NEON registers.
7534 (<shift>di3_neon): Likewise.
7536 2018-03-20 Tom de Vries <tom@codesourcery.com>
7539 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7540 (nvptx_process_pars): Emit bar.sync asap and alap.
7542 2018-03-20 Tom de Vries <tom@codesourcery.com>
7545 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7546 seen_label if seen_label is already set.
7548 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7551 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7552 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7553 Use 1U instead of 1. Formatting fixes.
7556 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7557 instead of TREE_TYPE (s1) for the return value.
7559 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7561 PR tree-optimization/84946
7562 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7563 bitsize + bitsize in poly_uint64 rather than poly_int64.
7566 * dwarf2asm.c: Include fold-const.c.
7567 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7568 of decl rather than decl itself.
7570 PR rtl-optimization/84643
7571 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7573 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7576 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7577 calling assemble_variable.
7579 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7582 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7583 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7585 2018-03-19 Jim Wilson <jimw@sifive.com>
7588 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7589 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7590 (riscv_first_stack_step): Likewise.
7591 (riscv_option_override): Use STACK_BOUNDARY instead of
7593 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7595 (BIGGEST_ALIGNMENT): Set to 128.
7596 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7597 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7600 2018-03-19 Richard Biener <rguenther@suse.de>
7602 PR tree-optimization/84933
7603 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7604 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7606 2018-03-19 Richard Biener <rguenther@suse.de>
7608 PR tree-optimization/84859
7609 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7610 (cond_if_else_store_replacement): Perform sinking operation on
7611 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7612 Generalize what a BB with a single eligible store is.
7614 2018-03-19 Richard Biener <rguenther@suse.de>
7616 PR tree-optimization/84929
7617 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7618 chrec_is_positive against non-chrec arg.
7620 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7623 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7625 2018-03-18 Martin Liska <mliska@suse.cz>
7627 PR rtl-optimization/84635
7628 * regrename.c (build_def_use): Use matches_mode only when
7631 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7633 PR tree-optimization/84913
7634 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7635 vectorize chains of COND_EXPRs.
7637 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7639 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7641 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7643 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7645 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7647 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7649 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7650 Kito Cheng <kito.cheng@gmail.com>
7652 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7653 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7654 (nds32_adjust_reg_alloc_order): New function.
7655 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7657 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7659 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7660 nds32_print_operand, nds32_print_operand_address): Use
7661 HOST_WIDE_INT_PRINT_DEC instead.
7663 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7665 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7667 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7670 * config/i386/i386.c (initial_ix86_tune_features,
7671 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7673 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7674 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7675 rather than 1u << ix86_tune. Formatting fix.
7676 (ix86_option_override_internal): Change ix86_arch_mask from
7677 unsigned int to unsigned HOST_WIDE_INT, initialize to
7678 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7679 (ix86_function_specific_restore): Likewise.
7681 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7684 * postreload.c (reload_combine_recognize_pattern): Perform
7685 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7686 truncate_int_for_mode the result for the destination's mode.
7689 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7690 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7692 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7695 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7698 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7699 regno_allocno_class_array and sorted_pseudos.
7700 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7701 insns where regno is used.
7703 2018-03-16 Martin Liska <mliska@suse.cz>
7706 * multiple_target.c (create_dispatcher_calls): Redirect
7707 reference in the symbol table.
7709 2018-03-16 Martin Liska <mliska@suse.cz>
7712 * multiple_target.c (create_dispatcher_calls): Redirect also
7715 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7719 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7720 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7721 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7723 2018-03-16 Julia Koval <julia.koval@intel.com>
7725 * doc/invoke.texi (Skylake Server): Add CLWB.
7726 Cannonlake): Remove CLWB.
7728 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7730 PR tree-optimization/84841
7731 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7733 (FLOAT_ONE_CONST_TYPE): Define.
7734 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7735 (sort_by_operand_rank): Put entries with higher constant_type last
7736 rather than first to match comments.
7738 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7740 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7743 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7746 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7747 check and use address of target always.
7749 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7752 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7753 (indirect_thunk_bnd_needed): Likewise.
7754 (indirect_thunks_used): Likewise.
7755 (indirect_thunks_bnd_used): Likewise.
7756 (indirect_return_needed): New.
7757 (indirect_return_bnd_needed): Likewise.
7758 (output_indirect_thunk_function): Add a bool argument for
7760 (output_indirect_thunk_function): Don't generate alias for
7761 function return thunk.
7762 (ix86_code_end): Call output_indirect_thunk_function to generate
7763 function return thunks.
7764 (ix86_output_function_return): Set indirect_return_bnd_needed
7765 and indirect_return_needed instead of indirect_thunk_bnd_needed
7766 and indirect_thunk_needed.
7768 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7770 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7771 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7772 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7774 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7775 Paul Hua <paul.hua.gm@gmail.com>
7778 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7780 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7782 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7783 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7786 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7789 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7790 instead of GET_MODE_SIZE when comparing Units.
7792 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7795 * varasm.c (hash_section): Return an unchangeble hash value
7796 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7797 Return !aarch64_can_use_per_function_literal_pools_p ().
7799 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7802 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7803 rather than address of cmode as last argument to prepare_cmp_insn.
7805 2018-03-15 Julia Koval <julia.koval@intel.com>
7807 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7808 F_AVX512VNNI, F_AVX512BITALG): New.
7810 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7813 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7814 insn for floating-point loads and stores.
7816 2018-03-14 Carl Love <cel@us.ibm.com>
7818 * config/rs6000/rs6000-c.c: Add macro definitions for
7819 ALTIVEC_BUILTIN_VEC_PERMXOR.
7820 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7821 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7822 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7823 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7825 * config/doc/extend.texi: Add prototypes for vec_permxor.
7827 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7830 * diagnostic-show-locus.c (class layout_point): Convert m_line
7831 from int to linenum_type.
7832 (line_span::comparator): Use linenum "compare" function when
7833 comparing line numbers.
7834 (test_line_span): New function.
7835 (layout_range::contains_point): Convert param "row" from int to
7837 (layout_range::intersects_line_p): Likewise.
7838 (layout::will_show_line_p): Likewise.
7839 (layout::print_source_line): Likewise.
7840 (layout::should_print_annotation_line_p): Likewise.
7841 (layout::print_annotation_line): Likewise.
7842 (layout::print_leading_fixits): Likewise.
7843 (layout::annotation_line_showed_range_p): Likewise.
7844 (struct line_corrections): Likewise for field m_row.
7845 (line_corrections::line_corrections): Likewise for param "row".
7846 (layout::print_trailing_fixits): Likewise.
7847 (layout::get_state_at_point): Likewise.
7848 (layout::get_x_bound_for_row): Likewise.
7849 (layout::print_line): Likewise.
7850 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7851 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7852 * input.c (selftest::test_linenum_comparisons): New function.
7853 (selftest::input_c_tests): Call it.
7854 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7855 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7856 * selftest.h (ASSERT_GT): New macro.
7857 (ASSERT_GT_AT): New macro.
7858 (ASSERT_LT): New macro.
7859 (ASSERT_LT_AT): New macro.
7861 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7863 PR rtl-optimization/84780
7864 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7866 2018-03-14 Martin Liska <mliska@suse.cz>
7868 * tree.c (record_node_allocation_statistics): Use
7869 get_stats_node_kind.
7870 (get_stats_node_kind): New function extracted from
7871 record_node_allocation_statistics.
7872 (free_node): Use get_stats_node_kind.
7874 2018-03-14 Richard Biener <rguenther@suse.de>
7876 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7877 that the value-set of ANTIC_IN doesn't grow.
7880 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7882 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7883 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7885 2018-03-14 Julia Koval <julia.koval@intel.com>
7887 * config.gcc (icelake-client, icelake-server): New.
7889 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7890 (initial_ix86_arch_features): Ditto.
7891 (PTA_SKYLAKE): Add SGX.
7892 (PTA_ICELAKE): Remove.
7893 (PTA_ICELAKE_CLIENT): New.
7894 (PTA_ICELAKE_SERVER): New.
7895 (ix86_option_override_internal): Split up icelake on icelake client and
7897 (get_builtin_code_for_version): Ditto.
7898 (fold_builtin_cpu): Ditto.
7899 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7900 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7901 * config/i386/i386.h (processor_type): Ditto.
7902 * doc/invoke.texi: Ditto.
7904 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7907 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7908 INTEGER_CST offset, add it together with bitpos / 8 and
7909 sign extend based on POINTER_SIZE.
7913 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7916 * config/i386/constraints.md (Yc): New register constraint.
7917 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7918 Use Yc constraint for alternative 2 of operand 0. Remove
7919 preferred_for_speed attribute.
7921 2018-03-14 Richard Biener <rguenther@suse.de>
7923 PR tree-optimization/84830
7924 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
7925 with the old one to avoid oscillations.
7927 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
7930 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
7932 (assign_by_spills): Return a flag of reload assignment failure.
7933 Do not process the reload assignment failures. Do not spill other
7934 reload pseudos if they has the same reg class. Update n if
7936 (lra_assign): Add a return arg. Set up from the result of
7937 assign_by_spills call.
7938 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
7939 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
7940 usage_insns if it is not NULL.
7941 (spill_hard_reg_in_range): New function.
7942 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
7943 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
7944 function prototypes.
7945 (lra_assign): Change prototype.
7946 * lra.c (lra): Add code to deal with fails by splitting hard reg
7949 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
7951 * config/riscv/riscv.opt (mrelax): New option.
7952 * config/riscv/riscv.c (riscv_file_start): Emit ".option
7953 "norelax" when riscv_mrelax is disabled.
7954 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
7956 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7959 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
7960 reassociation for int modes.
7962 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7964 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
7965 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
7967 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
7968 * config/aarch64/aarch64-sve.md
7969 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
7970 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
7971 (*extend<mode><Vwide>2): Rename to...
7972 (aarch64_sve_extend<mode><Vwide>2): ...this.
7973 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
7974 renaming the old pattern to...
7975 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
7977 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
7978 define_expand, renaming the old pattern to...
7979 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
7980 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
7981 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
7982 account when deciding which SVE instruction the optab should use.
7983 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
7985 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7987 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
7988 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
7989 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
7990 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
7991 (tlsdesc_small_<mode>): Turn a define_expand and use
7992 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
7993 (tlsdesc_small_advsimd_<mode>): ...this.
7994 (tlsdesc_small_sve_<mode>): New pattern.
7996 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
7998 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
7999 (UNSPEC_UMUL_HIGHPART): New constants.
8000 (MUL_HIGHPART): New int iteraor.
8001 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8002 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8004 (*<su>mul<mode>3_highpart): New define_insn.
8006 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8009 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8012 2018-03-13 Martin Liska <mliska@suse.cz>
8015 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8017 (sem_item_optimizer::~sem_item_optimizer): Release it.
8018 (sem_item_optimizer::merge_classes): Register variable aliases.
8019 (sem_item_optimizer::fixup_pt_set): New function.
8020 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8021 * ipa-icf.h: Declare new variables and functions.
8023 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8026 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8027 integer_pow2p@2 and test integer_pow2p in condition.
8028 (A < 0 ? C : 0): Similarly for @1.
8031 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8032 characters starting at p contain '\0' character, don't look beyond
8036 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8037 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8040 * reg-stack.c (change_stack): Change update_end var from int to
8041 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8042 also call set_block_for_insn on the newly added insns and rescan.
8045 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8046 on the last operand.
8049 * tree.c (stabilize_reference_1): Return save_expr (e) for
8050 STATEMENT_LIST even if it doesn't have side-effects.
8052 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8054 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8056 2018-03-12 Renlin Li <renlin.li@arm.com>
8058 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8059 aarch64_output_scalar_simd_mov_immediate.
8061 2018-03-12 Martin Sebor <msebor@redhat.com>
8063 PR tree-optimization/83456
8064 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8065 for perfectly overlapping calls to memcpy.
8066 (gimple_fold_builtin_memory_chk): Same.
8067 (gimple_fold_builtin_strcpy): Handle no-warning.
8068 (gimple_fold_builtin_stxcpy_chk): Same.
8069 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8071 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8073 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8074 parameter. Use it for SFmode.
8075 (rs6000_function_arg_advance_1): Adjust.
8076 (rs6000_function_arg): Adjust.
8077 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8079 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8081 PR rtl-optimization/84169
8082 PR rtl-optimization/84780
8083 * combine.c (can_combine_p): Check for a 2-insn combination whether
8084 the destination register is used between the two insns, too.
8086 2018-03-12 Richard Biener <rguenther@suse.de>
8088 PR tree-optimization/84803
8089 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8090 for refs DR analysis didn't process.
8092 2018-03-12 Richard Biener <rguenther@suse.de>
8094 PR tree-optimization/84777
8095 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8096 force-vectorize loops ignore whether we are optimizing for size.
8098 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8100 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8101 (TARGET_MD_ASM_ADJUST): Define.
8103 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8104 Kito Cheng <kito.cheng@gmail.com>
8105 Chung-Ju Wu <jasonwucj@gmail.com>
8107 * config/nds32/nds32.c (nds32_compute_stack_frame,
8108 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8109 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8110 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8111 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8112 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8113 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8114 * config/nds32/nds32.md (prologue, epilogue): Use macro
8115 NDS32_V3PUSH_AVAILABLE_P to do checking.
8117 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8120 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8121 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8122 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8123 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8124 addition of most attributes on !orig_type_die or the attribute not
8125 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8127 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8128 Chung-Ju Wu <jasonwucj@gmail.com>
8130 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8132 * config/nds32/nds32.opt (mvh): New option.
8134 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8135 Chung-Ju Wu <jasonwucj@gmail.com>
8137 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8139 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8140 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8143 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8144 Chung-Ju Wu <jasonwucj@gmail.com>
8146 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8148 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8149 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8151 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8152 Kito Cheng <kito.cheng@gmail.com>
8153 Chung-Ju Wu <jasonwucj@gmail.com>
8155 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8156 UNSPEC_FFMISM and UNSPEC_FLMISM.
8157 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8158 for ffb, ffmism and flmism.
8159 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8160 (unspec_ffmism): Ditto.
8161 (unspec_flmism): Ditto.
8162 (nds32_expand_builtin_impl): Check if string extension is available.
8163 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8164 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8166 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8169 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8172 * lra-assigns.c (assign_by_spills): Return a flag of reload
8173 assignment failure. Do not process the reload assignment
8174 failures. Do not spill other reload pseudos if they has the same
8176 (lra_assign): Add a return arg. Set up from the result of
8177 assign_by_spills call.
8178 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8179 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8180 usage_insns if it is not NULL.
8181 (spill_hard_reg_in_range): New function.
8182 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8183 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8184 function prototypes.
8185 (lra_assign): Change prototype.
8186 * lra.c (lra): Add code to deal with fails by splitting hard reg
8189 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8192 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8194 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8197 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8198 (dw_val_node): Add val_symbolic_view.
8199 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8200 (symview_upper_bound): New.
8201 (new_line_info_table): Initialize symviews_since_reset.
8202 (dwarf2out_source_line): Count symviews_since_reset and set
8203 symview_upper_bound.
8204 (dw_val_equal_p): Handle symview.
8205 (add_AT_symview): New.
8206 (print_dw_val): Handle symview.
8207 (attr_checksum, attr_checksum_ordered): Likewise.
8208 (same_dw_val_p, size_of_die): Likewise.
8209 (value_format, output_die): Likewise.
8210 (add_high_low_attributes): Use add_AT_symview for entry_view.
8211 (dwarf2out_finish): Reset symview_upper_bound, clear
8214 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8217 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8218 Add strict argument and use it.
8219 (rs6000_split_multireg_move): Update for new strict argument.
8220 (mem_operand_gpr): Disallow all non-offsettable addresses.
8221 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8223 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8226 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8227 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8228 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8231 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8232 decl, use remap_type if we want to use the type.
8234 2018-03-09 Martin Sebor <msebor@redhat.com>
8236 PR tree-optimization/84526
8237 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8239 (builtin_access::generic_overlap): Be prepared to handle non-array
8242 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8244 PR rtl-optimization/84682
8245 * lra-constraints.c (process_address_1): Check is_address flag
8246 for address constraints.
8247 (process_alt_operands): Likewise.
8248 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8249 preprocess_constraints.
8250 * recog.h (preprocess_constraints): Add oploc parameter.
8252 * recog.c (preprocess_constraints): Test address_operand for
8253 CT_ADDRESS constraints.
8255 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8258 * lra-assigns.c (assign_by_spills): Return a flag of reload
8259 assignment failure. Do not process the reload assignment
8260 failures. Do not spill other reload pseudos if they has the same
8262 (lra_assign): Add a return arg. Set up from the result of
8263 assign_by_spills call.
8264 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8265 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8266 usage_insns if it is not NULL.
8267 (spill_hard_reg_in_range): New function.
8268 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8269 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8270 function prototypes.
8271 (lra_assign): Change prototype.
8272 * lra.c (lra): Add code to deal with fails by splitting hard reg
8275 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8278 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8279 Accept complain bool parameter. Only emit errors if it is true.
8280 (arm_parse_cpu_option_name): Likewise.
8281 (arm_target_thumb_only): Adjust callers of the above.
8282 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8283 prototype to take a default true bool parameter.
8284 (arm_parse_arch_option_name): Likewise.
8286 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8287 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8291 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8292 * configure: Regenerate.
8293 * configure.ac ("linker --version-script option"): New.
8294 ("linker soname option"): New.
8296 2018-03-09 Richard Biener <rguenther@suse.de>
8298 PR tree-optimization/84775
8299 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8300 immediate uses of predicate stmts and mark them modified.
8303 PR tree-optimization/84178
8304 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8306 (version_loop_for_if_conversion): Delay update_ssa call.
8307 (tree_if_conversion): Delay update_ssa until after predicate
8310 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8313 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8314 when the function accesses prior frames.
8316 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8319 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8320 gen_llsym, otherwise call maybe_gen_llsym.
8323 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8324 has ',' character inside of it.
8326 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8329 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8330 as clobbering CC_REGNUM.
8332 2018-03-08 Richard Biener <rguenther@suse.de>
8335 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8336 (follow_copies_to_constant): Do not follow SSA names registered
8339 2018-03-08 Richard Biener <rguenther@suse.de>
8341 PR tree-optimization/84178
8342 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8344 (version_loop_for_if_conversion): Delay update_ssa call.
8345 (tree_if_conversion): Delay update_ssa until after predicate
8348 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8350 PR tree-optimization/84178
8351 * tree-if-conv.c (release_bb_predicate): Remove the
8352 the assertion that the stmts have NULL use_ops.
8353 Discard the statements, asserting that they haven't
8354 yet been added to a BB.
8356 2018-03-08 Richard Biener <rguenther@suse.de>
8358 PR tree-optimization/84746
8359 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8360 (phi_translate): Pass in destination ANTIC_OUT set.
8361 (phi_translate_1): Likewise. For a simplified result lookup
8362 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8363 (phi_translate_set): Adjust.
8364 (do_pre_regular_insertion): Likewise.
8365 (do_pre_partial_partial_insertion): Likewise.
8367 2018-03-08 Martin Liska <mliska@suse.cz>
8369 PR gcov-profile/84735
8370 * doc/gcov.texi: Document usage of profile files.
8371 * gcov-io.h: Document changes in the format.
8373 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8377 * dwarf2out.c (struct dw_line_info_table): Update comments for
8379 (FORCE_RESET_NEXT_VIEW): New.
8380 (FORCE_RESETTING_VIEW_P): New.
8381 (RESETTING_VIEW_P): Check for -1 too.
8382 (ZERO_VIEW_P): Likewise.
8383 (new_line_info_table): Force-reset next view.
8384 (dwarf2out_begin_function): Likewise.
8385 (dwarf2out_source_line): Simplify zero_view_p initialization.
8386 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8387 view directly. Omit view when omitting .loc at line 0.
8389 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8391 PR tree-optimization/84740
8392 * tree-switch-conversion.c (process_switch): Call build_constructors
8393 only if info.phi_count is non-zero.
8395 PR tree-optimization/84739
8396 * tree-tailcall.c (find_tail_calls): Check call arguments against
8397 DECL_ARGUMENTS (current_function_decl) rather than
8398 DECL_ARGUMENTS (func) when checking for tail recursion.
8400 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8402 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8403 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8404 Volker Reichelt's entry and add entries for people that perform
8405 GCC fuzzy testing and report numerous bugs.
8407 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8410 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8411 readonly data in sdata, if that is disabled.
8412 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8413 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8414 -mreadonly-in-sdata option.
8416 2018-03-07 Martin Sebor <msebor@redhat.com>
8418 PR tree-optimization/84468
8419 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8420 basic block when looking for nul assignment.
8422 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8425 * except.h (output_function_exception_table): Adjust prototype.
8426 * except.c (output_function_exception_table): Remove FNNAME parameter
8427 and add SECTION parameter. Ouput one part of the table at a time.
8428 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8429 the first part of the exception table and emit unwind directives.
8430 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8431 (i386_pe_seh_cold_init): Likewise.
8432 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8433 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8434 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8435 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8436 * config/i386/winnt.c: Include except.h.
8437 (struct seh_frame_state): Add reg_offset, after_prologue and
8438 in_cold_section fields.
8439 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8440 (i386_pe_seh_cold_init): New function.
8441 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8442 to seh->in_cold_section.
8443 (seh_emit_push): Record the offset of the push.
8444 (seh_emit_save): Record the offet of the save.
8445 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8446 Test seh->after_prologue to disregard the epilogue.
8447 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8448 (i386_pe_end_cold_function): New function.
8450 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8453 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8454 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8457 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8459 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8460 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8463 * multiple_target.c: Include tree-inline.h and intl.h.
8464 (expand_target_clones): Diagnose and fail if node->definition and
8465 !tree_versionable_function_p (node->decl).
8467 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8469 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8471 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8472 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8473 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8475 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8478 * combine.c (try_combine): Use reg_or_subregno instead of handling
8479 just paradoxical SUBREGs and REGs.
8481 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8483 * config/arc/arc.c (arc_finalize_pic): Remove function.
8484 (arc_must_save_register): We use single base PIC register, remove
8485 checks to save/restore the PIC register.
8486 (arc_expand_prologue): Likewise.
8487 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8489 (arc_verify_short): Likewise.
8490 (arc_attr_type): Likewise.
8491 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8492 (walk_stores): Likewise.
8493 (arc_address_cost): Make it static.
8494 (arc_verify_short): Likewise.
8495 (branch_dest): Likewise.
8496 (arc_attr_type): Likewise.
8497 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8498 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8499 (arc_final_prescan_insn): Remove inserting the nops due to
8500 hardware hazards. It is done in reorg step.
8501 (insn_length_variant_t): Remove.
8502 (insn_length_parameters_t): Likewise.
8503 (arc_insn_length_parameters): Likewise.
8504 (arc_get_insn_variants): Likewise.
8505 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8507 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8510 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8513 PR tree-optimization/84687
8514 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8516 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8518 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8520 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8521 Rename to ppc_speculation_barrier.
8522 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8523 __builtin_ppc_speculation_barrier.
8525 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8528 * combine.c (combine_simplify_rtx): Don't try to simplify if
8529 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8532 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8534 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8535 to 32 bytes when compiling for POWER9.
8537 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8540 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8541 regparm >= 3 with no arg reg available also for calls with
8542 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8545 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8547 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8549 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8552 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8554 2018-03-05 Richard Biener <rguenther@suse.de>
8556 PR tree-optimization/84486
8557 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8558 When inserting a __builtin_assume_aligned call set the LHS
8559 SSA name alignment info accordingly.
8561 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8563 PR tree-optimization/84114
8564 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8565 Avoid reassociation of FLOAT_MODE addition.
8567 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8569 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8570 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8571 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8572 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8573 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8574 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8575 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8577 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8578 __builtin_ia32_wbinvd): New builtins.
8579 (SPECIAL_ARGS2): New.
8580 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8581 (SPECIAL_ARGS2): New.
8582 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8583 (ix86_valid_target_attribute_inner_p): Ditto.
8584 (ix86_init_mmx_sse_builtins): Add special_args2.
8585 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8586 TARGET_WBNOINVD_P): New.
8587 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8588 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8589 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8590 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8591 * config/i386/pconfigintrin.h: New file.
8592 * config/i386/wbnoinvdintrin.h: Ditto.
8593 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8595 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8597 2018-03-05 Richard Biener <rguenther@suse.de>
8599 PR tree-optimization/84670
8600 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8602 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8603 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8604 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8605 doesn't grow if all successors (recursively) were visited at least
8608 2018-03-05 Richard Biener <rguenther@suse.de>
8610 PR tree-optimization/84650
8611 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8612 if executed in the loop pipeline.
8614 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8616 * doc/configfiles.texi (Configuration Files): Move info about
8617 conditionalizing $target-protos.h to...
8618 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8619 differs from $target-protos.h.
8621 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8622 Chung-Ju Wu <jasonwucj@gmail.com>
8624 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8625 * config/nds32/nds32-multiple.md (setmemsi): Define.
8626 * config/nds32/nds32-memory-manipulation.c
8627 (nds32_gen_dup_4_byte_to_word_value): New.
8628 (emit_setmem_word_loop): New.
8629 (emit_setmem_byte_loop): New.
8630 (nds32_expand_setmem_loop): New.
8631 (nds32_expand_setmem_loop_v3m): New.
8632 (nds32_expand_setmem_unroll): New.
8633 (nds32_expand_setmem): New.
8635 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8636 Chung-Ju Wu <jasonwucj@gmail.com>
8638 * config/nds32/nds32-memory-manipulation.c
8639 (nds32_emit_load_store): New.
8640 (nds32_emit_post_inc_load_store): New.
8641 (nds32_emit_mem_move): New.
8642 (nds32_emit_mem_move_block): New.
8643 (nds32_expand_movmemsi_loop_unknown_size): New.
8644 (nds32_expand_movmemsi_loop_known_size): New.
8645 (nds32_expand_movmemsi_loop): New.
8646 (nds32_expand_movmemsi_unroll): New.
8647 (nds32_expand_movmemqi): Rename ...
8648 (nds32_expand_movmemsi): ... to this.
8649 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8650 (movmemsi): ... to this.
8651 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8652 (nds32_expand_movmemsi): ... to this.
8654 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8655 Monk Chiang <sh.chiang04@gmail.com>
8656 Chung-Ju Wu <jasonwucj@gmail.com>
8658 * config/nds32/nds32-protos.h
8659 (nds32_expand_load_multiple): New arguments.
8660 (nds32_expand_store_multiple): Ditto.
8661 (nds32_valid_multiple_load_store): Rename ...
8662 (nds32_valid_multiple_load_store_p): ... to this.
8663 * config/nds32/nds32-memory-manipulation.c
8664 (nds32_expand_load_multiple): Refine implementation.
8665 (nds32_expand_store_multiple): Ditto.
8666 * config/nds32/nds32-multiple.md
8667 (load_multiple): Update nds32_expand_load_multiple interface.
8668 (store_multiple): Update nds32_expand_store_multiple interface.
8669 * config/nds32/nds32-predicates.c
8670 (nds32_valid_multiple_load_store): Rename ...
8671 (nds32_valid_multiple_load_store_p): ... to this and refine
8673 * config/nds32/predicates.md
8674 (nds32_load_multiple_and_update_address_operation): New predicate.
8675 (nds32_store_multiple_and_update_address_operation): New predicate.
8677 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8678 Chung-Ju Wu <jasonwucj@gmail.com>
8680 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8681 (combo): New attribute.
8682 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8684 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8686 * config/nds32/nds32.opt: Change -mcmodel= default value.
8688 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8689 Monk Chiang <sh.chiang04@gmail.com>
8690 Chung-Ju Wu <jasonwucj@gmail.com>
8692 * config/nds32/constants.md (unspec_element): New enum.
8693 * config/nds32/constraints.md (Umw): New constraint.
8694 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8695 * config/nds32/nds32-intrinsic.md: Likewise.
8696 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8697 (nds32_valid_smw_lwm_base_p): New.
8698 (nds32_output_smw_single_word): New.
8699 (nds32_output_lmw_single_word): New.
8700 (nds32_expand_unaligned_load): New.
8701 (nds32_expand_unaligned_store): New.
8702 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8703 (nds32_output_smw_single_word): Declare.
8704 (nds32_output_lmw_single_word): Declare.
8705 (nds32_expand_unaligned_load): Declare.
8706 (nds32_expand_unaligned_store): Declare.
8707 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8708 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8709 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8710 NDS32_BUILTIN_UASTORE_DW.
8711 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8714 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8715 Kito Cheng <kito.cheng@gmail.com>
8716 Chung-Ju Wu <jasonwucj@gmail.com>
8718 * config/nds32/nds32-intrinsic.c
8719 (nds32_expand_builtin_null_ftype_reg): Delete.
8720 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8721 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8722 (nds32_read_argument): New.
8723 (nds32_legitimize_target): Ditto.
8724 (nds32_legitimize_argument): Ditto.
8725 (nds32_check_constant_argument): Ditto.
8726 (nds32_expand_unop_builtin): Ditto.
8727 (nds32_expand_unopimm_builtin): Ditto.
8728 (nds32_expand_binop_builtin): Ditto.
8729 (nds32_builtin_decl_impl): Ditto.
8730 (builtin_description): Ditto.
8731 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8732 (nds32_init_builtins_impl): Ditto.
8733 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8734 (nds32_builtin_decl): New.
8735 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8736 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8738 2018-03-02 Jeff Law <law@redhat.com>
8740 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8741 (redundant_insn, fill_simple_delay_slots): Likewise.
8742 (fill_slots_from_thread): Likewise.
8743 * resource.c (mark_referenced_resources): Likewise.
8744 (mark_set_resources, find_dead_or_set_registers): Likewise.
8746 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8748 * substring-locations.h (format_warning_va): Formatting fix for
8750 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8752 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8753 * substring-locations.c: Include intl.h.
8754 (format_warning_va): Turned into small wrapper around
8755 format_warning_n_va, renamed to ...
8756 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8757 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8759 (format_warning_at_substring_n): New function.
8760 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8761 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8762 format_warning_at_substring with just a shorter name instead of
8763 const function pointer.
8764 (fmtwarn_n): New function.
8765 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8766 appropriate, get rid of all the fmtstr temporaries, move conditionals
8767 with G_() wrapped string literals directly into fmtwarn arguments,
8768 cast dir.len to (int), formatting fixes.
8770 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8772 * doc/invoke.texi: Remove "Cilk Plus" references.
8774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8775 Richard Biener <rguenther@suse.de>
8778 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8779 for error or warning attributes if CALL_FROM_THUNK_P is set.
8782 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8785 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8786 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8789 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8790 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8792 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8793 instead of -1U in last predictors element's probability member.
8795 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8798 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8799 arguments if they are comparable.
8801 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8803 PR tree-optimization/84634
8804 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8805 masks and masked_loop_p with a single loop_masks, making sure it's
8806 null for bb vectorization.
8808 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8810 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8811 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8812 loop->force_vectorize to check whether there is no alias.
8814 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8817 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8819 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8821 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8822 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8823 instead of a loop around prev_real_insn.
8824 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8828 * config/i386/i386.c (ix86_print_operand): Use conditional
8829 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8832 2018-03-02 Richard Biener <rguenther@suse.de>
8834 PR tree-optimization/84427
8835 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8836 (bitmap_set_subtract_values): Rewrite to handle multiple
8839 (prune_clobbered_mems): Likewise.
8840 (phi_translate): Take edge instead of pred/phiblock.
8841 (phi_translate_1): Likewise.
8842 (phi_translate_set): Likewise. Insert all translated
8843 exprs for a value into the set, keeping possibly multiple
8844 expressions per value.
8845 (compute_antic_aux): Adjust for phi_translate changes.
8846 When intersecting union the expressions and prune those
8847 not in the final value set, keeping possibly multiple
8848 expressions per value. Do not use value-insertion
8849 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8850 all expressions. Add verification that the value-sets
8851 only shrink during iteration.
8852 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8853 (do_pre_regular_insertion): Likewise.
8854 (do_pre_partial_partial_insertion): Likewise.
8856 2018-03-02 Richard Biener <rguenther@suse.de>
8859 * config/darwin.c (saved_debug_info_level): New static global.
8860 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8861 (darwin_asm_lto_end): Restore debug info generation settings.
8863 2018-03-01 Martin Liska <mliska@suse.cz>
8866 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8869 2018-03-01 Richard Biener <rguenther@suse.de>
8872 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8873 type annotation with LTO when debug was disabled at compile-time.
8875 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8877 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8879 (mips_final_postscan_insn): Likewise.
8881 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8883 PR rtl-optimization/84528
8884 * alias.c (init_alias_target): Add commentary.
8885 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8886 a unique base value if the frame pointer is not eliminated
8887 to the stack pointer.
8889 2018-03-01 Tom de Vries <tom@codesourcery.com>
8891 PR rtl-optimization/83327
8892 * lra-int.h (hard_regs_spilled_into): Declare.
8893 * lra.c (hard_regs_spilled_into): Define.
8894 (init_reg_info): Init hard_regs_spilled_into.
8895 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8896 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8897 (process_bb_lives): Handle hard_regs_spilled_into.
8898 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8899 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8901 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8903 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8904 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8905 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8906 * config/rs6000/aix72.h: New file.
8908 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8910 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8911 instead of warning_at with conditional singular and plural messages
8915 * stor-layout.c (update_alignment_for_field): For
8916 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8917 && !DECL_PACKED (field), do the alignment update, just use
8918 only desired_align instead of MAX (type_align, desired_align)
8920 (place_field): Don't do known_align < desired_align handling
8921 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
8922 is non-NULL, instead do it after rli->prev_field handling and
8923 only if not within a bitfield word. For DECL_PACKED (field)
8924 use type_align of BITS_PER_UNIT.
8926 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
8928 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
8929 superfluous parentheses and trailing spaces.
8931 2018-02-28 Richard Biener <rguenther@suse.de>
8933 PR tree-optimization/84584
8934 * graphite-scop-detection.c (scop_detection::add_scop): Discard
8935 SCoPs with fake exit edge.
8937 2018-02-28 Martin Liska <mliska@suse.cz>
8940 * timevar.c (timer::print): Fix format to properly print 100%
8943 2018-02-28 Richard Biener <rguenther@suse.de>
8946 * genmatch.c (capture_info::walk_match): Do not mark
8947 captured expressions without operands as expr_p given
8948 they act more like predicates and should be subject to
8949 "lost tail" side-effect preserving.
8951 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
8953 PR rtl-optimization/81611
8954 * auto-inc-dec.c (attempt_change): Move dead note from
8955 mem_insn if it's the next use of regno
8956 (find_address): Take address use of reg holding
8957 non-incremented value. Add parm to limit search to the named
8959 (merge_in_block): Attempt to use a mem insn that is the next
8960 use of the original regno.
8962 2018-02-27 Martin Sebor <msebor@redhat.com>
8965 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
8966 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
8968 2018-02-27 Martin Sebor <msebor@redhat.com>
8970 PR translation/84207
8971 * diagnostic-core.h (warning_n, error_n, inform_n): Change
8972 n argument to unsigned HOST_WIDE_INT.
8973 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
8974 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
8975 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
8976 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
8978 2018-02-27 Richard Biener <rguenther@suse.de>
8980 PR tree-optimization/84512
8981 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8982 Do not use the estimate returned from record_stmt_cost for
8983 the scalar iteration cost but sum properly using add_stmt_cost.
8985 2018-02-27 Richard Biener <rguenther@suse.de>
8987 PR tree-optimization/84466
8988 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8989 Adjust last change to less strictly validate use operands.
8991 2018-02-27 Martin Liska <mliska@suse.cz>
8993 PR gcov-profile/84548
8994 * gcov.c (process_file): Allow partial overlap and consider it
8995 also as group functions.
8996 (output_lines): Properly calculate range of lines for a group.
8998 2018-02-27 Martin Liska <mliska@suse.cz>
9000 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9001 'ggc' suffixes. Change first column width.
9002 (timer::print): Fix formatting of the column.
9004 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9006 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9007 preserve inline entry blocks for the sake of debug inline
9008 entry point markers alone.
9009 (remove_unused_locals): Suggest in comments a better place to
9010 force the preservation of inline entry blocks that are
9011 otherwise unused, but do not preserve them.
9013 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9015 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9017 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9020 * config/i386/constraints.md (Bs): Replace
9021 ix86_indirect_branch_register with
9022 TARGET_INDIRECT_BRANCH_REGISTER.
9024 * config/i386/i386.md (indirect_jump): Likewise.
9025 (tablejump): Likewise.
9026 (*sibcall_memory): Likewise.
9027 (*sibcall_value_memory): Likewise.
9028 Peepholes of indirect call and jump via memory: Likewise.
9029 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9030 (*sibcall_value_GOT_32): Likewise.
9031 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9032 (GOT_memory_operand): Likewise.
9033 (call_insn_operand): Likewise.
9034 (sibcall_insn_operand): Likewise.
9035 (GOT32_symbol_operand): Likewise.
9036 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9038 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9040 PR rtl-optimization/83496
9041 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9042 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9043 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9044 redundant insn, if any.
9045 (relax_delay_slots): Likewise.
9046 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9048 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9050 PR tree-optimization/83965
9051 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9052 that grouped statements are part of a reduction chain. Return
9053 true if the statement is not marked as a reduction itself but
9055 (vect_recog_dot_prod_pattern): Don't check whether the statement
9056 is part of a group here.
9057 (vect_recog_sad_pattern): Likewise.
9058 (vect_recog_widen_sum_pattern): Likewise.
9060 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9063 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9065 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9068 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9070 (ix86_output_indirect_function_return): New prototype.
9071 (ix86_split_simple_return_pop_internal): Likewise.
9072 * config/i386/i386.c (indirect_return_via_cx): New.
9073 (indirect_return_via_cx_bnd): Likewise.
9074 (indirect_thunk_name): Handle return va CX_REG.
9075 (output_indirect_thunk_function): Create alias for
9076 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9077 (ix86_output_indirect_jmp): Remove the bool argument.
9078 (ix86_output_indirect_function_return): New function.
9079 (ix86_split_simple_return_pop_internal): Likewise.
9080 * config/i386/i386.md (*indirect_jump): Don't pass false
9081 to ix86_output_indirect_jmp.
9082 (*tablejump_1): Likewise.
9083 (simple_return_pop_internal): Change it to define_insn_and_split.
9084 Call ix86_split_simple_return_pop_internal to split it for
9086 (simple_return_indirect_internal): Call
9087 ix86_output_indirect_function_return instead of
9088 ix86_output_indirect_jmp.
9090 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9093 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9094 memset and value initialization afterwards.
9096 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9098 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9100 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9103 * common/config/aarch64/aarch64-common.c
9104 (aarch_option_optimization_table[]): Switch
9105 off fomit-frame-pointer
9107 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9108 Chung-Ju Wu <jasonwucj@gmail.com>
9110 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9112 (store_multiple): Ditto.
9114 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9116 * config.gcc: Add --with-cpu support for nds32 target.
9117 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9118 * config/nds32/nds32.opt: Add -mcpu= option.
9120 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9122 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9123 isel=yes): Warn for these deprecated options.
9125 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9127 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9128 ISA_2_5_MASKS_EMBEDDED.
9130 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9132 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9133 p->max as pointers rather than using iterative_hash_expr.
9135 2018-02-23 Carl Love <cel@us.ibm.com>
9137 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9138 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9140 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9141 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9142 P8V_BUILTIN_VEC_VUNSIGNED2.
9144 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9147 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9148 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9150 * lra-constraints.c (curr_insn_transform): Set up
9151 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9153 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9154 * lra-eliminations.c (spill_pseudos): Ditto.
9155 (process_insn_for_elimination): Ditto.
9156 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9157 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9158 LRA_NON_CLOBBERED_ALT.
9160 2018-02-22 Martin Sebor <msebor@redhat.com>
9162 PR tree-optimization/84480
9163 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9164 to maybe_diag_stxncpy_trunc. Call it.
9165 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9166 from gimple_fold_builtin_strcpy. Print inlining stack.
9167 (handle_builtin_stxncpy): Print inlining stack.
9168 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9170 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9173 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9174 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9175 and -fcheck-pointer-bounds are used together.
9176 (indirect_thunk_prefix): New enum.
9177 (indirect_thunk_need_prefix): New function.
9178 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9179 "_nt" instead of "_bnd" for NOTRACK prefix.
9180 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9181 (output_indirect_thunk_function): Likewise.
9183 (ix86_code_end): Update output_indirect_thunk_function calls.
9184 (ix86_output_indirect_branch_via_reg): Replace
9185 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9186 (ix86_output_indirect_branch_via_push): Likewise.
9187 (ix86_output_function_return): Likewise.
9188 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9189 incompatible with -fcf-protection=branch and
9190 -fcheck-pointer-bounds.
9192 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9195 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9196 Change gcc_assert call to output_operand_lossage.
9198 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9200 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9202 2018-02-22 DJ Delorie <dj@redhat.com>
9203 Sebastian Perta <sebastian.perta@renesas.com>
9204 Oleg Endo <olegendo@gcc.gnu.org>
9206 * config/rx/rx.c (rx_rtx_costs): New function.
9207 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9209 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9211 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9213 2018-02-22 Martin Liska <mliska@suse.cz>
9216 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9217 Add "native" as a possible value.
9219 2018-02-22 Martin Liska <mliska@suse.cz>
9222 * config/i386/i386.c (ix86_option_override_internal):
9223 Add "native" as a possible value for -march and -mtune.
9225 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9228 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9229 to all type variants.
9231 PR tree-optimization/84503
9232 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9233 width as info->bitpos + info->bitsize - start.
9234 (merged_store_group::merge_overlapping): Simplify width computation.
9235 (check_no_overlap): New function.
9236 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9237 start + width and last_order of the group, fail if check_no_overlap
9239 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9240 to group if check_no_overlap fails.
9242 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9244 * config/rs6000/altivec.md: Delete contraint arguments to
9245 define_expand, define_split, and define_peephole2, and in
9246 define_insn_and_split if always unused.
9247 * config/rs6000/darwin.md: Ditto.
9248 * config/rs6000/dfp.md: Ditto.
9249 * config/rs6000/rs6000.md: Ditto.
9250 * config/rs6000/sync.md: Ditto.
9251 * config/rs6000/vector.md: Ditto.
9252 * config/rs6000/vsx.md: Ditto.
9254 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9256 * config/rs6000/altivec.md: Write output control strings as braced
9257 blocks instead of double-quoted strings.
9258 * config/rs6000/darwin.md: Ditto.
9259 * config/rs6000/rs6000.md: Ditto.
9260 * config/rs6000/vector.md: Ditto.
9261 * config/rs6000/vsx.md: Ditto.
9263 2018-02-21 Jason Merrill <jason@redhat.com>
9265 PR c++/84314 - ICE with templates and fastcall attribute.
9266 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9268 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9270 * ipa-cp.c (determine_versionability): Fix comment typos.
9272 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9275 * ipa-cp.c (determine_versionability): Do not version functions caling
9278 2018-02-21 Martin Liska <mliska@suse.cz>
9281 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9282 Add "native" as a possible value.
9283 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9284 the macro when native cpu detection is available.
9286 2018-02-21 Martin Liska <mliska@suse.cz>
9289 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9290 Add "native" as a possible value.
9291 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9292 when native cpu detection is available.
9294 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9295 Martin Sebor <msebor@redhat.com>
9297 PR tree-optimization/84478
9298 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9300 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9301 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9302 support which is conservatively correct, for 2 only stay conservative
9303 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9304 argument to the 2 argument get_range_strlen, adjust 6 arg
9305 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9307 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9308 (gimple_fold_builtin_strlen): Pass true as last argument to
9311 2018-02-20 Martin Sebor <msebor@redhat.com>
9314 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9315 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9316 (builtin_memref::builtin_memref): Factor out parts into
9317 set_base_and_offset and call it.
9319 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9322 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9323 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9324 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9325 search at the associated MODE_INT.
9327 2018-02-20 Jeff Law <law@redhat.com>
9330 PR tree-optimization/81592
9332 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9333 for range data rather than using global data.
9334 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9335 range data rather than using global data.
9336 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9337 pass it to children as needed.
9338 (struct directive::fmtresult): Similarly.
9339 (struct directive::set_width): Similarly.
9340 (struct directive::set_precision): Similarly.
9341 (format_integer, format_directive, parse_directive): Similarly.
9342 (format_none): Accept unnamed vr_values parameter.
9343 (format_percent, format_floating, format_character): Similarly.
9344 (format_string, format_plain): Similarly.
9345 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9346 the EVRP range analyzer for range data rather than using global data.
9347 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9348 gimple-ssa-evrp-analyze.h
9349 (class sprintf_dom_walker): Add after_dom_children member function.
9350 Add evrp_range_analyzer member.
9351 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9352 range analyzer as needed.
9353 (sprintf_dom_walker::after_dom_children): New member function.
9354 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9356 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9357 (evrp_range_analyzer::pop_to_marker): Likewise.
9359 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9361 PR tree-optimization/84419
9362 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9363 with the required type if its current type is compatible but
9366 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9369 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9370 after vectorization.
9372 2018-02-20 Martin Liska <mliska@suse.cz>
9375 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9376 possible values if we don't have a hint.
9378 2018-02-20 Martin Liska <mliska@suse.cz>
9382 * final.c (shorten_branches): Build align_tab array with one
9384 * opts.c (finish_options): Add alignment option limit check.
9385 (MAX_CODE_ALIGN): Likewise.
9386 (MAX_CODE_ALIGN_VALUE): Likewise.
9387 * doc/invoke.texi: Document maximum allowed option value for
9388 all -falign-* options.
9390 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9393 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9394 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9395 * var-tracking.c (emit_note_insn_var_location): Remove all references
9396 to NOTE_INSN_CALL_ARG_LOCATION.
9397 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9398 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9399 Use copy_rtx_if_shared.
9400 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9401 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9402 (dwarf2out_var_location): Remove handling of
9403 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9405 * final.c (final_scan_insn): Remove all references to
9406 NOTE_INSN_CALL_ARG_LOCATION.
9407 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9408 before dumping final insns.
9409 * except.c (emit_note_eh_region_end): Remove all references to
9410 NOTE_INSN_CALL_ARG_LOCATION.
9411 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9412 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9413 * config/arc/arc.c (hwloop_optimize): Likewise.
9414 * config/arm/arm.c (create_fix_barrier): Likewise.
9415 * config/s390/s390.c (s390_chunkify_start): Likewise.
9416 * config/sh/sh.c (find_barrier): Likewise.
9417 * config/i386/i386.c (rest_of_insert_endbranch,
9418 ix86_seh_fixup_eh_fallthru): Likewise.
9419 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9420 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9421 * config/frv/frv.c (frv_function_prologue): Likewise.
9422 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9424 (note_outside_basic_block_p): Remove all references to
9425 NOTE_INSN_CALL_ARG_LOCATION.
9426 * gengtype.c (adjust_field_rtx_def): Likewise.
9427 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9429 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9430 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9433 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9436 PR tree-optimization/84452
9437 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9438 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9441 2018-02-19 Martin Liska <mliska@suse.cz>
9444 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9446 2018-02-19 Martin Liska <mliska@suse.cz>
9447 Richard Sandiford <richard.sandiford@linaro.org>
9449 PR tree-optimization/82491
9450 * gimple-fold.c (get_base_constructor): Make earlier bail out
9453 2018-02-19 Carl Love <cel@us.ibm.com>
9455 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9456 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9458 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9459 P8V_BUILTIN_VEC_NEG.
9461 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9463 * config/rl78/rl78.md (movdf): New define expand.
9465 2018-02-19 Martin Liska <mliska@suse.cz>
9468 * doc/invoke.texi: Fix typo.
9469 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9471 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9473 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9474 handle rs6000_single_float and rs6000_double_float specially for
9477 2018-02-16 Jeff Law <law@redhat.com>
9479 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9482 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9485 * ipa-inline.c (inline_small_functions): Fix a typo.
9487 2018-02-16 Nathan Sidwell <nathan@acm.org>
9489 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9491 2018-02-16 Carl Love <cel@us.ibm.com>
9493 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9494 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9495 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9496 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9497 expansion to P8V_BUILTIN_VEC_FLOAT2.
9499 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9501 PR rtl-optimization/70023
9502 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9503 src_regno into account.
9505 2018-02-16 Carl Love <cel@us.ibm.com>
9507 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9508 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9509 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9510 * config/rs6000/rs6000.c: Remove case statements for
9511 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9512 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9513 and P9V_BUILTIN_VEC_VINSERT4B.
9514 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9515 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9516 * config/rs6000/vsx.md:
9517 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9520 2018-02-16 Carl Love <cel@us.ibm.com>
9522 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9524 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9526 * config/rs6000/rs6000-c.c: Add the definitions for
9527 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9528 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9529 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9530 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9531 definition for insert4b and define insn *insert3b_internal.
9532 * doc/extend.texi: Add documentation for vec_extract4b.
9534 2018-02-16 Nathan Sidwell <nathan@acm.org>
9536 * doc/extend.texi (Backwards Compatibility): Mention friend
9537 injection. Note for-scope is deprecated.
9538 * doc/invoke.texi (-ffriend-injection): Deprecate.
9540 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9542 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9543 that moved to I2, also allow destinations that are a paradoxical
9544 subreg (instead of a normal reg).
9546 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9549 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9552 2018-02-16 Richard Biener <rguenther@suse.de>
9554 PR tree-optimization/84037
9555 PR tree-optimization/84016
9557 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9558 Adjust vec_construct for the fact we need additional higher latency
9559 128bit inserts for AVX256 and AVX512 vector builds.
9560 (ix86_add_stmt_cost): Scale vector construction cost for
9563 2018-02-16 Richard Biener <rguenther@suse.de>
9565 PR tree-optimization/84417
9566 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9567 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9568 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9570 2018-02-16 Martin Liska <mliska@suse.cz>
9573 * internal-fn.def (ASAN_CHECK): Set proper flags.
9574 (ASAN_MARK): Likewise.
9576 2018-02-16 Julia Koval <julia.koval@intel.com>
9578 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9579 from PTA_CANNONLAKE.
9581 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9584 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9585 Use ++iter rather than iter++ for std::list iterators.
9586 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9587 defer deleting them until all nodes in the forest are processed. Do
9588 free even leaf nodes. Change to_process into auto_vec.
9591 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9592 * vec.h (vec_default_construct): Use memset instead of placement new
9593 if BROKEN_VALUE_INITIALIZATION is defined.
9594 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9595 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9598 PR rtl-optimization/83723
9599 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9600 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9601 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9603 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9605 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9607 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9609 PR rtl-optimization/81443
9610 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9611 from inner REGs to paradoxical SUBREGs.
9613 2018-02-16 Richard Biener <rguenther@suse.de>
9615 PR tree-optimization/84399
9616 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9617 For operands we can analyze at their definition make sure we can
9618 analyze them at each use as well.
9620 2018-02-16 Richard Biener <rguenther@suse.de>
9622 PR tree-optimization/84190
9623 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9624 volatile accesses if the decl isn't volatile.
9626 2018-02-15 Jason Merrill <jason@redhat.com>
9628 PR c++/84314 - ICE with templates and fastcall attribute.
9629 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9630 TYPE_CANONICAL on an existing type.
9632 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9634 PR tree-optimization/84383
9635 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9636 dstoff nor call operand_equal_p if dstbase is NULL.
9638 PR tree-optimization/84334
9639 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9640 also a CONSTANT_CLASS_P, punt.
9642 2018-02-14 Jim Wilson <jimw@sifive.com>
9644 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9645 first SMALL_OPERAND check. New local min_second_step. Move assert
9646 to where locals are set. Add TARGET_RVC support.
9647 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9649 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9651 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9653 2018-02-14 Martin Sebor <msebor@redhat.com>
9655 PR tree-optimization/83698
9656 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9657 arrays constrain the offset range to their bounds.
9658 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9659 (builtin_access::overlap): Avoid setting the size of overlap if it's
9661 (maybe_diag_overlap): Also consider arrays when deciding what values
9662 of offsets to include in diagnostics.
9664 2018-02-14 Martin Sebor <msebor@redhat.com>
9667 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9668 that correspond to the kind of a declaration.
9670 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9673 * config/pa/pa.md: Load address of PIC label using the linkage table
9674 if the label is nonlocal.
9676 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9678 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9679 warning message if user requests -maltivec=be.
9680 * doc/invoke.texi: Document deprecation of -maltivec=be.
9682 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9685 * config/rs6000/rs6000-c.c: Update definitions for
9686 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9687 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9689 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9692 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9693 add _get_ssp intrinsics. Remove argument from
9694 __builtin_ia32_rdssp[d|q].
9695 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9696 * config/i386/i386-builtin.def: Remove argument from
9697 __builtin_ia32_rdssp[d|q].
9698 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9699 ix86_expand_special_args_builtin for _rdssp[d|q].
9700 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9701 Clear register before usage.
9702 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9703 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9705 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9707 PR tree-optimization/84357
9708 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9709 operand 1 of an ARRAY_REF too.
9711 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9714 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9715 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9717 (set_of_reg): New struct.
9718 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9719 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9720 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9722 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9723 Split into bitclr, bitset, bitinvert patterns if appropriate.
9724 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9725 use rx_fuse_in_memory_bitop.
9726 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9727 to named insn, correct maximum insn length.
9729 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9732 * machmode.def: Define a complex mode for PARTIAL_INT.
9733 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9735 * doc/rtl.texi: Document CSPImode.
9736 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9738 (msp430_hard_regno_nregs_with_padding): Likewise.
9740 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9743 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9745 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9747 PR rtl-optimization/84169
9748 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9749 we generated a parallel as new i3 and we split that to new i2 and i3
9750 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9751 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9752 those to i2, not i1. Partially rewrite this scan code.
9754 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9757 * stor-layout.c (place_field): For variable length fields, adjust
9758 offset_align afterwards not just based on the field's alignment,
9759 but also on the size.
9762 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9763 of exps and logs in the use_exp2 case.
9765 2018-02-13 Jeff Law <law@redhat.com>
9767 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9770 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9773 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9777 * common.opt (gas-loc-support, gas-locview-support): New.
9778 (ginline-points, ginternal-reset-location-views): New.
9779 * doc/invoke.texi: Document them. Use @itemx where intended.
9780 (gvariable-location-views): Adjust.
9781 * target.def (reset_location_view): New.
9782 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9783 (TARGET_RESET_LOCATION_VIEW): New.
9784 * doc/tm.texi: Rebuilt.
9785 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9786 (dwarf2out_default_as_locview_support): New.
9787 (output_asm_line_debug_info): Use option variables.
9788 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9789 (output_loc_list): Likewise.
9790 (add_high_low_attributes): Check option variables.
9791 Don't output entry view attribute in strict mode.
9792 (gen_inlined_subroutine_die): Check option variables.
9793 (dwarf2out_inline_entry): Likewise.
9794 (init_sections_and_labels): Likewise.
9795 (dwarf2out_early_finish): Likewise.
9796 (maybe_reset_location_view): New, from...
9797 (dwarf2out_var_location): ... here. Call it.
9798 * debug.h (dwarf2out_default_as_loc_support): Declare.
9799 (dwarf2out_default_as_locview_support): Declare.
9800 * hooks.c (hook_int_rtx_insn_0): New.
9801 * hooks.h (hook_int_rtx_insn_0): Declare.
9802 * toplev.c (process_options): Take -gas-loc-support and
9803 -gas-locview-support from dwarf2out. Enable
9804 -gvariable-location-views by default only with locview
9805 assembler support. Enable -ginternal-reset-location-views by
9806 default only if the target defines the corresponding hook.
9807 Enable -ginline-points by default if location views are
9808 enabled; force it disabled if statement frontiers are
9810 * tree-inline.c (expand_call_inline): Check option variables.
9811 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9813 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9815 PR tree-optimization/84321
9816 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9817 handling. Also check whether the anti-range contains any values
9818 that satisfy the mask; switch to a VR_RANGE if not.
9820 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9823 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9825 2018-02-13 Martin Jambor <mjambor@suse.cz>
9828 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9829 of call statements, also set location of a load to a temporary.
9831 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9833 * config/rl78/rl78.c (add_vector_labels): New function.
9834 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9835 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9836 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9837 which checks that no arguments are passed.
9838 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9839 * doc/extend.texi: Documentation for the new attribute.
9841 2018-02-13 Andreas Schwab <schwab@suse.de>
9843 * config/riscv/linux.h (CPP_SPEC): Define.
9845 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9848 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9849 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9850 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9851 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9852 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9853 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9854 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9855 temporarily for AES and PCLMUL builtins.
9857 PR tree-optimization/84339
9858 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9859 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9863 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9864 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9865 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9867 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9869 * omp-simd-clone.h: New file.
9870 * omp-simd-clone.c: Include omp-simd-clone.h.
9871 (expand_simd_clones): No longer static.
9872 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9873 cgraph.h and omp-simd-clone.h.
9874 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9875 (vect_recog_widen_shift_pattern): Formatting fix.
9876 (vect_pattern_recog_1): Don't check optab for calls.
9879 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9880 operands[2] into a REG before using gen_lowpart on it.
9882 2018-02-12 Jeff Law <law@redhat.com>
9885 * config/sh/sh.c (find_barrier): Consider a sibling call
9888 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9889 successfully back substituting a reg.
9891 2018-02-12 Richard Biener <rguenther@suse.de>
9893 PR tree-optimization/84037
9894 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9895 parameter, move visited init to caller.
9896 (vect_slp_analyze_operations): Separate cost from validity
9897 check, initialize visited once for all instances.
9898 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9900 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9902 (vect_model_promotion_demotion_cost): Likewise.
9903 (vectorizable_bswap): Guard cost modeling with !slp_node
9904 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9906 (vectorizable_call): Likewise.
9907 (vectorizable_conversion): Likewise.
9908 (vectorizable_assignment): Likewise.
9909 (vectorizable_shift): Likewise.
9910 (vectorizable_operation): Likewise.
9911 (vectorizable_store): Likewise.
9912 (vectorizable_load): Likewise.
9913 (vectorizable_condition): Likewise.
9914 (vectorizable_comparison): Likewise.
9916 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9919 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9920 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9921 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
9923 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
9926 * params.def (inline-min-speedup): Increase from 8 to 15.
9927 (max-inline-insns-auto): Decrease from 40 to 30.
9928 * ipa-split.c (consider_split): Add some buffer for function to
9929 be considered inlining candidate.
9930 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
9933 2018-02-12 Richard Biener <rguenther@suse.de>
9935 PR tree-optimization/84037
9936 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
9937 matched stmts if we cannot swap the non-matched ones.
9939 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9941 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
9942 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
9943 _mm_maskz_scalef_round_ss): New intrinsics.
9944 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
9945 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
9946 __builtin_ia32_scalefss_round): Remove.
9947 (__builtin_ia32_scalefsd_mask_round,
9948 __builtin_ia32_scalefss_mask_round): New intrinsics.
9949 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
9950 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
9951 ((match_operand:VF_128 2 "<round_nimm_predicate>"
9952 "<round_constraint>")): Changed to ...
9953 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
9954 "<round_scalar_constraint>")): ... this.
9955 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
9956 %0, %1, %2<round_op3>}"): Changed to ...
9957 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
9958 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
9959 %2<round_scalar_mask_op3>}"): ... this.
9960 * config/i386/subst.md (round_scalar_nimm_predicate): New.
9962 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
9964 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
9965 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
9966 (_mm_maskz_sqrt_round_ss): New intrinsics.
9967 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
9968 (__builtin_ia32_sqrtsd_mask_round)
9969 (__builtin_ia32_sqrtss_mask_round): New builtins.
9970 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
9971 (__builtin_ia32_sqrtss_round): Remove.
9972 (__builtin_ia32_sqrtsd_mask_round)
9973 (__builtin_ia32_sqrtss_mask_round): New builtins.
9974 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
9975 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
9976 ((match_operand:VF_128 1 "vector_operand"
9977 "xBm,<round_constraint>")): Changed to ...
9978 ((match_operand:VF_128 1 "vector_operand"
9979 "xBm,<round_scalar_constraint>")): ... this.
9980 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
9981 %0, %2, %<iptr>1<round_op3>}): Changed to ...
9982 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
9983 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
9984 %<iptr>1<round_scalar_mask_op3>}): ... this.
9985 ((set_attr "prefix" "<round_prefix>")): Changed to ...
9986 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
9988 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
9991 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
9992 Cast vec_cmpeq result to correct type.
9993 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
9994 Cast vec_cmpgt result to correct type.
9996 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
9998 * final.c (final_scan_insn_1): Renamed from...
9999 (final_scan_insn): ... this. New wrapper, to recover
10000 seen from the outermost call in recursive ones.
10001 * config/sparc/sparc.c (output_return): Drop seen from call.
10002 (output_sibcall): Likewise.
10003 * config/visium/visium.c (output_branch): Likewise.
10005 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10007 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10010 2018-02-10 Alan Modra <amodra@gmail.com>
10013 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10014 Specify LR as an input.
10016 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10019 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10020 remove_member_access_dummy_vars): New functions.
10021 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10022 lower_omp_1, execute_lower_omp): Use them.
10024 PR rtl-optimization/84308
10025 * shrink-wrap.c (spread_components): Release todo vector.
10027 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10029 PR rtl-optimization/57193
10030 * ira-color.c (struct allocno_color_data): Add member
10031 conflict_allocno_hard_prefs.
10032 (update_conflict_allocno_hard_prefs): New.
10033 (bucket_allocno_compare_func): Add a preference based on
10034 conflict_allocno_hard_prefs.
10035 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10036 (color_allocnos): Remove a dead code. Initiate
10037 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10039 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10042 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10043 constraint from =wa to wa. Avoid a subreg on the output operand,
10044 instead use a pseudo and subreg it in a move.
10045 (p9_xxbrd_<mode>): Changed to ...
10046 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10047 (p9_xxbrd_v2df): New expander.
10048 (p9_xxbrw_<mode>): Changed to ...
10049 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10050 (p9_xxbrw_v4sf): New expander.
10052 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10054 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10056 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10059 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10060 multiply in 32-bit mode.
10061 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10062 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10065 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10067 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10068 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10069 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10070 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10072 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10075 * dwarf2out.c (is_trivial_indirect_ref): New function.
10076 (dwarf2out_late_global_decl): Do not generate a location
10077 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10078 and that are not defined in the current unit.
10080 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10082 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10083 instead of a libcall for UNORDERED.
10085 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10088 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10089 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10091 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10094 * config/s390/s390.c (s390_set_current_function): Invoke
10095 s390_indirect_branch_settings also if fndecl didn't change.
10097 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10099 * config/rs6000/rs6000.md (blockage): Set length to zero.
10101 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10103 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10105 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10108 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10109 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10113 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10114 PARALLEL incoming that failed vt_get_decl_and_offset check.
10116 PR middle-end/84237
10117 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10118 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10120 (get_variable_section): For decls in named .bss* sections pass true as
10121 second argument to bss_initializer_p.
10123 2018-02-09 Marek Polacek <polacek@redhat.com>
10124 Jakub Jelinek <jakub@redhat.com>
10127 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10128 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10129 Sync some changes from cxx_fold_indirect_ref.
10131 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10133 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10135 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10136 (BLOCK_INLINE_ENTRY_LABEL): New.
10137 (dwarf2out_var_location): Disregard inline entry markers.
10138 (inline_entry_data): New struct.
10139 (inline_entry_data_hasher): New hashtable type.
10140 (inline_entry_data_hasher::hash): New.
10141 (inline_entry_data_hasher::equal): New.
10142 (inline_entry_data_table): New variable.
10143 (add_high_low_attributes): Add DW_AT_entry_pc and
10144 DW_AT_GNU_entry_view attributes if a pending entry is found
10145 in inline_entry_data_table. Add old entry_pc attribute only
10146 if debug nonbinding markers are disabled.
10147 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10148 markers are enabled.
10149 (block_within_block_p, dwarf2out_inline_entry): New.
10150 (dwarf2out_finish): Check that no entries remained in
10151 inline_entry_data_table.
10152 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10153 (final_scan_insn, notice_source_line): Likewise.
10154 (rest_of_clean_state): Skip inline entry markers.
10155 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10157 * gimple.c (gimple_build_debug_inline_entry): New.
10158 * gimple.h (enum gimple_debug_subcode): Add
10159 GIMPLE_DEBUG_INLINE_ENTRY.
10160 (gimple_build_debug_inline_entry): Declare.
10161 (gimple_debug_inline_entry_p): New.
10162 (gimple_debug_nonbind_marker_p): Adjust.
10163 * insn-notes.def (INLINE_ENTRY): New.
10164 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10165 inline entry marker notes.
10166 (print_insn): Likewise.
10167 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10168 (INSN_DEBUG_MARKER_KIND): Likewise.
10169 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10170 * tree-inline.c (expand_call_inline): Build and insert
10171 debug_inline_entry stmt.
10172 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10173 inline entry blocks early, if nonbind markers are enabled.
10174 (dump_scope_block): Dump fragment info.
10175 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10176 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10177 (gimple_build_debug_inline_entry): New.
10178 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10179 Enable/disable inline entry points too.
10180 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10181 (DEBUG_INSN): Describe inline entry markers.
10183 * common.opt (gvariable-location-views): New.
10184 (gvariable-location-views=incompat5): New.
10185 * config.in: Rebuilt.
10186 * configure: Rebuilt.
10187 * configure.ac: Test assembler for view support.
10188 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10189 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10190 * dwarf2out.c (var_loc_view): New typedef.
10191 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10192 (dwarf2out_locviews_in_attribute): New.
10193 (dwarf2out_locviews_in_loclist): New.
10194 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10195 (enum dw_line_info_opcode): Add LI_adv_address.
10196 (struct dw_line_info_table): Add view.
10197 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10198 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10199 (zero_view_p): New variable.
10200 (ZERO_VIEW_P): New macro.
10201 (output_asm_line_debug_info): New.
10202 (struct var_loc_node): Add view.
10203 (add_AT_view_list, AT_loc_list): New.
10204 (add_var_loc_to_decl): Add view param. Test it against last.
10205 (new_loc_list): Add view params. Record them.
10206 (AT_loc_list_ptr): Handle loc and view lists.
10207 (view_list_to_loc_list_val_node): New.
10208 (print_dw_val): Handle dw_val_class_view_list.
10209 (size_of_die): Likewise.
10210 (value_format): Likewise.
10211 (loc_list_has_views): New.
10212 (gen_llsym): Set vl_symbol too.
10213 (maybe_gen_llsym, skip_loc_list_entry): New.
10214 (dwarf2out_maybe_output_loclist_view_pair): New.
10215 (output_loc_list): Output view list or entries too.
10216 (output_view_list_offset): New.
10217 (output_die): Handle dw_val_class_view_list.
10218 (output_dwarf_version): New.
10219 (output_compilation_unit_header): Use it.
10220 (output_skeleton_debug_sections): Likewise.
10221 (output_rnglists, output_line_info): Likewise.
10222 (output_pubnames, output_aranges): Update version comments.
10223 (output_one_line_info_table): Output view numbers in asm comments.
10224 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10225 Call maybe_gen_llsym.
10226 (loc_list_from_tree_1): Adjust.
10227 (add_AT_location_description): Create view list attribute if
10228 needed, check it's absent otherwise.
10229 (convert_cfa_to_fb_loc_list): Adjust.
10230 (maybe_emit_file): Call output_asm_line_debug_info for test.
10231 (dwarf2out_var_location): Reset views as needed. Precompute
10232 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10233 attribute. Set view.
10234 (new_line_info_table): Reset next view.
10235 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10236 (dwarf2out_source_line): Likewise. Output view resets and labels to
10237 the assembler, or select appropriate line info opcodes.
10238 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10239 (optimize_string_length): Catch it. Adjust.
10240 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10241 dw_val_class_view_list, and remove it if no longer needed.
10242 (hash_loc_list): Hash view numbers.
10243 (loc_list_hasher::equal): Compare them.
10244 (optimize_location_lists): Check whether a view list symbol is
10245 needed, and whether the locview attribute is present, and
10246 whether they match. Remove the locview attribute if no longer
10248 (index_location_lists): Call skip_loc_list_entry for test.
10249 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10250 Use output_dwarf_version.
10251 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10252 (struct dw_val_node): Add val_view_list.
10253 * final.c (SEEN_NEXT_VIEW): New.
10254 (set_next_view_needed): New.
10255 (clear_next_view_needed): New.
10256 (maybe_output_next_view): New.
10257 (final_start_function): Rename to...
10258 (final_start_function_1): ... this. Take pointer to FIRST,
10259 add SEEN parameter. Emit param bindings in the initial view.
10260 (final_start_function): Reintroduce SEEN-less interface.
10261 (final): Rename to...
10262 (final_1): ... this. Take SEEN parameter. Output final pending
10263 next view at the end.
10264 (final): Reintroduce seen-less interface.
10265 (final_scan_insn): Output pending next view before switching
10266 sections or ending a block. Mark the next view as needed when
10267 outputting variable locations. Notify debug backend of section
10268 changes, and of location view changes.
10269 (rest_of_handle_final): Adjust.
10270 * toplev.c (process_options): Autodetect value for debug variable
10271 location views option. Warn on incompat5 without -gdwarf-5.
10272 * doc/invoke.texi (gvariable-location-views): New.
10273 (gvariable-location-views=incompat5): New.
10274 (gno-variable-location-views): New.
10276 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10278 PR tree-optimization/84136
10279 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10280 that the result of find_edge is non-NULL.
10282 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10285 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10286 storing integer register in SImode. Fix cost of 256 and 512
10287 byte aligned SSE register store.
10289 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10291 * config/i386/i386.c (ix86_multiplication_cost): Fix
10292 multiplication cost for TARGET_AVX512DQ.
10294 2018-02-08 Marek Polacek <polacek@redhat.com>
10296 PR tree-optimization/84238
10297 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10300 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10302 PR tree-optimization/84265
10303 * tree-vect-stmts.c (vectorizable_store): Don't treat
10304 VMAT_CONTIGUOUS accesses as grouped.
10305 (vectorizable_load): Likewise.
10307 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10309 PR tree-optimization/81635
10310 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10311 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10312 (test_round_for_mask): New functions.
10313 (wide_int_cc_tests): Call test_round_for_mask.
10314 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10315 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10316 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10317 range returned by get_range_info.
10319 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10322 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10323 * symtab.c: Include builtins.h
10324 (symtab_node::output_to_lto_symbol_table_p): Move here
10325 from lto-streamer-out.c:output_symbol_p.
10326 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10327 (output_symbol_p): Move all logic to symtab.c
10328 (produce_symtab): Update.
10330 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10332 * config/s390/s390-opts.h (enum indirect_branch): Define.
10333 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10334 (s390_indirect_branch_via_thunk)
10335 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10336 (enum s390_indirect_branch_type): Define.
10337 * config/s390/s390.c (struct s390_frame_layout, struct
10338 machine_function): Remove.
10339 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10340 (indirect_branch_table_label_no, indirect_branch_table_name):
10342 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10343 (enum s390_indirect_branch_option): Define.
10344 (s390_return_addr_from_memory): New function.
10345 (s390_handle_string_attribute): New function.
10346 (s390_attribute_table): Add new attribute handler.
10347 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10348 (s390_indirect_branch_via_thunk): New function.
10349 (s390_indirect_branch_via_inline_thunk): New function.
10350 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10351 sibling call optimization for non z10 compiles.
10352 (s390_emit_call): Force indirect branch target to be a single
10353 register. Add r1 clobber for non-z10 compiles.
10354 (s390_emit_epilogue): Emit return jump via return_use expander.
10355 (s390_reorg): Handle JUMP_INSNs as execute targets.
10356 (s390_option_override_internal): Perform validity checks for the
10357 new command line options.
10358 (s390_indirect_branch_attrvalue): New function.
10359 (s390_indirect_branch_settings): New function.
10360 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10361 (s390_output_indirect_thunk_function): New function.
10362 (s390_code_end): Implement target hook.
10363 (s390_case_values_threshold): Implement target hook.
10364 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10366 * config/s390/s390.h (struct s390_frame_layout)
10367 (struct machine_function): Move here from s390.c.
10368 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10369 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10370 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10371 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10372 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10373 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10374 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10375 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10376 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10377 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10378 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10379 (mnemonic attribute): Add values which aren't recognized
10381 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10382 pattern for branch conversion. Fix mnemonic attribute.
10383 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10384 indirect branch via thunk if requested.
10385 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10386 ("*indirect_jump"): Disable for branch conversion using out of
10388 ("indirect_jump_via_thunk<mode>_z10")
10389 ("indirect_jump_via_thunk<mode>")
10390 ("indirect_jump_via_inlinethunk<mode>_z10")
10391 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10392 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10393 ("casesi_jump_via_inlinethunk<mode>_z10")
10394 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10395 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10396 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10397 ("*indirect2_jump"): Disable for branch conversion.
10398 ("casesi_jump"): Turn into expander and expand patterns for branch
10400 ("return_use"): New expander.
10401 ("*return"): Emit return via thunk and rename it to ...
10402 ("*return<mode>"): ... this one.
10403 * config/s390/s390.opt: Add new options and and enum for the
10406 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10408 * lra-constraints.c (match_reload): Unconditionally use
10409 gen_lowpart_SUBREG, rather than selecting between that
10410 and equivalent gen_rtx_SUBREG code.
10412 2018-02-08 Richard Biener <rguenther@suse.de>
10414 PR tree-optimization/84233
10415 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10416 changed flag instead of boguously re-using phi_inserted.
10418 2018-02-08 Martin Jambor <mjambor@suse.cz>
10420 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10421 static local variables.
10423 2018-02-08 Richard Biener <rguenther@suse.de>
10425 PR tree-optimization/84278
10426 * tree-vect-stmts.c (vectorizable_store): When looking for
10427 smaller vector types to perform grouped strided loads/stores
10428 make sure the mode is supported by the target.
10429 (vectorizable_load): Likewise.
10431 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10433 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10434 Increase LDP/STP opportunities by adding adjacent callee-saves.
10436 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10438 PR rtl-optimization/84068
10439 PR rtl-optimization/83459
10440 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10442 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10444 PR tree-optimization/84224
10445 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10446 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10447 non-zero arguments.
10449 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10452 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10453 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10454 count, remove one USE.
10456 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10458 * doc/install.texi (Configuration): Document the
10459 --with-long-double-format={ibm,ieee} PowerPC configuration
10463 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10464 Convert from define_expand to be define_insn_and_split. Rework
10465 float/double/_Float128 conversions to QI/HI/SImode to work with
10466 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10467 conversions to QI/HImode types did a store and then a load to
10468 truncate the value. For conversions to VSX registers, don't split
10469 the insn, instead emit the code directly. Use the code iterator
10470 any_fix to combine signed and unsigned conversions.
10471 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10472 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10473 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10474 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10475 (fix_<mode>di2_hw): Likewise.
10476 (fixuns_<mode>di2_hw): Likewise.
10477 (fix_<mode>si2_hw): Likewise.
10478 (fixuns_<mode>si2_hw): Likewise.
10479 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10480 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10481 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10482 fix<uns>_trunc<SFDF:mode>si2_p8.
10483 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10484 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10485 (fix<uns>_<mode>_mem): Likewise.
10486 (fctiw<u>z_<mode>_mem): Likewise.
10487 (fix<uns>_<mode>_mem): Likewise.
10488 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10489 the register allocator from doing a direct move to the GPRs to do
10490 a store, and instead use the ISA 3.0 store byte/half-word from
10491 vector register instruction. For IEEE 128-bit floating point,
10492 also optimize stores of 32-bit ints.
10493 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10495 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10497 * genextract.c (push_pathstr_operand): New function to support
10499 (walk_rtx): Call push_pathstr_operand.
10500 (print_path): Support [a-zA-Z].
10502 2018-02-07 Richard Biener <rguenther@suse.de>
10504 PR tree-optimization/84037
10505 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10506 (cse_and_gimplify_to_preheader): Declare.
10507 (vect_get_place_in_interleaving_chain): Likewise.
10508 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10510 (_loop_vec_info::~_loop_vec_info): Delete it.
10511 (cse_and_gimplify_to_preheader): New function.
10512 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10513 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10514 (vectorizable_load): Likewise. For grouped stores always base
10515 the IV on the first element.
10516 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10517 condition before gimplifying.
10519 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10521 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10522 *DIV_EXPR and *MOD_EXPR.
10524 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10527 * config/i386/i386.c (ix86_option_override_internal): Mask out
10528 the CF_SET bit when checking -fcf-protection.
10530 2018-02-07 Tom de Vries <tom@codesourcery.com>
10533 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10536 2018-02-07 Richard Biener <rguenther@suse.de>
10538 PR tree-optimization/84204
10539 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10542 PR tree-optimization/84205
10543 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10544 special-case isl_ast_op_zdiv_r.
10546 PR tree-optimization/84223
10547 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10548 Only add conditions from within the region.
10549 (gather_bbs::after_dom_children): Adjust.
10551 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10554 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10555 * config/avr/avr.md: Only post-reload split REG-REG moves if
10556 either register is GENERAL_REG_P.
10558 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10560 PR tree-optimization/84235
10561 * tree-ssa-scopedtables.c
10562 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10563 if the subtraction is performed in floating point type where NaNs are
10564 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10565 build 1. Formatting fix.
10567 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10570 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10571 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10572 and skip it regardless of bb boundaries. Use CALL_P macro,
10573 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10574 unnecessarily, formatting fix.
10576 2018-02-06 Michael Collison <michael.collison@arm.com>
10578 * config/arm/thumb2.md:
10579 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10580 (*thumb_mov_notscc): Ditto.
10582 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10585 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10586 unsigned_fix, not "s".
10588 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10590 * configure.ac (gcc_fn_eh_frame_ro): New function.
10591 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10592 correct .eh_frame permissions.
10593 * configure: Regenerate.
10595 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10597 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10598 irrelevant options.
10600 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10602 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10603 Display warning message for -mno-speculate-indirect-jumps.
10605 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10607 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10609 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10611 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10613 PR tree-optimization/84225
10614 * tree-eh.c (find_trapping_overflow): Only call
10615 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10617 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10620 * config/i386/i386.c: Reimplement the check of possible options
10621 -mibt/-mshstk conbination. Change error messages.
10622 * doc/invoke.texi: Fix a typo: remove extra '='.
10624 2018-02-06 Marek Polacek <polacek@redhat.com>
10626 PR tree-optimization/84228
10627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10629 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10632 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10633 emitted arch directives.
10634 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10635 __ARM_FEATURE_COPROC before changing architectures.
10637 2018-02-06 Richard Biener <rguenther@suse.de>
10639 * config/i386/i386.c (print_reg): Fix typo.
10640 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10642 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10644 * configure: Regenerate.
10646 2018-02-05 Martin Sebor <msebor@redhat.com>
10648 PR tree-optimization/83369
10649 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10652 2018-02-05 Martin Liska <mliska@suse.cz>
10654 * doc/invoke.texi: Cherry-pick upstream r323995.
10656 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10658 * ira.c (ira_init_register_move_cost): Adjust comment.
10660 2018-02-05 Martin Liska <mliska@suse.cz>
10662 PR gcov-profile/84137
10663 * doc/gcov.texi: Fix typo in documentation.
10665 2018-02-05 Martin Liska <mliska@suse.cz>
10667 PR gcov-profile/83879
10668 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10669 using dlopen functionality.
10671 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10673 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10674 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10675 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10676 _mm_maskz_range_round_ss): New intrinsics.
10677 (__builtin_ia32_rangesd128_round)
10678 (__builtin_ia32_rangess128_round): Remove.
10679 (__builtin_ia32_rangesd128_mask_round,
10680 __builtin_ia32_rangess128_mask_round): New builtins.
10681 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10682 __builtin_ia32_rangess128_round): Remove.
10683 (__builtin_ia32_rangesd128_mask_round,
10684 __builtin_ia32_rangess128_mask_round): New builtins.
10685 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10686 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10687 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10688 "<round_saeonly_constraint>")): Changed to ...
10689 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10690 "<round_saeonly_scalar_constraint>")): ... this.
10691 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10692 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10693 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10694 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10695 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10697 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10699 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10701 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10702 Remove all values except native, 8540 and 8548.
10704 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10706 * config/i386/i386.c (ix86_output_function_return): Pass
10707 INVALID_REGNUM, instead of -1, as invalid register number to
10708 indirect_thunk_name and output_indirect_thunk.
10710 2018-02-02 Julia Koval <julia.koval@intel.com>
10712 * config.gcc: Add -march=icelake.
10713 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10714 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10715 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10716 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10717 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10718 (processor_target_table): Add icelake.
10719 (ix86_option_override_internal): Handle new PTAs.
10720 (get_builtin_code_for_version): Handle icelake.
10721 (M_INTEL_COREI7_ICELAKE): New.
10722 (fold_builtin_cpu): Handle icelake.
10723 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10724 * doc/invoke.texi: Add -march=icelake.
10726 2018-02-02 Julia Koval <julia.koval@intel.com>
10728 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10729 to wide_int_bitmask.
10730 * wide-int-bitmask.h: New.
10732 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10735 * config/i386/i386.md: Replace Pmode with word_mode in
10736 builtin_setjmp_setup and builtin_longjmp to support x32.
10738 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10742 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10744 (rs6000_supported_cpu_names): New static variable.
10745 (linux_cpu_translation_table): Likewise.
10746 (elf_platform) <cpu>: Define new static variable and use it.
10747 Translate kernel AT_PLATFORM name to canonical name if needed.
10748 Error if platform name is unknown.
10750 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10753 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10755 2018-02-01 Jeff Law <law@redhat.com>
10758 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10759 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10760 the scratch if RELEASE_VIA_POP is false.
10761 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10762 If we have to save a temporary register, decrement SIZE appropriately.
10763 Pass new arguments to release_scratch_register_on_entry.
10764 (ix86_adjust_stack_and_probe): Likewise.
10765 (ix86_emit_probe_stack_range): Pass new arguments to
10766 release_scratch_register_on_entry.
10768 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10770 PR rtl-optimization/84157
10771 * combine.c (change_zero_ext): Use REG_P predicate in
10772 front of HARD_REGISTER_P predicate.
10774 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10776 * config/avr/avr.c (avr_option_override): Move disabling of
10777 -fdelete-null-pointer-checks to...
10778 * common/config/avr/avr-common.c (avr_option_optimization_table):
10781 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10783 PR tree-optimization/81635
10784 * tree-data-ref.c (split_constant_offset_1): For types that
10785 wrap on overflow, try to use range info to prove that wrapping
10788 2018-02-01 Renlin Li <renlin.li@arm.com>
10791 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10792 TAILCALL_ADDR_REGS.
10793 (aarch64_register_move_cost): Likewise.
10794 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10795 TAILCALL_ADDR_REGS.
10796 (REG_CLASS_NAMES): Likewise.
10797 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10798 TAILCALL_ADDR_REGS. Remove IP registers.
10799 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10801 2018-02-01 Richard Biener <rguenther@suse.de>
10803 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10804 for specifying RPO order and allow NULL for that.
10805 * domwalk.c (dom_walker::dom_walker): Likewise.
10806 (dom_walker::walk): Handle NULL RPO order.
10807 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10809 (rewrite_update_dom_walker): Likewise.
10810 (mark_def_dom_walker): Likewise.
10812 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10814 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10815 (aarch64_maybe_expand_sve_subreg_move): Declare.
10816 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10817 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10819 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10820 that are semantically a reverse operation.
10821 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10822 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10823 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10825 (aarch64_can_change_mode_class): For big-endian, forbid changes
10826 between two SVE modes if they have different element sizes.
10828 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10830 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10831 the TImode handling for big-endian targets.
10833 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10835 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10836 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10838 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10839 Remove BSWAP handing for big-endian targets and use the form of
10840 LD1RQ appropariate for the mode.
10842 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10844 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10845 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10846 duplicated element.
10848 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10851 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10852 check for operands that need to go through aarch64_sve_reload_be.
10854 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10856 PR tree-optimization/81661
10857 PR tree-optimization/84117
10858 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10859 * tree-eh.c: Include gimplify.h.
10860 (find_trapping_overflow, replace_trapping_overflow,
10861 rewrite_to_non_trapping_overflow): New functions.
10862 * tree-vect-loop.c: Include tree-eh.h.
10863 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10864 * tree-data-ref.c: Include tree-eh.h.
10865 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10867 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10869 PR rtl-optimization/84123
10870 * combine.c (change_zero_ext): Check if hard register satisfies
10871 can_change_dest_mode before calling gen_lowpart_SUBREG.
10873 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10876 * ira.c (ira_init_register_move_cost): Remove assert.
10878 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10880 PR rtl-optimization/84071
10881 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10882 * doc/tm.texi: Regenerate.
10884 2018-01-31 Richard Biener <rguenther@suse.de>
10886 PR tree-optimization/84132
10887 * tree-data-ref.c (analyze_miv_subscript): Properly
10888 check whether evolution_function_is_affine_multivariate_p
10889 before calling gcd_of_steps_may_divide_p.
10891 2018-01-31 Julia Koval <julia.koval@intel.com>
10894 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10895 * config/i386/i386.md (rdpid_rex64) New.
10896 (rdpid): Make 32bit only.
10898 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10901 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10902 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10904 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10907 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10909 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10911 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10913 PR rtl-optimization/84071
10914 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10915 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10917 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10919 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10920 (arc_attribute_table): Add 'aux' attribute.
10921 (arc_in_small_data_p): Consider aux like variables.
10922 (arc_is_aux_reg_p): New function.
10923 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
10924 (arc_get_aux_arg): New function.
10925 (prepare_move_operands): Handle aux-register access.
10926 (arc_handle_aux_attribute): New function.
10927 * doc/extend.texi (ARC Variable attributes): Add subsection.
10929 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10931 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
10932 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
10933 (arc_attribute_table): Add 'uncached' attribute.
10934 (arc_print_operand): Print '.di' flag for uncached memory
10936 (arc_in_small_data_p): Do not consider for small data the uncached
10938 (arc_is_uncached_mem_p): New function.
10939 * config/arc/predicates.md (compact_store_memory_operand): Check
10940 for uncached memory accesses.
10941 (nonvol_nonimm_operand): Likewise.
10942 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
10944 2018-01-31 Jakub Jelinek <jakub@redhat.com>
10947 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
10948 falign-loops=): Add Optimization flag.
10950 2018-01-30 Jeff Law <law@redhat.com>
10953 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
10954 INT_REGISTERS_SAVED. Check it prior to calling
10955 get_scratch_register_on_entry.
10956 (ix86_adjust_stack_and_probe): Similarly.
10957 (ix86_emit_probe_stack_range): Similarly.
10958 (ix86_expand_prologue): Corresponding changes.
10960 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10963 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
10964 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
10966 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
10969 * lra-constraints.c (curr_insn_transform): Process AND in the
10972 2018-01-30 Jakub Jelinek <jakub@redhat.com>
10974 PR rtl-optimization/83986
10975 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
10976 dependence against last_pending_memory_flush in addition to
10977 pending_jump_insns.
10979 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
10981 PR tree-optimization/81611
10982 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
10985 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10988 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
10991 2018-01-30 Richard Biener <rguenther@suse.de>
10992 Jakub Jelinek <jakub@redhat.com>
10994 PR tree-optimization/84111
10995 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
10996 inner loops added during recursion, as they don't have up-to-date
10999 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11002 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11003 (can_inline_edge_by_limits_p): ... here.
11004 (can_early_inline_edge_p, check_callers,
11005 update_caller_keys, update_callee_keys, recursive_inlining,
11006 add_new_edges_to_heap, speculation_useful_p,
11007 inline_small_functions,
11008 inline_small_functions, flatten_function,
11009 inline_to_all_callers_1): Update.
11011 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11013 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11016 2018-01-30 Richard Biener <rguenther@suse.de>
11018 PR tree-optimization/83008
11019 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11020 invariant and constant vector uses in stmts when they need
11021 more than one stmt.
11023 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11026 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11027 * configure: Regenerate.
11029 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11031 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11033 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11034 Use gen_rtx_REG rather than gen_lowpart.
11036 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11038 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11039 rather than 0 when creating partial subregs.
11041 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11043 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11046 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11049 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11050 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11051 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11052 flags. This restores the settings used before the 2017-07-24.
11053 Turning off pre increment/decrement/modify allows IVOPTS to
11054 optimize DF/SF loops where the index is an int.
11056 2018-01-29 Richard Biener <rguenther@suse.de>
11057 Kelvin Nilsen <kelvin@gcc.gnu.org>
11060 * tree-vect-stmts.c (vectorizable_call): Don't call
11061 targetm.vectorize_builtin_md_vectorized_function if callee is
11064 2018-01-22 Carl Love <cel@us.ibm.com>
11066 * doc/extend.tex: Fix typo in second arg in
11067 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11069 2018-01-29 Richard Biener <rguenther@suse.de>
11071 PR tree-optimization/84086
11072 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11073 (flush_ssaname_freelist): When SSA names were released reset
11074 the SCEV hash table.
11076 2018-01-29 Richard Biener <rguenther@suse.de>
11078 PR tree-optimization/84057
11079 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11080 removed paths when removing edges.
11082 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11084 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11085 -mfunction-return=@var{choice}.
11087 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11089 PR diagnostic/84034
11090 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11091 Handle CR like TAB.
11092 (layout::print_source_line): Likewise.
11093 (test_get_line_width_without_trailing_whitespace): Add test cases.
11095 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11097 PR middle-end/84040
11098 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11101 2018-01-26 Jim Wilson <jimw@sifive.com>
11103 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11105 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11108 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11110 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11111 and CMP + SUB-immediate -> SUBS.
11113 2018-01-26 Martin Sebor <msebor@redhat.com>
11115 PR tree-optimization/83896
11116 * tree-ssa-strlen.c (get_string_len): Rename...
11117 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11118 Avoid assuming length is constant.
11119 (handle_char_store): Use HOST_WIDE_INT for string length.
11121 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11124 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11125 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11127 2018-01-26 Richard Biener <rguenther@suse.de>
11129 PR rtl-optimization/84003
11130 * dse.c (record_store): Only record redundant stores when
11131 the earlier store aliases at least all accesses the later one does.
11133 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11135 PR rtl-optimization/83985
11136 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11137 REG_CFA_RESTORE insns.
11138 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11139 REG_CFA_RESTORE insns here.
11142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11143 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11145 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11147 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11148 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11149 (arc_init): Likewise.
11150 (arc_override_options): Likewise.
11151 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11153 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11155 * config/arc/arc.h (TARGET_DBNZ): Define.
11156 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11157 properly set the tune attribute.
11158 (dbnz): Use TARGET_DBNZ guard.
11159 * config/arc/arc.opt (mtune): Add core3 option.
11161 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11163 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11164 recognize new pic like addresses.
11165 (arc_delegitimize_address): Clean up.
11167 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11169 * config/arc/arc-arches.def: Option mrf16 valid for all
11171 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11172 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11173 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11174 * config/arc/arc-tables.opt: Regenerate.
11175 * config/arc/arc.c (arc_conditional_register_usage): Handle
11176 reduced register file case.
11177 (arc_file_start): Set must have build attributes.
11178 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11179 mrf16 option value.
11180 * config/arc/arc.opt (mrf16): Add new option.
11181 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11182 * config/arc/genmultilib.awk: Handle new mrf16 option.
11183 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11184 * config/arc/t-multilib: Regenerate.
11185 * doc/invoke.texi (ARC Options): Document mrf16 option.
11187 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11189 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11190 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11191 (arc_attribute_table): Add 'secure_call' attribute.
11192 (arc_print_operand): Print secure call operand.
11193 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11195 (arc_is_secure_call_p): New function. * config/arc/arc.md
11196 (call_i): Add support for sjli instruction.
11197 (call_value_i): Likewise.
11198 * config/arc/constraints.md (Csc): New constraint.
11200 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11201 John Eric Martin <John.Martin@emmicro-us.com>
11203 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11204 * config/arc/arc.c (_arc_jli_section): New struct.
11205 (arc_jli_section): New type.
11206 (rc_jli_sections): New static variable.
11207 (arc_handle_jli_attribute): New function.
11208 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11209 (arc_file_end): New function.
11210 (TARGET_ASM_FILE_END): Define.
11211 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11212 (arc_add_jli_section): New function.
11213 (jli_call_scan): Likewise.
11214 (arc_reorg): Call jli_call_scan.
11215 (arc_output_addsi): Remove 'S' from printing asm operand.
11216 (arc_is_jli_call_p): New function.
11217 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11219 (movhi_insn): Likewise.
11220 (movsi_insn): Likewise.
11221 (movsi_set_cc_insn): Likewise.
11222 (loadqi_update): Likewise.
11223 (load_zeroextendqisi_update): Likewise.
11224 (load_signextendqisi_update): Likewise.
11225 (loadhi_update): Likewise.
11226 (load_zeroextendhisi_update): Likewise.
11227 (load_signextendhisi_update): Likewise.
11228 (loadsi_update): Likewise.
11229 (loadsf_update): Likewise.
11230 (movsicc_insn): Likewise.
11231 (bset_insn): Likewise.
11232 (bxor_insn): Likewise.
11233 (bclr_insn): Likewise.
11234 (bmsk_insn): Likewise.
11235 (bicsi3_insn): Likewise.
11236 (cmpsi_cc_c_insn): Likewise.
11237 (movsi_ne): Likewise.
11238 (movsi_cond_exec): Likewise.
11239 (clrsbsi2): Likewise.
11240 (norm_f): Likewise.
11250 (call_i): Remove 'S' asm letter, add jli instruction.
11251 (call_value_i): Likewise.
11252 * config/arc/arc.op (mjli-always): New option.
11253 * config/arc/constraints.md (Cji): New constraint.
11254 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11256 (subsf3_fpx): Likewise.
11257 (mulsf3_fpx): Likewise.
11258 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11260 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11261 function attrbutes.
11262 * doc/invoke.texi (ARC): Document mjli-always option.
11264 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11266 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11267 avoid addition with 0 and use incw and decw where possible.
11269 2018-01-26 Richard Biener <rguenther@suse.de>
11271 PR tree-optimization/81082
11272 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11273 association if it requires casting to unsigned.
11274 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11275 from fold_plusminus_mult_expr to catch important cases late when
11276 range info is available.
11278 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11280 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11281 * configure.ac (hidden_linkonce): New test.
11282 * configure: Regenerate.
11283 * config.in: Regenerate.
11285 2018-01-26 Julia Koval <julia.koval@intel.com>
11287 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11288 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11289 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11290 _mm_mask_bitshuffle_epi64_mask): Fix type.
11291 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11292 USI_FTYPE_V4DI_V4DI_USI): Remove.
11293 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11294 __builtin_ia32_vpshufbitqmb256_mask,
11295 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11296 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11297 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11299 2018-01-26 Alan Modra <amodra@gmail.com>
11302 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11303 UNSPEC_VBPERMQ. Sort other unspecs.
11305 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11307 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11309 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11311 PR middle-end/83055
11312 * predict.c (drop_profile): Do not push/pop cfun; update also
11314 (handle_missing_profiles): Fix logic looking for zero profiles.
11316 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11318 PR middle-end/83977
11319 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11320 on functions with #pragma omp declare simd or functions with simd
11322 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11323 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11324 Remove trailing \n from warning_at calls.
11326 2018-01-25 Tom de Vries <tom@codesourcery.com>
11329 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11330 for neutered workers.
11332 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11335 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11336 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11338 2018-01-24 Jeff Law <law@redhat.com>
11341 * i386.c (get_probe_interval): Move to earlier point.
11342 (ix86_compute_frame_layout): If -fstack-clash-protection and
11343 the frame is larger than the probe interval, then use pushes
11344 to save registers rather than reg->mem moves.
11345 (ix86_expand_prologue): Remove conditional for int_registers_saved
11348 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11351 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11352 min/max for never referenced object.
11354 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11356 PR middle-end/83977
11357 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11359 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11360 attributes from DECL_ATTRIBUTES (decl) without affecting
11361 DECL_ATTRIBUTES (current_function_decl).
11362 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11363 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11365 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11367 PR tree-optimization/83979
11368 * fold-const.c (fold_comparison): Use constant_boolean_node
11369 instead of boolean_{true,false}_node.
11371 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11373 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11376 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11378 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11379 Simplify the clause that sets the length attribute.
11380 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11381 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11382 clause that sets the length attribute.
11383 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11385 2018-01-24 Tom de Vries <tom@codesourcery.com>
11388 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11389 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11390 Add strict parameter.
11391 (prevent_branch_around_nothing): Insert dummy insn between branch to
11392 label and label with no ptx insn inbetween.
11393 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11395 2018-01-24 Tom de Vries <tom@codesourcery.com>
11398 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11399 for neutered threads in warp.
11400 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11402 2018-01-24 Richard Biener <rguenther@suse.de>
11404 PR tree-optimization/83176
11405 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11408 2018-01-24 Richard Biener <rguenther@suse.de>
11410 PR tree-optimization/82819
11411 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11412 code generating pluses that are no-ops in the target precision.
11414 2018-01-24 Richard Biener <rguenther@suse.de>
11416 PR middle-end/84000
11417 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11419 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11421 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11422 to merge probabilities.
11423 * predict.c (probably_never_executed): Also mark as cold functions
11424 with global 0 profile and guessed local profile.
11425 * profile-count.c (profile_probability::combine_with_count): New
11427 * profile-count.h (profile_probability::operator*,
11428 profile_probability::operator*=, profile_probability::operator/,
11429 profile_probability::operator/=): Reduce precision to adjusted
11430 and set value to guessed on contradictory divisions.
11431 (profile_probability::combine_with_freq): Remove.
11432 (profile_probability::combine_wiht_count): Declare.
11433 (profile_count::force_nonzero):: Set to adjusted.
11434 (profile_count::probability_in):: Set quality to adjusted.
11435 * tree-ssa-tail-merge.c (replace_block_by): Use
11436 combine_with_count.
11438 2018-01-23 Andrew Waterman <andrew@sifive.com>
11439 Jim Wilson <jimw@sifive.com>
11441 * config/riscv/riscv.c (riscv_stack_boundary): New.
11442 (riscv_option_override): Set riscv_stack_boundary. Handle
11443 riscv_preferred_stack_boundary_arg.
11444 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11445 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11446 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11447 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11448 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11449 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11451 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11454 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11455 of struct ix86_frame.
11456 (ix86_expand_epilogue): Likewise. Add a local variable for
11457 the reg_save_offset field in struct ix86_frame.
11459 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11461 PR tree-optimization/82604
11462 * tree-loop-distribution.c (enum partition_kind): New enum item
11463 PKIND_PARTIAL_MEMSET.
11464 (partition_builtin_p): Support above new enum item.
11465 (generate_code_for_partition): Ditto.
11466 (compute_access_range): Differentiate cases that equality can be
11467 proven at all loops, the innermost loops or no loops.
11468 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11469 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11470 (finalize_partitions, distribute_loop): Don't fuse partition of
11471 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11472 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11473 parloop is enabled.
11475 2018-01-23 Martin Liska <mliska@suse.cz>
11477 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11478 order to ignore the predictor.
11479 (PRED_POLYMORPHIC_CALL): Likewise.
11480 (PRED_RECURSIVE_CALL): Likewise.
11482 2018-01-23 Martin Liska <mliska@suse.cz>
11484 * tree-profile.c (tree_profiling): Print function header to
11485 aware reader which function we are working on.
11486 * value-prof.c (gimple_find_values_to_profile): Do not print
11487 not interesting value histograms.
11489 2018-01-23 Martin Liska <mliska@suse.cz>
11491 * profile-count.h (enum profile_quality): Add
11492 profile_uninitialized as the first value. Do not number values
11493 as they are zero based.
11494 (profile_count::verify): Update sanity check.
11495 (profile_probability::verify): Likewise.
11497 2018-01-23 Nathan Sidwell <nathan@acm.org>
11499 * doc/invoke.texi (ffor-scope): Deprecate.
11501 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11503 PR tree-optimization/83510
11504 * domwalk.c (set_all_edges_as_executable): New function.
11505 (dom_walker::dom_walker): Convert bool param
11506 "skip_unreachable_blocks" to enum reachability. Move setup of
11507 edge flags to set_all_edges_as_executable and only do it when
11508 reachability is REACHABLE_BLOCKS.
11509 * domwalk.h (enum dom_walker::reachability): New enum.
11510 (dom_walker::dom_walker): Convert bool param
11511 "skip_unreachable_blocks" to enum reachability.
11512 (set_all_edges_as_executable): New decl.
11513 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11514 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11516 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11517 but converting true to REACHABLE_BLOCKS.
11518 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11520 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11521 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11522 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11524 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11525 if check_all_array_refs will be called.
11527 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11529 * tree.c (selftest::test_location_wrappers): Add more test
11532 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11534 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11535 (selftest::test_bit_in_range): Likewise.
11537 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11540 * doc/sourcebuild.texi (vect_float): Say that the selector
11541 only describes the situation when -funsafe-math-optimizations is on.
11542 (vect_float_strict): Document.
11544 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11546 PR tree-optimization/83965
11547 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11548 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11549 instead of checking only for a reduction.
11550 (vect_recog_widen_sum_pattern): Likewise.
11552 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11554 * predict.c (probably_never_executed): Only use precise profile info.
11555 (compute_function_frequency): Skip after inlining hack since we now
11556 have quality checking.
11558 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11560 * profile-count.h (profile_probability::very_unlikely,
11561 profile_probability::unlikely, profile_probability::even): Set
11562 precision to guessed.
11564 2018-01-23 Richard Biener <rguenther@suse.de>
11566 PR tree-optimization/83963
11567 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11568 Properly terminate dominator walk when crossing the exit edge not
11569 when visiting its source block.
11571 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11574 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11575 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11577 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11579 PR tree-optimization/83957
11580 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11581 semicolon after for body surrounded by braces.
11583 PR tree-optimization/83081
11584 * profile-count.h (profile_probability::split): New method.
11585 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11586 Use profile_probability::split.
11587 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11588 when splitting a single conditional jump into 2.
11590 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11592 PR tree-optimization/69452
11593 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11596 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11598 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11599 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11600 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11602 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11604 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11606 * config/rl78/rl78.md (movdi): New define_expand.
11607 * config/rl78/rl78.c (rl78_split_movdi): New function.
11609 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11612 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11614 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11615 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11616 128-bit to produce an UNSPEC move to get the double word with the
11617 signbit and then a shift directly to do signbit.
11618 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11619 implementation with a new version that just does either a direct
11620 move or a regular move. Move memory interface to separate insns.
11621 Move insns so they are next to the expander.
11622 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11623 with signbit move. Split big and little endian case.
11624 (signbit<mode>2_dm_mem_le): Likewise.
11625 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11626 (signbit<mode>2_dm2): Likewise.
11628 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11630 * config/rl78/rl78.md (anddi3): New define_expand.
11632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11634 * config/rl78/rl78.md (umindi3): New define_expand.
11636 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11638 * config/rl78/rl78.md (smindi3): New define_expand.
11640 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11642 * config/rl78/rl78.md (smaxdi3): New define_expand.
11644 2018-01-22 Carl Love <cel@us.ibm.com>
11646 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11647 LVX_V1TI): Add macro expansion.
11648 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11649 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11650 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11651 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11652 Change check to determine if the instruction is a byte reversing
11653 entry. Fix typo in comment.
11654 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11655 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11656 Add def_builtin calls for new builtins.
11657 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11658 Add define_insn expansion.
11660 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11662 * config/rl78/rl78.md (umaxdi3): New define_expand.
11664 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11666 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11667 for non-QImode registers.
11669 2018-01-22 Richard Biener <rguenther@suse.de>
11671 PR tree-optimization/83963
11672 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11673 including the loop exit block.
11674 (scop_detection::merge_sese): Likewise.
11675 (scop_detection::add_scop): Do it here instead.
11677 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11679 * doc/sourcebuild.texi (arm_softfloat): Document.
11681 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11684 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11685 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11686 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11688 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11689 David Edelsohn <dje.gcc@gmail.com>
11692 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11693 Change "crset eq" to "crset 2".
11694 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11695 (*call_indirect_aix<mode>_nospec): Likewise.
11696 (*call_value_indirect_aix<mode>_nospec): Likewise.
11697 (*call_indirect_elfv2<mode>_nospec): Likewise.
11698 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11699 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11700 change assembly output from . to $.
11701 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11702 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11703 (*tablejump<mode>_internal1_nospec): Likewise.
11705 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11708 * config/sh/sh_optimize_sett_clrt.cc:
11709 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11711 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11713 PR tree-optimization/83940
11714 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11715 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11716 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11717 use it to pass back the definition type.
11718 (vect_check_store_rhs): Likewise rhs_dt_out.
11719 (vect_build_gather_load_calls): Add a mask_dt argument and use
11720 it instead of a call to vect_is_simple_use.
11721 (vectorizable_store): Update calls to vect_check_load_store_mask
11722 and vect_check_store_rhs. Use the dt returned by the latter instead
11723 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11724 instead of calls to vect_is_simple_use. Pass the scalar rather
11725 than the vector operand to vect_is_simple_use when handling
11726 second and subsequent copies of an rhs value.
11727 (vectorizable_load): Update calls to vect_check_load_store_mask
11728 and vect_build_gather_load_calls. Use the cached mask_dt and
11729 gs_info.offset_dt instead of calls to vect_is_simple_use.
11731 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11733 PR middle-end/83945
11734 * tree-emutls.c: Include gimplify.h.
11735 (lower_emutls_2): New function.
11736 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11737 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11738 it before further processing.
11741 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11742 UINTVAL (trueop1) instead of INTVAL (op1).
11744 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11748 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11749 INCOMING_FRAME_SP_OFFSET if not defined.
11750 (scan_trace): Add ENTRY argument. If true and
11751 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11752 emit a note to adjust the CFA offset.
11753 (create_cfi_notes): Adjust scan_trace callers.
11754 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11755 INCOMING_FRAME_SP_OFFSET in the CIE.
11756 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11757 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11759 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11760 * doc/tm.texi: Regenerated.
11762 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11764 PR rtl-optimization/83147
11765 * lra-constraints.c (remove_inheritance_pseudos): Use
11766 lra_substitute_pseudo_within_insn.
11768 2018-01-19 Tom de Vries <tom@codesourcery.com>
11769 Cesar Philippidis <cesar@codesourcery.com>
11772 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11774 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11777 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11778 spaces for function labels.
11780 2018-01-19 Martin Liska <mliska@suse.cz>
11782 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11783 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11784 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11785 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11786 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11787 (PRED_CONST_RETURN): Change from 69 to 65.
11788 (PRED_NULL_RETURN): Change from 91 to 71.
11789 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11790 (PRED_LOOP_GUARD): Change from 66 to 73.
11792 2018-01-19 Martin Liska <mliska@suse.cz>
11794 * predict.c (predict_insn_def): Add new assert.
11795 (struct branch_predictor): Change type to signed integer.
11796 (test_prediction_value_range): Amend test to cover
11797 PROB_UNINITIALIZED.
11798 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11799 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11800 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11801 (PRED_LOOP_IV_COMPARE): Likewise.
11802 * predict.h (PROB_UNINITIALIZED): Define new constant.
11804 2018-01-19 Martin Liska <mliska@suse.cz>
11806 * predict.c (dump_prediction): Add new format for
11807 analyze_brprob.py script which is enabled with -details
11809 * profile-count.h (precise_p): New function.
11811 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11813 PR tree-optimization/83922
11814 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11815 there are no statements that need masking.
11816 (vect_active_double_reduction_p): New function.
11817 (vect_analyze_loop_operations): Use it when handling phis that
11818 are not in the loop header.
11820 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11822 PR tree-optimization/83914
11823 * tree-vect-loop.c (vectorizable_induction): Don't convert
11824 init_expr or apply the peeling adjustment for inductions
11825 that are nested within the vectorized loop.
11827 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11829 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11832 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11836 * function.h (gimplify_parameters): Add gimple_seq * argument.
11837 * function.c: Include gimple.h and options.h.
11838 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11839 for the added local temporaries if needed.
11840 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11841 if there are any parameter cleanups, wrap whole body into a
11842 try/finally with the cleanups.
11844 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11847 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11848 Use GET_MODE_CLASS for scalar floating point.
11850 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11854 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11855 Fix call of call_cgraph_insertion_hooks.
11857 2018-01-18 Martin Sebor <msebor@redhat.com>
11859 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11861 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11864 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11867 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11870 * common.opt: (-ffile-prefix-map): New option.
11871 * opts.c (common_handle_option): Defer it.
11872 * opts-global.c (handle_common_deferred_options): Handle it.
11873 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11874 * file-prefix-map.h: New file.
11875 (remap_debug_filename, add_debug_prefix_map): ...here.
11876 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11877 * final.c (debug_prefix_map, add_debug_prefix_map
11878 remap_debug_filename): Move to...
11879 * file-prefix-map.c: New file.
11880 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11881 generalize, get rid of alloca(), use strrchr() instead of strchr().
11882 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11883 Implement in terms of add_prefix_map().
11884 (remap_macro_filename, remap_debug_filename): Implement in term of
11886 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11887 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11888 * dbxout.c: Include file-prefix-map.h.
11889 * varasm.c: Likewise.
11890 * vmsdbgout.c: Likewise.
11891 * xcoffout.c: Likewise.
11892 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11893 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11894 * doc/invoke.texi (-ffile-prefix-map): Document.
11895 (-fdebug-prefix-map): Update description.
11897 2018-01-18 Martin Liska <mliska@suse.cz>
11899 * config/i386/i386.c (indirect_thunk_name): Document that also
11901 (output_indirect_thunk): Document why both instructions
11902 (pause and lfence) are generated.
11904 2018-01-18 Richard Biener <rguenther@suse.de>
11906 PR tree-optimization/83887
11907 * graphite-scop-detection.c
11908 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11909 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11910 (scop_detection::merge_sese): Re-implement with a flood-fill
11911 algorithm that properly finds a SESE region if it exists.
11913 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11916 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11917 pointer_diff optimizations use view_convert instead of convert.
11919 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11921 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11922 Generate different code for -mno-speculate-indirect-jumps.
11923 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11924 (*call_indirect_aix<mode>): Disable for
11925 -mno-speculate-indirect-jumps.
11926 (*call_indirect_aix<mode>_nospec): New define_insn.
11927 (*call_value_indirect_aix<mode>): Disable for
11928 -mno-speculate-indirect-jumps.
11929 (*call_value_indirect_aix<mode>_nospec): New define_insn.
11930 (*sibcall_nonlocal_sysv<mode>): Generate different code for
11931 -mno-speculate-indirect-jumps.
11932 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11934 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11936 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
11937 long double type, set the flags for noting the default long double
11938 type, even if we don't pass or return a long double type.
11940 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
11943 * ipa-inline.c (flatten_function): Do not overwrite final inlining
11946 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
11948 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
11949 support for merge[hl].
11950 (fold_mergehl_helper): New helper function.
11951 (tree-vector-builder.h): New #include for tree_vector_builder usage.
11952 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
11953 (altivec_vmrglw_direct): Add xxmrglw insn.
11955 2018-01-17 Andrew Waterman <andrew@sifive.com>
11957 * config/riscv/riscv.c (riscv_conditional_register_usage): If
11958 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
11960 2018-01-17 David Malcolm <dmalcolm@redhat.com>
11963 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
11964 call the lto_location_cache before reading the
11965 DECL_SOURCE_LOCATION of the types.
11967 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
11968 Richard Sandiford <richard.sandiford@linaro.org>
11970 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
11971 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
11972 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
11973 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
11974 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
11976 * config/aarch64/constraints.md (aarch64_movti_operand):
11978 * config/aarch64/predicates.md (Uti): Add new constraint.
11980 2018-01-17 Carl Love <cel@us.ibm.com>
11982 * config/rs6000/vsx.md (define_expand xl_len_r,
11983 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
11984 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
11986 (define_expand, define_insn): Move the shift left from the
11987 define_insn to the define_expand for lxvl and stxvl instructions.
11988 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
11989 and XL_LEN_R definitions to PURE.
11991 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
11993 * config/i386/i386.c (indirect_thunk_name): Declare regno
11994 as unsigned int. Compare regno with INVALID_REGNUM.
11995 (output_indirect_thunk): Ditto.
11996 (output_indirect_thunk_function): Ditto.
11997 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
11998 in the call to output_indirect_thunk_function.
12000 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12002 PR middle-end/83884
12003 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12004 rather than the size of inner_type to determine the stack slot size
12005 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12007 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12010 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12013 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12015 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12016 endian Linux systems to optionally enable multilibs for selecting
12017 the long double type if the user configured an explicit type.
12018 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12019 have no long double multilibs if not defined.
12020 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12021 warn if the user used -mabi={ieee,ibm}longdouble and we built
12022 multilibs for long double.
12023 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12024 appropriate multilib option.
12025 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12027 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12028 for building long double multilibs.
12029 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12031 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12033 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12036 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12038 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12041 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12044 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12047 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12049 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12050 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12052 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12054 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12055 different rtl trees depending on TARGET_64BIT.
12056 (rs6000_gen_lvx): Likewise.
12058 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12060 * config/visium/visium.md (nop): Tweak comment.
12061 (hazard_nop): Likewise.
12063 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12065 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12066 -mspeculate-indirect-jumps.
12067 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12068 for -mno-speculate-indirect-jumps.
12069 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12070 (*call_value_indirect_elfv2<mode>): Disable for
12071 -mno-speculate-indirect-jumps.
12072 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12073 (indirect_jump): Emit different RTL for
12074 -mno-speculate-indirect-jumps.
12075 (*indirect_jump<mode>): Disable for
12076 -mno-speculate-indirect-jumps.
12077 (*indirect_jump<mode>_nospec): New define_insn.
12078 (tablejump): Emit different RTL for
12079 -mno-speculate-indirect-jumps.
12080 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12081 (tablejumpsi_nospec): New define_expand.
12082 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12083 (tablejumpdi_nospec): New define_expand.
12084 (*tablejump<mode>_internal1): Disable for
12085 -mno-speculate-indirect-jumps.
12086 (*tablejump<mode>_internal1_nospec): New define_insn.
12087 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12090 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12092 * caller-save.c (insert_save): Drop unnecessary parameter. All
12095 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12096 Richard Biener <rguenth@suse.de>
12099 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12100 return early, inline manually is_gimple_sizepos. Make sure if we
12101 call gimplify_expr we don't end up with a gimple constant.
12102 * tree.c (variably_modified_type_p): Don't return true for
12103 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12104 * gimplify.h (is_gimple_sizepos): Remove.
12106 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12108 PR tree-optimization/83857
12109 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12110 vectorizable_live_operation for pure SLP statements.
12111 (vectorizable_live_operation): Handle PHIs.
12113 2018-01-16 Richard Biener <rguenther@suse.de>
12115 PR tree-optimization/83867
12116 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12117 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12119 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12122 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12123 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12124 If off is not INTEGER_CST, issue a may not be aligned warning
12125 rather than isn't aligned. Use isn%'t rather than isn't.
12126 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12128 <case MULT_EXPR>: Improve the case when bottom and one of the
12129 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12130 operand, in that case check if the other operand is multiple of
12131 bottom divided by the INTEGER_CST operand.
12133 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12136 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12137 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12138 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12139 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12140 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12141 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12142 (pa_function_arg_size): New function.
12144 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12146 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12147 in a separate statement.
12149 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12151 PR tree-optimization/83847
12152 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12153 group gathers and scatters.
12155 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12157 PR rtl-optimization/86620
12158 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12160 PR rtl-optimization/83213
12161 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12162 to last if both are JUMP_INSNs.
12164 PR tree-optimization/83843
12165 * gimple-ssa-store-merging.c
12166 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12167 store_immediate_info for bswap/nop orig_stores.
12169 2018-01-15 Andrew Waterman <andrew@sifive.com>
12171 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12173 <UDIV>: Increase cost if !TARGET_DIV.
12175 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12177 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12178 (define_attr "cr_logical_3op"): New.
12179 (cceq_ior_compare): Adjust.
12180 (cceq_ior_compare_complement): Adjust.
12181 (*cceq_rev_compare): Adjust.
12182 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12183 (is_cracked_insn): Adjust.
12184 (insn_must_be_first_in_group): Adjust.
12185 * config/rs6000/40x.md: Adjust.
12186 * config/rs6000/440.md: Adjust.
12187 * config/rs6000/476.md: Adjust.
12188 * config/rs6000/601.md: Adjust.
12189 * config/rs6000/603.md: Adjust.
12190 * config/rs6000/6xx.md: Adjust.
12191 * config/rs6000/7450.md: Adjust.
12192 * config/rs6000/7xx.md: Adjust.
12193 * config/rs6000/8540.md: Adjust.
12194 * config/rs6000/cell.md: Adjust.
12195 * config/rs6000/e300c2c3.md: Adjust.
12196 * config/rs6000/e500mc.md: Adjust.
12197 * config/rs6000/e500mc64.md: Adjust.
12198 * config/rs6000/e5500.md: Adjust.
12199 * config/rs6000/e6500.md: Adjust.
12200 * config/rs6000/mpc.md: Adjust.
12201 * config/rs6000/power4.md: Adjust.
12202 * config/rs6000/power5.md: Adjust.
12203 * config/rs6000/power6.md: Adjust.
12204 * config/rs6000/power7.md: Adjust.
12205 * config/rs6000/power8.md: Adjust.
12206 * config/rs6000/power9.md: Adjust.
12207 * config/rs6000/rs64.md: Adjust.
12208 * config/rs6000/titan.md: Adjust.
12210 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12212 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12213 ix86_indirect_branch_register logic.
12215 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12217 * config/i386/constraints.md (Bs): Update
12218 ix86_indirect_branch_register check. Don't check
12219 ix86_indirect_branch_register with GOT_memory_operand.
12221 * config/i386/predicates.md (GOT_memory_operand): Don't check
12222 ix86_indirect_branch_register here.
12223 (GOT32_symbol_operand): Likewise.
12225 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12227 * config/i386/predicates.md (constant_call_address_operand):
12228 Rewrite ix86_indirect_branch_register logic.
12229 (sibcall_insn_operand): Likewise.
12231 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12233 * config/i386/constraints.md (Bs): Replace
12234 ix86_indirect_branch_thunk_register with
12235 ix86_indirect_branch_register.
12237 * config/i386/i386.md (indirect_jump): Likewise.
12238 (tablejump): Likewise.
12239 (*sibcall_memory): Likewise.
12240 (*sibcall_value_memory): Likewise.
12241 Peepholes of indirect call and jump via memory: Likewise.
12242 * config/i386/i386.opt: Likewise.
12243 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12244 (GOT_memory_operand): Likewise.
12245 (call_insn_operand): Likewise.
12246 (sibcall_insn_operand): Likewise.
12247 (GOT32_symbol_operand): Likewise.
12249 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12251 PR middle-end/83837
12252 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12253 type rather than type addr's type points to.
12254 (expand_omp_atomic_mutex): Likewise.
12255 (expand_omp_atomic): Likewise.
12257 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12260 * config/i386/i386.c (output_indirect_thunk_function): Use
12261 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12262 for __x86_return_thunk.
12264 2018-01-15 Richard Biener <rguenther@suse.de>
12266 PR middle-end/83850
12267 * expmed.c (extract_bit_field_1): Fix typo.
12269 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12272 * config/arm/iterators.md (VF): New mode iterator.
12273 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12274 Remove integer-related logic from pattern.
12275 (neon_vabd<mode>_3): Likewise.
12277 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12279 PR middle-end/82694
12280 * common.opt (fstrict-overflow): No longer an alias.
12281 (fwrapv-pointer): New option.
12282 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12283 also for pointer types based on flag_wrapv_pointer.
12284 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12285 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12286 opts->x_flag_wrapv got set.
12287 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12288 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12289 POINTER_TYPE_OVERFLOW_UNDEFINED.
12290 * match.pd: Likewise in address comparison pattern.
12291 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12293 2018-01-15 Richard Biener <rguenther@suse.de>
12296 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12297 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12298 Reset type names to their identifier if their TYPE_DECL doesn't
12299 have linkage (and thus is used for ODR and devirt).
12300 (save_debug_info_for_decl): Remove.
12301 (save_debug_info_for_type): Likewise.
12302 (add_tree_to_fld_list): Adjust.
12303 * tree-pretty-print.c (dump_generic_node): Make dumping of
12304 type names more robust.
12306 2018-01-15 Richard Biener <rguenther@suse.de>
12308 * BASE-VER: Bump to 8.0.1.
12310 2018-01-14 Martin Sebor <msebor@redhat.com>
12313 * builtins.c (check_access): Avoid warning when the no-warning bit
12316 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12318 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12319 * ira-color (allocno_hard_regs_compare): Likewise.
12321 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12324 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12325 Use .pushsection/.popsection.
12327 2018-01-14 Martin Sebor <msebor@redhat.com>
12330 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12332 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12334 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12335 entry from extra_headers.
12336 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12337 extra_headers, make the list bitwise identical to the i?86-*-* one.
12339 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12341 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12342 -mcmodel=large with -mindirect-branch=thunk,
12343 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12344 -mfunction-return=thunk-extern.
12345 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12346 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12347 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12349 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12351 * config/i386/i386.c (print_reg): Print the name of the full
12352 integer register without '%'.
12353 (ix86_print_operand): Handle 'V'.
12354 * doc/extend.texi: Document 'V' modifier.
12356 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12358 * config/i386/constraints.md (Bs): Disallow memory operand for
12359 -mindirect-branch-register.
12361 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12362 (GOT_memory_operand): Likewise.
12363 (call_insn_operand): Likewise.
12364 (sibcall_insn_operand): Likewise.
12365 (GOT32_symbol_operand): Likewise.
12366 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12367 for -mindirect-branch-register.
12368 (tablejump): Likewise.
12369 (*sibcall_memory): Likewise.
12370 (*sibcall_value_memory): Likewise.
12371 Disallow peepholes of indirect call and jump via memory for
12372 -mindirect-branch-register.
12373 (*call_pop): Replace m with Bw.
12374 (*call_value_pop): Likewise.
12375 (*sibcall_pop_memory): Replace m with Bs.
12376 * config/i386/i386.opt (mindirect-branch-register): New option.
12377 * doc/invoke.texi: Document -mindirect-branch-register option.
12379 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12381 * config/i386/i386-protos.h (ix86_output_function_return): New.
12382 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12383 set function_return_type.
12384 (indirect_thunk_name): Add ret_p to indicate thunk for function
12386 (output_indirect_thunk_function): Pass false to
12387 indirect_thunk_name.
12388 (ix86_output_indirect_branch_via_reg): Likewise.
12389 (ix86_output_indirect_branch_via_push): Likewise.
12390 (output_indirect_thunk_function): Create alias for function
12391 return thunk if regno < 0.
12392 (ix86_output_function_return): New function.
12393 (ix86_handle_fndecl_attribute): Handle function_return.
12394 (ix86_attribute_table): Add function_return.
12395 * config/i386/i386.h (machine_function): Add
12396 function_return_type.
12397 * config/i386/i386.md (simple_return_internal): Use
12398 ix86_output_function_return.
12399 (simple_return_internal_long): Likewise.
12400 * config/i386/i386.opt (mfunction-return=): New option.
12401 (indirect_branch): Mention -mfunction-return=.
12402 * doc/extend.texi: Document function_return function attribute.
12403 * doc/invoke.texi: Document -mfunction-return= option.
12405 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12407 * config/i386/i386-opts.h (indirect_branch): New.
12408 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12409 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12410 with local indirect jump when converting indirect call and jump.
12411 (ix86_set_indirect_branch_type): New.
12412 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12413 (indirectlabelno): New.
12414 (indirect_thunk_needed): Likewise.
12415 (indirect_thunk_bnd_needed): Likewise.
12416 (indirect_thunks_used): Likewise.
12417 (indirect_thunks_bnd_used): Likewise.
12418 (INDIRECT_LABEL): Likewise.
12419 (indirect_thunk_name): Likewise.
12420 (output_indirect_thunk): Likewise.
12421 (output_indirect_thunk_function): Likewise.
12422 (ix86_output_indirect_branch_via_reg): Likewise.
12423 (ix86_output_indirect_branch_via_push): Likewise.
12424 (ix86_output_indirect_branch): Likewise.
12425 (ix86_output_indirect_jmp): Likewise.
12426 (ix86_code_end): Call output_indirect_thunk_function if needed.
12427 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12429 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12430 (ix86_attribute_table): Add indirect_branch.
12431 * config/i386/i386.h (machine_function): Add indirect_branch_type
12432 and has_local_indirect_jump.
12433 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12435 (tablejump): Likewise.
12436 (*indirect_jump): Use ix86_output_indirect_jmp.
12437 (*tablejump_1): Likewise.
12438 (simple_return_indirect_internal): Likewise.
12439 * config/i386/i386.opt (mindirect-branch=): New option.
12440 (indirect_branch): New.
12443 (thunk-inline): Likewise.
12444 (thunk-extern): Likewise.
12445 * doc/extend.texi: Document indirect_branch function attribute.
12446 * doc/invoke.texi: Document -mindirect-branch= option.
12448 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12451 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12453 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12455 * ipa-inline.c (want_inline_small_function_p): Return false if
12456 inlining has already failed with CIF_FINAL_ERROR.
12457 (update_caller_keys): Call want_inline_small_function_p before
12459 (update_callee_keys): Likewise.
12461 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12463 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12465 (rs6000_quadword_masked_address_p): Likewise.
12466 (quad_aligned_load_p): Likewise.
12467 (quad_aligned_store_p): Likewise.
12468 (const_load_sequence_p): Add comment to describe the outer-most loop.
12469 (mimic_memory_attributes_and_flags): New function.
12470 (rs6000_gen_stvx): Likewise.
12471 (replace_swapped_aligned_store): Likewise.
12472 (rs6000_gen_lvx): Likewise.
12473 (replace_swapped_aligned_load): Likewise.
12474 (replace_swapped_load_constant): Capitalize argument name in
12475 comment describing this function.
12476 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12477 and stores that access quad-word aligned addresses and replace
12478 with stvx or lvx instructions when appropriate.
12479 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12480 New function prototype.
12481 (rs6000_quadword_masked_address_p): Likewise.
12482 (rs6000_gen_lvx): Likewise.
12483 (rs6000_gen_stvx): Likewise.
12484 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12485 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12486 when memory address is aligned.
12487 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12488 this split to select lvx instruction when memory address is aligned.
12489 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12490 instruction when memory address is aligned.
12491 (*vsx_le_perm_load_v16qi): Likewise.
12492 (four unnamed splitters): Modify to select the stvx instruction
12493 when memory is aligned.
12495 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12497 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12498 which appears in the queue multiple times.
12500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12501 Alan Hayward <alan.hayward@arm.com>
12502 David Sherwood <david.sherwood@arm.com>
12504 * tree-vectorizer.h (vec_lower_bound): New structure.
12505 (_loop_vec_info): Add check_nonzero and lower_bounds.
12506 (LOOP_VINFO_CHECK_NONZERO): New macro.
12507 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12508 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12509 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12510 fields. Make seg_len the distance travelled, not including the
12512 (dr_direction_indicator): Declare.
12513 (dr_zero_step_indicator): Likewise.
12514 (dr_known_forward_stride_p): Likewise.
12515 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12517 (runtime_alias_check_p): Allow runtime alias checks with
12519 (operator ==): Compare access_size and align.
12520 (prune_runtime_alias_test_list): Rework for new distinction between
12521 the access_size and seg_len.
12522 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12524 (get_segment_min_max): New function.
12525 (create_intersect_range_checks): Use it.
12526 (dr_step_indicator): New function.
12527 (dr_direction_indicator): Likewise.
12528 (dr_zero_step_indicator): Likewise.
12529 (dr_known_forward_stride_p): Likewise.
12530 * tree-loop-distribution.c (data_ref_segment_size): Return
12531 DR_STEP * (niters - 1).
12532 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12534 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12535 (vect_preserves_scalar_order_p): New function, split out from...
12536 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12537 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12538 (vect_vfa_access_size): New function.
12539 (vect_vfa_align): Likewise.
12540 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12541 (dump_lower_bound): New function.
12542 (vect_check_lower_bound): Likewise.
12543 (vect_small_gap_p): Likewise.
12544 (vectorizable_with_step_bound_p): Likewise.
12545 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12546 depencies if the vectorization factor is 1. Convert the checks
12547 for nonzero steps into checks on the bounds of DR_STEP. Try using
12548 a bunds check for variable steps if the minimum required step is
12549 relatively small. Update calls to the dr_with_seg_len
12550 constructor and to vect_compile_time_alias.
12551 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12553 (vect_loop_versioning): Call it.
12554 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12556 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12559 Alan Hayward <alan.hayward@arm.com>
12560 David Sherwood <david.sherwood@arm.com>
12562 * doc/sourcebuild.texi (vect_scatter_store): Document.
12563 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12565 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12567 * genopinit.c (main): Add supports_vec_scatter_store and
12568 supports_vec_scatter_store_cached to target_optabs.
12569 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12570 IFN_MASK_SCATTER_STORE.
12571 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12573 * internal-fn.h (internal_store_fn_p): Declare.
12574 (internal_fn_stored_value_index): Likewise.
12575 * internal-fn.c (scatter_store_direct): New macro.
12576 (expand_scatter_store_optab_fn): New function.
12577 (direct_scatter_store_optab_supported_p): New macro.
12578 (internal_store_fn_p): New function.
12579 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12580 IFN_MASK_SCATTER_STORE.
12581 (internal_fn_mask_index): Likewise.
12582 (internal_fn_stored_value_index): New function.
12583 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12584 for scatter stores.
12585 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12586 * optabs-query.c (supports_vec_scatter_store_p): New function.
12587 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12588 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12589 true for scatter stores.
12590 (vect_gather_scatter_fn_p): Handle scatter stores too.
12591 (vect_check_gather_scatter): Consider using scatter stores if
12592 supports_vec_scatter_store_p.
12593 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12594 scatter stores too.
12595 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12596 internal_fn_stored_value_index.
12597 (check_load_store_masking): Handle scatter stores too.
12598 (vect_get_store_rhs): Make public.
12599 (vectorizable_call): Use internal_store_fn_p.
12600 (vectorizable_store): Handle scatter store internal functions.
12601 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12602 when deciding whether the end of the group has been reached.
12603 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12604 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12605 (mask_scatter_store<mode>): New insns.
12607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12608 Alan Hayward <alan.hayward@arm.com>
12609 David Sherwood <david.sherwood@arm.com>
12611 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12612 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12613 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12615 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12616 Use vect_truncate_gather_scatter_offset if we can't treat the
12617 operation as a normal gather load or scatter store.
12618 (get_group_load_store_type): Take the gather_scatter_info
12619 as argument. Try using a gather load or scatter store for
12620 single-element groups.
12621 (get_load_store_type): Update calls to get_group_load_store_type
12622 and vect_use_strided_gather_scatters_p.
12624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12625 Alan Hayward <alan.hayward@arm.com>
12626 David Sherwood <david.sherwood@arm.com>
12628 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12629 optional tree argument.
12630 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12632 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12633 but continue to use the current value as a fallback.
12634 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12635 to compare the updates.
12636 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12637 (get_load_store_type): Use it when handling a strided access.
12638 (vect_get_strided_load_store_ops): New function.
12639 (vect_get_data_ptr_increment): Likewise.
12640 (vectorizable_load): Handle strided gather loads. Always pass
12641 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12644 Alan Hayward <alan.hayward@arm.com>
12645 David Sherwood <david.sherwood@arm.com>
12647 * doc/md.texi (gather_load@var{m}): Document.
12648 (mask_gather_load@var{m}): Likewise.
12649 * genopinit.c (main): Add supports_vec_gather_load and
12650 supports_vec_gather_load_cached to target_optabs.
12651 * optabs-tree.c (init_tree_optimization_optabs): Use
12652 ggc_cleared_alloc to allocate target_optabs.
12653 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12654 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12656 * internal-fn.h (internal_load_fn_p): Declare.
12657 (internal_gather_scatter_fn_p): Likewise.
12658 (internal_fn_mask_index): Likewise.
12659 (internal_gather_scatter_fn_supported_p): Likewise.
12660 * internal-fn.c (gather_load_direct): New macro.
12661 (expand_gather_load_optab_fn): New function.
12662 (direct_gather_load_optab_supported_p): New macro.
12663 (direct_internal_fn_optab): New function.
12664 (internal_load_fn_p): Likewise.
12665 (internal_gather_scatter_fn_p): Likewise.
12666 (internal_fn_mask_index): Likewise.
12667 (internal_gather_scatter_fn_supported_p): Likewise.
12668 * optabs-query.c (supports_at_least_one_mode_p): New function.
12669 (supports_vec_gather_load_p): Likewise.
12670 * optabs-query.h (supports_vec_gather_load_p): Declare.
12671 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12672 and memory_type field.
12673 (NUM_PATTERNS): Bump to 15.
12674 * tree-vect-data-refs.c: Include internal-fn.h.
12675 (vect_gather_scatter_fn_p): New function.
12676 (vect_describe_gather_scatter_call): Likewise.
12677 (vect_check_gather_scatter): Try using internal functions for
12678 gather loads. Recognize existing calls to a gather load function.
12679 (vect_analyze_data_refs): Consider using gather loads if
12680 supports_vec_gather_load_p.
12681 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12682 (vect_get_gather_scatter_offset_type): Likewise.
12683 (vect_convert_mask_for_vectype): Likewise.
12684 (vect_add_conversion_to_patterm): Likewise.
12685 (vect_try_gather_scatter_pattern): Likewise.
12686 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12687 (vect_vect_recog_func_ptrs): Add it.
12688 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12689 internal_fn_mask_index and internal_gather_scatter_fn_p.
12690 (check_load_store_masking): Take the gather_scatter_info as an
12691 argument and handle gather loads.
12692 (vect_get_gather_scatter_ops): New function.
12693 (vectorizable_call): Check internal_load_fn_p.
12694 (vectorizable_load): Likewise. Handle gather load internal
12696 (vectorizable_store): Update call to check_load_store_masking.
12697 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12698 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12699 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12700 (aarch64_gather_scale_operand_d): New predicates.
12701 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12702 (mask_gather_load<mode>): New insns.
12704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12705 Alan Hayward <alan.hayward@arm.com>
12706 David Sherwood <david.sherwood@arm.com>
12708 * optabs.def (fold_left_plus_optab): New optab.
12709 * doc/md.texi (fold_left_plus_@var{m}): Document.
12710 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12711 * internal-fn.c (fold_left_direct): Define.
12712 (expand_fold_left_optab_fn): Likewise.
12713 (direct_fold_left_optab_supported_p): Likewise.
12714 * fold-const-call.c (fold_const_fold_left): New function.
12715 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12716 * tree-parloops.c (valid_reduction_p): New function.
12717 (gather_scalar_reductions): Use it.
12718 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12719 (vect_finish_replace_stmt): Declare.
12720 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12721 (needs_fold_left_reduction_p): New function, split out from...
12722 (vect_is_simple_reduction): ...here. Accept reductions that
12723 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12724 (vect_force_simple_reduction): Also store the reduction type in
12725 the assignment's STMT_VINFO_REDUC_TYPE.
12726 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12727 (merge_with_identity): New function.
12728 (vect_expand_fold_left): Likewise.
12729 (vectorize_fold_left_reduction): Likewise.
12730 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12731 scalar phi in place for it. Check for target support and reject
12732 cases that would reassociate the operation. Defer the transform
12733 phase to vectorize_fold_left_reduction.
12734 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12735 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12736 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12740 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12741 call to ifc_temp_var.
12743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12744 Alan Hayward <alan.hayward@arm.com>
12745 David Sherwood <david.sherwood@arm.com>
12747 * target.def (legitimize_address_displacement): Take the original
12748 offset as a poly_int.
12749 * targhooks.h (default_legitimize_address_displacement): Update
12751 * targhooks.c (default_legitimize_address_displacement): Likewise.
12752 * doc/tm.texi: Regenerate.
12753 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12754 as an argument, moving assert of ad->disp == ad->disp_term to...
12755 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12756 Try calling targetm.legitimize_address_displacement before expanding
12757 the address rather than afterwards, and adjust for the new interface.
12758 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12759 Match the new hook interface. Handle SVE addresses.
12760 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12761 new hook interface.
12763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12765 * Makefile.in (OBJS): Add early-remat.o.
12766 * target.def (select_early_remat_modes): New hook.
12767 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12768 * doc/tm.texi: Regenerate.
12769 * targhooks.h (default_select_early_remat_modes): Declare.
12770 * targhooks.c (default_select_early_remat_modes): New function.
12771 * timevar.def (TV_EARLY_REMAT): New timevar.
12772 * passes.def (pass_early_remat): New pass.
12773 * tree-pass.h (make_pass_early_remat): Declare.
12774 * early-remat.c: New file.
12775 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12777 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12780 Alan Hayward <alan.hayward@arm.com>
12781 David Sherwood <david.sherwood@arm.com>
12783 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12784 vfm1 with a bound_epilog parameter.
12785 (vect_do_peeling): Update calls accordingly, and move the prologue
12786 call earlier in the function. Treat the base bound_epilog as 0 for
12787 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12788 this base when peeling for gaps.
12789 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12790 with fully-masked loops.
12791 (vect_estimate_min_profitable_iters): Handle the single peeled
12792 iteration in that case.
12794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12795 Alan Hayward <alan.hayward@arm.com>
12796 David Sherwood <david.sherwood@arm.com>
12798 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12799 single-element interleaving even if the size is not a power of 2.
12800 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12801 accesses for single-element interleaving if the group size is
12804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12805 Alan Hayward <alan.hayward@arm.com>
12806 David Sherwood <david.sherwood@arm.com>
12808 * doc/md.texi (fold_extract_last_@var{m}): Document.
12809 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12810 * optabs.def (fold_extract_last_optab): New optab.
12811 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12812 * internal-fn.c (fold_extract_direct): New macro.
12813 (expand_fold_extract_optab_fn): Likewise.
12814 (direct_fold_extract_optab_supported_p): Likewise.
12815 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12816 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12817 EXTRACT_LAST_REDUCTION.
12818 (get_initial_def_for_reduction): Do not create an initial vector
12819 for EXTRACT_LAST_REDUCTION reductions.
12820 (vectorizable_reduction): Leave the scalar phi in place for
12821 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12822 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12823 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12824 transform phase to vectorizable_condition.
12825 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12827 (vect_finish_stmt_generation): ...here.
12828 (vect_finish_replace_stmt): New function.
12829 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12830 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12832 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12835 Alan Hayward <alan.hayward@arm.com>
12836 David Sherwood <david.sherwood@arm.com>
12838 * doc/md.texi (extract_last_@var{m}): Document.
12839 * optabs.def (extract_last_optab): New optab.
12840 * internal-fn.def (EXTRACT_LAST): New internal function.
12841 * internal-fn.c (cond_unary_direct): New macro.
12842 (expand_cond_unary_optab_fn): Likewise.
12843 (direct_cond_unary_optab_supported_p): Likewise.
12844 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12845 loops using EXTRACT_LAST.
12846 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12847 (extract_last_<mode>): ...this optab.
12848 (vec_extract<mode><Vel>): Update accordingly.
12850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12851 Alan Hayward <alan.hayward@arm.com>
12852 David Sherwood <david.sherwood@arm.com>
12854 * target.def (empty_mask_is_expensive): New hook.
12855 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12856 * doc/tm.texi: Regenerate.
12857 * targhooks.h (default_empty_mask_is_expensive): Declare.
12858 * targhooks.c (default_empty_mask_is_expensive): New function.
12859 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12860 if the target says that empty masks are expensive.
12861 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12863 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12865 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12866 Alan Hayward <alan.hayward@arm.com>
12867 David Sherwood <david.sherwood@arm.com>
12869 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12870 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12871 (vect_use_loop_mask_for_alignment_p): New function.
12872 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12873 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12874 niters_skip argument. Make sure that the first niters_skip elements
12875 of the first iteration are inactive.
12876 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12877 Update call to vect_set_loop_masks_directly.
12878 (get_misalign_in_elems): New function, split out from...
12879 (vect_gen_prolog_loop_niters): ...here.
12880 (vect_update_init_of_dr): Take a code argument that specifies whether
12881 the adjustment should be added or subtracted.
12882 (vect_update_init_of_drs): Likewise.
12883 (vect_prepare_for_masked_peels): New function.
12884 (vect_do_peeling): Skip prologue peeling if we're using a mask
12885 instead. Update call to vect_update_inits_of_drs.
12886 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12888 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12889 alignment. Do not include the number of peeled iterations in
12890 the minimum threshold in that case.
12891 (vectorizable_induction): Adjust the start value down by
12892 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12893 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12894 Take the number of skipped iterations into account when calculating
12896 * tree-vect-stmts.c (vect_gen_while_not): New function.
12898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12899 Alan Hayward <alan.hayward@arm.com>
12900 David Sherwood <david.sherwood@arm.com>
12902 * doc/sourcebuild.texi (vect_fully_masked): Document.
12903 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12904 default value to 0.
12905 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12907 (vect_analyze_loop_2): ...here. Don't check the vectorization
12908 factor against the number of loop iterations if the loop is
12911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12912 Alan Hayward <alan.hayward@arm.com>
12913 David Sherwood <david.sherwood@arm.com>
12915 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12916 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12917 (dump_groups): Update accordingly.
12918 (iv_use::mem_type): New member variable.
12919 (address_p): New function.
12920 (record_use): Add a mem_type argument and initialize the new
12922 (record_group_use): Add a mem_type argument. Use address_p.
12923 Remove obsolete null checks of base_object. Update call to record_use.
12924 (find_interesting_uses_op): Update call to record_group_use.
12925 (find_interesting_uses_cond): Likewise.
12926 (find_interesting_uses_address): Likewise.
12927 (get_mem_type_for_internal_fn): New function.
12928 (find_address_like_use): Likewise.
12929 (find_interesting_uses_stmt): Try find_address_like_use before
12930 calling find_interesting_uses_op.
12931 (addr_offset_valid_p): Use the iv mem_type field as the type
12932 of the addressed memory.
12933 (add_autoinc_candidates): Likewise.
12934 (get_address_cost): Likewise.
12935 (split_small_address_groups_p): Use address_p.
12936 (split_address_groups): Likewise.
12937 (add_iv_candidate_for_use): Likewise.
12938 (autoinc_possible_for_pair): Likewise.
12939 (rewrite_groups): Likewise.
12940 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
12941 (determine_group_iv_cost): Update after split of USE_ADDRESS.
12942 (get_alias_ptr_type_for_ptr_address): New function.
12943 (rewrite_use_address): Rewrite address uses in calls that were
12944 identified by find_address_like_use.
12946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12947 Alan Hayward <alan.hayward@arm.com>
12948 David Sherwood <david.sherwood@arm.com>
12950 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
12952 * gimple-expr.h (is_gimple_addressable: Likewise.
12953 * gimple-expr.c (is_gimple_address): Likewise.
12954 * internal-fn.c (expand_call_mem_ref): New function.
12955 (expand_mask_load_optab_fn): Use it.
12956 (expand_mask_store_optab_fn): Likewise.
12958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12959 Alan Hayward <alan.hayward@arm.com>
12960 David Sherwood <david.sherwood@arm.com>
12962 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
12963 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
12964 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
12965 (cond_umax@var{mode}): Document.
12966 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
12967 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
12968 (cond_umin_optab, cond_umax_optab): New optabs.
12969 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
12970 (COND_IOR, COND_XOR): New internal functions.
12971 * internal-fn.h (get_conditional_internal_fn): Declare.
12972 * internal-fn.c (cond_binary_direct): New macro.
12973 (expand_cond_binary_optab_fn): Likewise.
12974 (direct_cond_binary_optab_supported_p): Likewise.
12975 (get_conditional_internal_fn): New function.
12976 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
12977 Cope with reduction statements that are vectorized as calls rather
12979 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
12980 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
12981 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
12982 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
12983 (UNSPEC_COND_EOR): New unspecs.
12984 (optab): Add mappings for them.
12985 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
12986 (sve_int_op, sve_fp_op): New int attributes.
12988 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12989 Alan Hayward <alan.hayward@arm.com>
12990 David Sherwood <david.sherwood@arm.com>
12992 * optabs.def (while_ult_optab): New optab.
12993 * doc/md.texi (while_ult@var{m}@var{n}): Document.
12994 * internal-fn.def (WHILE_ULT): New internal function.
12995 * internal-fn.h (direct_internal_fn_supported_p): New override
12996 that takes two types as argument.
12997 * internal-fn.c (while_direct): New macro.
12998 (expand_while_optab_fn): New function.
12999 (convert_optab_supported_p): Likewise.
13000 (direct_while_optab_supported_p): New macro.
13001 * wide-int.h (wi::udiv_ceil): New function.
13002 * tree-vectorizer.h (rgroup_masks): New structure.
13003 (vec_loop_masks): New typedef.
13004 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13005 and fully_masked_p.
13006 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13007 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13008 (vect_max_vf): New function.
13009 (slpeel_make_loop_iterate_ntimes): Delete.
13010 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13011 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13012 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13013 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13014 internal-fn.h, stor-layout.h and optabs-query.h.
13015 (vect_set_loop_mask): New function.
13016 (add_preheader_seq): Likewise.
13017 (add_header_seq): Likewise.
13018 (interleave_supported_p): Likewise.
13019 (vect_maybe_permute_loop_masks): Likewise.
13020 (vect_set_loop_masks_directly): Likewise.
13021 (vect_set_loop_condition_masked): Likewise.
13022 (vect_set_loop_condition_unmasked): New function, split out from
13023 slpeel_make_loop_iterate_ntimes.
13024 (slpeel_make_loop_iterate_ntimes): Rename to..
13025 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13026 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13027 (vect_do_peeling): Update call accordingly.
13028 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13030 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13031 mask_compare_type, can_fully_mask_p and fully_masked_p.
13032 (release_vec_loop_masks): New function.
13033 (_loop_vec_info): Use it to free the loop masks.
13034 (can_produce_all_loop_masks_p): New function.
13035 (vect_get_max_nscalars_per_iter): Likewise.
13036 (vect_verify_full_masking): Likewise.
13037 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13038 retries, and free the mask rgroups before retrying. Check loop-wide
13039 reasons for disallowing fully-masked loops. Make the final decision
13040 about whether use a fully-masked loop or not.
13041 (vect_estimate_min_profitable_iters): Do not assume that peeling
13042 for the number of iterations will be needed for fully-masked loops.
13043 (vectorizable_reduction): Disable fully-masked loops.
13044 (vectorizable_live_operation): Likewise.
13045 (vect_halve_mask_nunits): New function.
13046 (vect_double_mask_nunits): Likewise.
13047 (vect_record_loop_mask): Likewise.
13048 (vect_get_loop_mask): Likewise.
13049 (vect_transform_loop): Handle the case in which the final loop
13050 iteration might handle a partial vector. Call vect_set_loop_condition
13051 instead of slpeel_make_loop_iterate_ntimes.
13052 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13053 (check_load_store_masking): New function.
13054 (prepare_load_store_mask): Likewise.
13055 (vectorizable_store): Handle fully-masked loops.
13056 (vectorizable_load): Likewise.
13057 (supportable_widening_operation): Use vect_halve_mask_nunits for
13059 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13060 (vect_gen_while): New function.
13061 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13062 (aarch64_uqdec<mode>): New insn.
13064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13065 Alan Hayward <alan.hayward@arm.com>
13066 David Sherwood <david.sherwood@arm.com>
13068 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13069 (reduc_xor_scal_optab): New optabs.
13070 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13071 (reduc_xor_scal_@var{m}): Document.
13072 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13073 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13074 internal functions.
13075 * fold-const-call.c (fold_const_call): Handle them.
13076 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13077 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13078 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13079 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13080 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13081 (UNSPEC_XORV): New unspecs.
13082 (optab): Add entries for them.
13083 (BITWISEV): New int iterator.
13084 (bit_reduc_op): New int attributes.
13086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13087 Alan Hayward <alan.hayward@arm.com>
13088 David Sherwood <david.sherwood@arm.com>
13090 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13091 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13092 * optabs.def (vec_shl_insert_optab): New optab.
13093 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13094 (duplicate_and_interleave): Likewise.
13095 * tree-vect-loop.c: Include internal-fn.h.
13096 (neutral_op_for_slp_reduction): New function, split out from
13097 get_initial_defs_for_reduction.
13098 (get_initial_def_for_reduction): Handle option 2 for variable-length
13099 vectors by loading the neutral value into a vector and then shifting
13100 the initial value into element 0.
13101 (get_initial_defs_for_reduction): Replace the code argument with
13102 the neutral value calculated by neutral_op_for_slp_reduction.
13103 Use gimple_build_vector for constant-length vectors.
13104 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13105 but the first group_size elements have a neutral value.
13106 Use duplicate_and_interleave otherwise.
13107 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13108 Update call to get_initial_defs_for_reduction. Handle SLP
13109 reductions for variable-length vectors by creating one vector
13110 result for each scalar result, with the elements associated
13111 with other scalar results stubbed out with the neutral value.
13112 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13113 Require IFN_VEC_SHL_INSERT for double reductions on
13114 variable-length vectors, or SLP reductions that have
13115 a neutral value. Require can_duplicate_and_interleave_p
13116 support for variable-length unchained SLP reductions if there
13117 is no neutral value, such as for MIN/MAX reductions. Also require
13118 the number of vector elements to be a multiple of the number of
13119 SLP statements when doing variable-length unchained SLP reductions.
13120 Update call to vect_create_epilog_for_reduction.
13121 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13122 and remove initial values.
13123 (duplicate_and_interleave): Make public.
13124 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13125 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13128 Alan Hayward <alan.hayward@arm.com>
13129 David Sherwood <david.sherwood@arm.com>
13131 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13132 (can_duplicate_and_interleave_p): New function.
13133 (vect_get_and_check_slp_defs): Take the vector of statements
13134 rather than just the current one. Remove excess parentheses.
13135 Restriction rejectinon of vect_constant_def and vect_external_def
13136 for variable-length vectors to boolean types, or types for which
13137 can_duplicate_and_interleave_p is false.
13138 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13139 (duplicate_and_interleave): New function.
13140 (vect_get_constant_vectors): Use gimple_build_vector for
13141 constant-length vectors and suitable variable-length constant
13142 vectors. Use duplicate_and_interleave for other variable-length
13143 vectors. Don't defer the update when inserting new statements.
13145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13146 Alan Hayward <alan.hayward@arm.com>
13147 David Sherwood <david.sherwood@arm.com>
13149 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13150 min_profitable_iters doesn't go negative.
13152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13153 Alan Hayward <alan.hayward@arm.com>
13154 David Sherwood <david.sherwood@arm.com>
13156 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13157 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13158 * optabs.def (vec_mask_load_lanes_optab): New optab.
13159 (vec_mask_store_lanes_optab): Likewise.
13160 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13161 (MASK_STORE_LANES): Likewise.
13162 * internal-fn.c (mask_load_lanes_direct): New macro.
13163 (mask_store_lanes_direct): Likewise.
13164 (expand_mask_load_optab_fn): Handle masked operations.
13165 (expand_mask_load_lanes_optab_fn): New macro.
13166 (expand_mask_store_optab_fn): Handle masked operations.
13167 (expand_mask_store_lanes_optab_fn): New macro.
13168 (direct_mask_load_lanes_optab_supported_p): Likewise.
13169 (direct_mask_store_lanes_optab_supported_p): Likewise.
13170 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13172 (vect_load_lanes_supported): Likewise.
13173 * tree-vect-data-refs.c (strip_conversion): New function.
13174 (can_group_stmts_p): Likewise.
13175 (vect_analyze_data_ref_accesses): Use it instead of checking
13176 for a pair of assignments.
13177 (vect_store_lanes_supported): Take a masked_p parameter.
13178 (vect_load_lanes_supported): Likewise.
13179 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13180 vect_store_lanes_supported and vect_load_lanes_supported.
13181 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13182 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13183 parameter. Don't allow gaps for masked accesses.
13184 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13185 and vect_load_lanes_supported.
13186 (get_load_store_type): Take a masked_p parameter and update
13187 call to get_group_load_store_type.
13188 (vectorizable_store): Update call to get_load_store_type.
13189 Handle IFN_MASK_STORE_LANES.
13190 (vectorizable_load): Update call to get_load_store_type.
13191 Handle IFN_MASK_LOAD_LANES.
13193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13194 Alan Hayward <alan.hayward@arm.com>
13195 David Sherwood <david.sherwood@arm.com>
13197 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13199 * config/aarch64/aarch64-protos.h
13200 (aarch64_sve_struct_memory_operand_p): Declare.
13201 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13202 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13203 (VPRED, vpred): Handle SVE structure modes.
13204 * config/aarch64/constraints.md (Utx): New constraint.
13205 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13206 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13207 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13208 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13209 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13210 structure modes. Split into pieces after RA.
13211 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13212 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13214 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13215 SVE structure modes.
13216 (aarch64_classify_address): Likewise.
13217 (sizetochar): Move earlier in file.
13218 (aarch64_print_operand): Handle SVE register lists.
13219 (aarch64_array_mode): New function.
13220 (aarch64_sve_struct_memory_operand_p): Likewise.
13221 (TARGET_ARRAY_MODE): Redefine.
13223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13224 Alan Hayward <alan.hayward@arm.com>
13225 David Sherwood <david.sherwood@arm.com>
13227 * target.def (array_mode): New target hook.
13228 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13229 * doc/tm.texi: Regenerate.
13230 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13231 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13232 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13233 targetm.array_mode.
13234 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13238 Alan Hayward <alan.hayward@arm.com>
13239 David Sherwood <david.sherwood@arm.com>
13241 * fold-const.c (fold_binary_loc): Check the argument types
13242 rather than the result type when testing for a vector operation.
13244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13246 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13247 * doc/tm.texi: Regenerate.
13249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13250 Alan Hayward <alan.hayward@arm.com>
13251 David Sherwood <david.sherwood@arm.com>
13253 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13254 (sve): Document new AArch64 extension.
13255 * doc/md.texi (w): Extend the description of the AArch64
13256 constraint to include SVE vectors.
13257 (Upl, Upa): Document new AArch64 predicate constraints.
13258 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13260 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13261 (msve-vector-bits=): New option.
13262 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13263 SVE when these are disabled.
13264 (sve): New extension.
13265 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13266 modes. Adjust their number of units based on aarch64_sve_vg.
13267 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13268 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13269 aarch64_addr_query_type.
13270 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13271 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13272 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13273 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13274 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13275 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13276 (aarch64_simd_imm_zero_p): Delete.
13277 (aarch64_check_zero_based_sve_index_immediate): Declare.
13278 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13279 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13280 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13281 (aarch64_sve_float_mul_immediate_p): Likewise.
13282 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13283 rather than an rtx.
13284 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13285 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13286 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13287 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13288 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13289 (aarch64_regmode_natural_size): Likewise.
13290 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13291 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13293 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13294 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13295 for VG and the SVE predicate registers.
13296 (V_ALIASES): Add a "z"-prefixed alias.
13297 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13298 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13299 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13300 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13301 (REG_CLASS_NAMES): Add entries for them.
13302 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13303 and the predicate registers.
13304 (aarch64_sve_vg): Declare.
13305 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13306 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13307 (REGMODE_NATURAL_SIZE): Define.
13308 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13310 * config/aarch64/aarch64.c: Include cfgrtl.h.
13311 (simd_immediate_info): Add a constructor for series vectors,
13312 and an associated step field.
13313 (aarch64_sve_vg): New variable.
13314 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13315 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13316 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13317 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13318 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13319 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13320 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13321 (aarch64_get_mask_mode): New functions.
13322 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13323 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13324 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13325 predicate modes and predicate registers. Explicitly restrict
13326 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13327 to store a vector mode if it is recognized by
13328 aarch64_classify_vector_mode.
13329 (aarch64_regmode_natural_size): New function.
13330 (aarch64_hard_regno_caller_save_mode): Return the original mode
13332 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13333 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13334 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13335 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13337 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13338 does not overlap dest if the function is frame-related. Handle
13340 (aarch64_split_add_offset): New function.
13341 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13342 them aarch64_add_offset.
13343 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13344 and update call to aarch64_sub_sp.
13345 (aarch64_add_cfa_expression): New function.
13346 (aarch64_expand_prologue): Pass extra temporary registers to the
13347 functions above. Handle the case in which we need to emit new
13348 DW_CFA_expressions for registers that were originally saved
13349 relative to the stack pointer, but now have to be expressed
13350 relative to the frame pointer.
13351 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13353 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13354 IP0 and IP1 values for SVE frames.
13355 (aarch64_expand_vec_series): New function.
13356 (aarch64_expand_sve_widened_duplicate): Likewise.
13357 (aarch64_expand_sve_const_vector): Likewise.
13358 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13359 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13360 into the register, rather than emitting a SET directly.
13361 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13362 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13363 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13364 (offset_9bit_signed_scaled_p): New functions.
13365 (aarch64_replicate_bitmask_imm): New function.
13366 (aarch64_bitmask_imm): Use it.
13367 (aarch64_cannot_force_const_mem): Reject expressions involving
13368 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13369 (aarch64_classify_index): Handle SVE indices, by requiring
13370 a plain register index with a scale that matches the element size.
13371 (aarch64_classify_address): Handle SVE addresses. Assert that
13372 the mode of the address is VOIDmode or an integer mode.
13373 Update call to aarch64_classify_symbol.
13374 (aarch64_classify_symbolic_expression): Update call to
13375 aarch64_classify_symbol.
13376 (aarch64_const_vec_all_in_range_p): New function.
13377 (aarch64_print_vector_float_operand): Likewise.
13378 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13379 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13380 and the FP immediates 1.0 and 0.5.
13381 (aarch64_print_address_internal): Handle SVE addresses.
13382 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13383 (aarch64_regno_regclass): Handle predicate registers.
13384 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13386 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13387 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13388 (aarch64_convert_sve_vector_bits): New function.
13389 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13390 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13391 rather than an rtx.
13392 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13393 Handle SVE vector and predicate modes. Accept VL-based constants
13394 that need only one temporary register, and VL offsets that require
13395 no temporary registers.
13396 (aarch64_conditional_register_usage): Mark the predicate registers
13397 as fixed if SVE isn't available.
13398 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13399 Return true for SVE vector and predicate modes.
13400 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13401 rather than an unsigned int. Handle SVE modes.
13402 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13404 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13406 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13407 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13408 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13409 (aarch64_sve_float_mul_immediate_p): New functions.
13410 (aarch64_sve_valid_immediate): New function.
13411 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13412 Explicitly reject structure modes. Check for INDEX constants.
13413 Handle PTRUE and PFALSE constants.
13414 (aarch64_check_zero_based_sve_index_immediate): New function.
13415 (aarch64_simd_imm_zero_p): Delete.
13416 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13417 vector modes. Accept constants in the range of CNT[BHWD].
13418 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13419 ask for an Advanced SIMD mode.
13420 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13421 (aarch64_simd_vector_alignment): Handle SVE predicates.
13422 (aarch64_vectorize_preferred_vector_alignment): New function.
13423 (aarch64_simd_vector_alignment_reachable): Use it instead of
13425 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13426 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13428 (MAX_VECT_LEN): Delete.
13429 (expand_vec_perm_d): Add a vec_flags field.
13430 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13431 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13432 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13434 (aarch64_evpc_rev): Rename to...
13435 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13436 (aarch64_evpc_rev_global): New function.
13437 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13438 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13440 (aarch64_evpc_sve_tbl): New function.
13441 (aarch64_expand_vec_perm_const_1): Update after rename of
13442 aarch64_evpc_rev. Handle SVE permutes too, trying
13443 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13444 than aarch64_evpc_tbl.
13445 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13446 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13447 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13448 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13449 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13450 (aarch64_expand_sve_vcond): New functions.
13451 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13452 of aarch64_vector_mode_p.
13453 (aarch64_dwarf_poly_indeterminate_value): New function.
13454 (aarch64_compute_pressure_classes): Likewise.
13455 (aarch64_can_change_mode_class): Likewise.
13456 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13457 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13458 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13459 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13460 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13461 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13462 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13463 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13465 (Dn, Dl, Dr): Accept const as well as const_vector.
13466 (Dz): Likewise. Compare against CONST0_RTX.
13467 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13468 of "vector" where appropriate.
13469 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13470 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13471 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13472 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13473 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13474 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13475 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13476 (v_int_equiv): Extend to SVE modes.
13477 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13479 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13480 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13481 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13482 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13483 (SVE_COND_FP_CMP): New int iterators.
13484 (perm_hilo): Handle the new unpack unspecs.
13485 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13487 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13488 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13489 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13490 (aarch64_equality_operator, aarch64_constant_vector_operand)
13491 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13492 (aarch64_sve_nonimmediate_operand): Likewise.
13493 (aarch64_sve_general_operand): Likewise.
13494 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13495 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13496 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13497 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13498 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13499 (aarch64_sve_float_arith_immediate): Likewise.
13500 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13501 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13502 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13503 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13504 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13505 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13506 (aarch64_sve_float_arith_operand): Likewise.
13507 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13508 (aarch64_sve_float_mul_operand): Likewise.
13509 (aarch64_sve_vec_perm_operand): Likewise.
13510 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13511 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13512 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13513 as well as const_vector.
13514 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13515 in file. Use CONST0_RTX and CONSTM1_RTX.
13516 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13517 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13518 Use aarch64_simd_imm_zero.
13519 * config/aarch64/aarch64-sve.md: New file.
13520 * config/aarch64/aarch64.md: Include it.
13521 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13522 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13523 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13524 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13525 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13526 (sve): New attribute.
13527 (enabled): Disable instructions with the sve attribute unless
13529 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13530 aarch64_expand_mov_immediate.
13531 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13532 CNT[BHSD] immediates.
13533 (movti): Split CONST_POLY_INT moves into two halves.
13534 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13535 Split additions that need a temporary here if the destination
13536 is the stack pointer.
13537 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13538 (*add<mode>3_poly_1): New instruction.
13539 (set_clobber_cc): New expander.
13541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13543 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13544 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13545 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13546 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13547 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13548 Change innermode from fixed_mode_size to machine_mode.
13549 (simplify_subreg): Update call accordingly. Handle a constant-sized
13550 subreg of a variable-length CONST_VECTOR.
13552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13553 Alan Hayward <alan.hayward@arm.com>
13554 David Sherwood <david.sherwood@arm.com>
13556 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13557 (add_offset_to_base): New function, split out from...
13558 (create_mem_ref): ...here. When handling a scale other than 1,
13559 check first whether the address is valid without the offset.
13560 Add it into the base if so, leaving the index and scale as-is.
13562 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13565 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13566 fold_for_warn before checking if arg2 is INTEGER_CST.
13568 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13570 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13571 (store_multiple_operation): Delete.
13572 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13573 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13574 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13575 guarded by TARGET_STRING.
13576 (rs6000_output_load_multiple): Delete.
13577 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13578 OPTION_MASK_STRING / TARGET_STRING handling.
13579 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13580 (const rs6000_opt_masks) <"string">: Change mask to 0.
13581 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13582 (MASK_STRING): Delete.
13583 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13585 (load_multiple): Delete.
13592 (store_multiple): Delete.
13599 (movmemsi_8reg): Delete.
13600 (corresponding unnamed define_insn): Delete.
13601 (movmemsi_6reg): Delete.
13602 (corresponding unnamed define_insn): Delete.
13603 (movmemsi_4reg): Delete.
13604 (corresponding unnamed define_insn): Delete.
13605 (movmemsi_2reg): Delete.
13606 (corresponding unnamed define_insn): Delete.
13607 (movmemsi_1reg): Delete.
13608 (corresponding unnamed define_insn): Delete.
13609 * config/rs6000/rs6000.opt (mno-string): New.
13610 (mstring): Replace by deprecation warning stub.
13611 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13613 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13615 * regrename.c (regrename_do_replace): If replacing the same
13616 reg multiple times, try to reuse last created gen_raw_REG.
13619 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13620 main to workaround a bug in GDB.
13622 2018-01-12 Tom de Vries <tom@codesourcery.com>
13625 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13627 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13629 PR rtl-optimization/80481
13630 * ira-color.c (get_cap_member): New function.
13631 (allocnos_conflict_by_live_ranges_p): Use it.
13632 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13633 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13635 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13638 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13639 (*saddl_se_1): Ditto.
13640 (*ssubsi_1): Ditto.
13641 (*ssubl_se_1): Ditto.
13643 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13645 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13646 rather than wi::to_widest for DR_INITs.
13647 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13648 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13649 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13651 (vect_analyze_group_access_1): Note that here.
13653 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13655 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13656 polynomial type sizes.
13658 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13660 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13661 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13662 (gimple_add_tmp_var): Likewise.
13664 2018-01-12 Martin Liska <mliska@suse.cz>
13666 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13667 (gimple_alloc_sizes): Likewise.
13668 (dump_gimple_statistics): Use PRIu64 in printf format.
13669 * gimple.h: Change uint64_t to int.
13671 2018-01-12 Martin Liska <mliska@suse.cz>
13673 * tree-core.h: Use uint64_t instead of int.
13674 * tree.c (tree_node_counts): Likewise.
13675 (tree_node_sizes): Likewise.
13676 (dump_tree_statistics): Use PRIu64 in printf format.
13678 2018-01-12 Martin Liska <mliska@suse.cz>
13680 * Makefile.in: As qsort_chk is implemented in vec.c, add
13681 vec.o to linkage of gencfn-macros.
13682 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13683 passing the info to record_node_allocation_statistics.
13684 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13686 * ggc-common.c (struct ggc_usage): Add operator== and use
13687 it in operator< and compare function.
13688 * mem-stats.h (struct mem_usage): Likewise.
13689 * vec.c (struct vec_usage): Remove operator< and compare
13690 function. Can be simply inherited.
13692 2018-01-12 Martin Jambor <mjambor@suse.cz>
13695 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13696 * tree-ssa-math-opts.c: Include domwalk.h.
13697 (convert_mult_to_fma_1): New function.
13698 (fma_transformation_info): New type.
13699 (fma_deferring_state): Likewise.
13700 (cancel_fma_deferring): New function.
13701 (result_of_phi): Likewise.
13702 (last_fma_candidate_feeds_initial_phi): Likewise.
13703 (convert_mult_to_fma): Added deferring logic, split actual
13704 transformation to convert_mult_to_fma_1.
13705 (math_opts_dom_walker): New type.
13706 (math_opts_dom_walker::after_dom_children): New method, body moved
13707 here from pass_optimize_widening_mul::execute, added deferring logic
13709 (pass_optimize_widening_mul::execute): Moved most of code to
13710 math_opts_dom_walker::after_dom_children.
13711 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13712 * config/i386/i386.c (ix86_option_override_internal): Added
13713 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13715 2018-01-12 Richard Biener <rguenther@suse.de>
13718 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13719 inline instance vars.
13721 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13724 * config/rx/rx.c (rx_is_restricted_memory_address):
13725 Handle SUBREG case.
13727 2018-01-12 Richard Biener <rguenther@suse.de>
13729 PR tree-optimization/80846
13730 * target.def (split_reduction): New target hook.
13731 * targhooks.c (default_split_reduction): New function.
13732 * targhooks.h (default_split_reduction): Declare.
13733 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13734 target requests first reduce vectors by combining low and high
13736 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13737 (get_vectype_for_scalar_type_and_size): Export.
13738 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13739 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13740 * doc/tm.texi: Regenerate.
13741 * config/i386/i386.c (ix86_split_reduction): Implement
13742 TARGET_VECTORIZE_SPLIT_REDUCTION.
13744 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13747 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13748 in PIC mode except for TARGET_VXWORKS_RTP.
13749 * config/sparc/sparc.c: Include cfgrtl.h.
13750 (TARGET_INIT_PIC_REG): Define.
13751 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13752 (sparc_pic_register_p): New predicate.
13753 (sparc_legitimate_address_p): Use it.
13754 (sparc_legitimize_pic_address): Likewise.
13755 (sparc_delegitimize_address): Likewise.
13756 (sparc_mode_dependent_address_p): Likewise.
13757 (gen_load_pcrel_sym): Remove 4th parameter.
13758 (load_got_register): Adjust call to above. Remove obsolete stuff.
13759 (sparc_expand_prologue): Do not call load_got_register here.
13760 (sparc_flat_expand_prologue): Likewise.
13761 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13762 (sparc_use_pseudo_pic_reg): New function.
13763 (sparc_init_pic_reg): Likewise.
13764 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13765 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13767 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13769 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13770 Add item for branch_cost.
13772 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13774 PR rtl-optimization/83565
13775 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13776 not extend the result to a larger mode for rotate operations.
13777 (num_sign_bit_copies1): Likewise.
13779 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13782 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13784 Use values-Xc.o for -pedantic.
13785 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13787 2018-01-12 Martin Liska <mliska@suse.cz>
13790 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13792 (possible_polymorphic_call_targets): Use it.
13793 (ipa_devirt): Likewise.
13795 2018-01-12 Martin Liska <mliska@suse.cz>
13797 * profile-count.h (enum profile_quality): Use 0 as invalid
13798 enum value of profile_quality.
13800 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13802 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13803 -mext-string options.
13805 2018-01-12 Richard Biener <rguenther@suse.de>
13807 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13808 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13809 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13811 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13813 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13815 * configure.ac (--with-long-double-format): Add support for the
13816 configuration option to change the default long double format on
13818 * config.gcc (powerpc*-linux*-*): Likewise.
13819 * configure: Regenerate.
13820 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13821 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13822 used without modification.
13824 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13826 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13827 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13828 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13829 MISC_BUILTIN_SPEC_BARRIER.
13830 (rs6000_init_builtins): Likewise.
13831 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13833 (speculation_barrier): New define_insn.
13834 * doc/extend.texi: Document __builtin_speculation_barrier.
13836 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13839 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13840 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13841 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13843 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13844 integral modes instead of "ss" and "sd".
13845 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13846 vectors with 32-bit and 64-bit elements.
13847 (vecdupssescalarmodesuffix): New mode attribute.
13848 (vec_dup<mode>): Use it.
13850 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13853 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13854 frame if argument is passed on stack.
13856 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13859 * ree.c (combine_reaching_defs): Optimize also
13860 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13861 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13863 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13865 PR middle-end/83189
13866 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13868 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13870 PR middle-end/83718
13871 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13872 after they are computed.
13874 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13876 PR tree-optimization/83695
13877 * gimple-loop-linterchange.cc
13878 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13879 reset cached scev information after interchange.
13880 (pass_linterchange::execute): Remove call to scev_reset_htab.
13882 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13884 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13885 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13886 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13887 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13888 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13889 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13890 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13891 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13892 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13893 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13894 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13895 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13896 (V_lane_reg): Likewise.
13897 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13899 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13900 (vfmal_lane_low<mode>_intrinsic,
13901 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13902 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13903 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13904 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13905 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13906 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13908 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13910 * config/arm/arm-cpus.in (fp16fml): New feature.
13911 (ALL_SIMD): Add fp16fml.
13912 (armv8.2-a): Add fp16fml as an option.
13913 (armv8.3-a): Likewise.
13914 (armv8.4-a): Add fp16fml as part of fp16.
13915 * config/arm/arm.h (TARGET_FP16FML): Define.
13916 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13918 * config/arm/arm-modes.def (V2HF): Define.
13919 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13920 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13921 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
13922 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
13923 vfmsl_low, vfmsl_high): New set of builtins.
13924 * config/arm/iterators.md (PLUSMINUS): New code iterator.
13925 (vfml_op): New code attribute.
13926 (VFMLHALVES): New int iterator.
13927 (VFML, VFMLSEL): New mode attributes.
13928 (V_reg): Define mapping for V2HF.
13929 (V_hi, V_lo): New mode attributes.
13930 (VF_constraint): Likewise.
13931 (vfml_half, vfml_half_selector): New int attributes.
13932 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
13934 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
13935 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
13937 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
13938 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
13939 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
13940 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
13942 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
13943 Document new effective target and option set.
13945 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13947 * config/arm/arm-cpus.in (armv8_4): New feature.
13948 (ARMv8_4a): New fgroup.
13949 (armv8.4-a): New arch.
13950 * config/arm/arm-tables.opt: Regenerate.
13951 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
13952 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
13953 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
13954 Add matching rules for -march=armv8.4-a and extensions.
13955 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
13957 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
13960 * config/rx/rx.md (BW): New mode attribute.
13961 (sync_lock_test_and_setsi): Add mode suffix to insn output.
13963 2018-01-11 Richard Biener <rguenther@suse.de>
13965 PR tree-optimization/83435
13966 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
13967 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
13968 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
13970 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13971 Alan Hayward <alan.hayward@arm.com>
13972 David Sherwood <david.sherwood@arm.com>
13974 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
13976 (aarch64_classify_address): Initialize it. Track polynomial offsets.
13977 (aarch64_print_address_internal): Use it to check for a zero offset.
13979 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
13980 Alan Hayward <alan.hayward@arm.com>
13981 David Sherwood <david.sherwood@arm.com>
13983 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
13984 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
13985 Return a poly_int64 rather than a HOST_WIDE_INT.
13986 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
13987 rather than a HOST_WIDE_INT.
13988 * config/aarch64/aarch64.h (aarch64_frame): Protect with
13989 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
13990 hard_fp_offset, frame_size, initial_adjust, callee_offset and
13991 final_offset from HOST_WIDE_INT to poly_int64.
13992 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
13993 to_constant when getting the number of units in an Advanced SIMD
13995 (aarch64_builtin_vectorized_function): Check for a constant number
13997 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
13999 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14000 attribute instead of GET_MODE_NUNITS.
14001 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14002 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14003 GET_MODE_SIZE for fixed-size registers.
14004 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14005 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14006 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14007 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14008 (aarch64_print_operand, aarch64_print_address_internal)
14009 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14010 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14011 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14012 Handle polynomial GET_MODE_SIZE.
14013 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14014 wider than SImode without modification.
14015 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14016 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14017 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14018 passing and returning SVE modes.
14019 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14020 rather than GEN_INT.
14021 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14022 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14023 (aarch64_allocate_and_probe_stack_space): Likewise.
14024 (aarch64_layout_frame): Cope with polynomial offsets.
14025 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14026 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14027 polynomial offsets.
14028 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14029 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14030 poly_int64 rather than a HOST_WIDE_INT.
14031 (aarch64_get_separate_components, aarch64_process_components)
14032 (aarch64_expand_prologue, aarch64_expand_epilogue)
14033 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14034 (aarch64_anchor_offset): New function, split out from...
14035 (aarch64_legitimize_address): ...here.
14036 (aarch64_builtin_vectorization_cost): Handle polynomial
14037 TYPE_VECTOR_SUBPARTS.
14038 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14040 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14041 number of elements from the PARALLEL rather than the mode.
14042 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14043 rather than GET_MODE_BITSIZE.
14044 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14045 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14046 (aarch64_expand_vec_perm_const_1): Handle polynomial
14047 d->perm.length () and d->perm elements.
14048 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14049 Apply to_constant to d->perm elements.
14050 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14051 polynomial CONST_VECTOR_NUNITS.
14052 (aarch64_move_pointer): Take amount as a poly_int64 rather
14054 (aarch64_progress_pointer): Avoid temporary variable.
14055 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14056 the mode attribute instead of GET_MODE.
14058 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14059 Alan Hayward <alan.hayward@arm.com>
14060 David Sherwood <david.sherwood@arm.com>
14062 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14063 x exists before using it.
14064 (aarch64_add_constant_internal): Rename to...
14065 (aarch64_add_offset_1): ...this. Replace regnum with separate
14066 src and dest rtxes. Handle the case in which they're different,
14067 including when the offset is zero. Replace scratchreg with an rtx.
14068 Use 2 additions if there is no spare register into which we can
14069 move a 16-bit constant.
14070 (aarch64_add_constant): Delete.
14071 (aarch64_add_offset): Replace reg with separate src and dest
14072 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14073 Use aarch64_add_offset_1.
14074 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14075 an rtx rather than an int. Take the delta as a poly_int64
14076 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14077 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14078 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14079 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14080 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14081 and aarch64_add_sp.
14082 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14083 aarch64_add_constant.
14085 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14087 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14088 Use scalar_float_mode.
14090 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14092 * config/aarch64/aarch64-simd.md
14093 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14094 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14095 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14096 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14097 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14098 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14099 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14100 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14101 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14102 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14104 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14107 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14108 targ_options->x_arm_arch_string is non NULL.
14110 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14112 * config/aarch64/aarch64.h
14113 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14115 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14118 * expmed.c (emit_store_flag_force): Swap if const op0
14119 and change VOIDmode to mode of op0.
14121 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14123 PR rtl-optimization/83761
14124 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14125 than bytes to mode_for_size.
14127 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14129 PR middle-end/83189
14130 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14131 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14134 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14136 PR middle-end/83575
14137 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14138 when in layout mode.
14139 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14140 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14143 2018-01-10 Michael Collison <michael.collison@arm.com>
14145 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14146 * config/aarch64/aarch64-option-extension.def: Add
14147 AARCH64_OPT_EXTENSION of 'fp16fml'.
14148 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14149 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14150 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14151 * config/aarch64/constraints.md (Ui7): New constraint.
14152 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14153 (VFMLA_SEL_W): Ditto.
14156 (VFMLA16_LOW): New int iterator.
14157 (VFMLA16_HIGH): Ditto.
14158 (UNSPEC_FMLAL): New unspec.
14159 (UNSPEC_FMLSL): Ditto.
14160 (UNSPEC_FMLAL2): Ditto.
14161 (UNSPEC_FMLSL2): Ditto.
14162 (f16mac): New code attribute.
14163 * config/aarch64/aarch64-simd-builtins.def
14164 (aarch64_fmlal_lowv2sf): Ditto.
14165 (aarch64_fmlsl_lowv2sf): Ditto.
14166 (aarch64_fmlalq_lowv4sf): Ditto.
14167 (aarch64_fmlslq_lowv4sf): Ditto.
14168 (aarch64_fmlal_highv2sf): Ditto.
14169 (aarch64_fmlsl_highv2sf): Ditto.
14170 (aarch64_fmlalq_highv4sf): Ditto.
14171 (aarch64_fmlslq_highv4sf): Ditto.
14172 (aarch64_fmlal_lane_lowv2sf): Ditto.
14173 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14174 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14175 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14176 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14177 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14178 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14179 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14180 (aarch64_fmlal_lane_highv2sf): Ditto.
14181 (aarch64_fmlsl_lane_highv2sf): Ditto.
14182 (aarch64_fmlal_laneq_highv2sf): Ditto.
14183 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14184 (aarch64_fmlalq_lane_highv4sf): Ditto.
14185 (aarch64_fmlsl_lane_highv4sf): Ditto.
14186 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14187 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14188 * config/aarch64/aarch64-simd.md:
14189 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14190 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14191 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14192 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14193 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14194 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14195 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14196 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14197 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14198 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14199 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14200 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14201 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14202 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14203 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14204 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14205 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14206 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14207 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14208 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14209 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14210 (vfmlsl_low_u32): Ditto.
14211 (vfmlalq_low_u32): Ditto.
14212 (vfmlslq_low_u32): Ditto.
14213 (vfmlal_high_u32): Ditto.
14214 (vfmlsl_high_u32): Ditto.
14215 (vfmlalq_high_u32): Ditto.
14216 (vfmlslq_high_u32): Ditto.
14217 (vfmlal_lane_low_u32): Ditto.
14218 (vfmlsl_lane_low_u32): Ditto.
14219 (vfmlal_laneq_low_u32): Ditto.
14220 (vfmlsl_laneq_low_u32): Ditto.
14221 (vfmlalq_lane_low_u32): Ditto.
14222 (vfmlslq_lane_low_u32): Ditto.
14223 (vfmlalq_laneq_low_u32): Ditto.
14224 (vfmlslq_laneq_low_u32): Ditto.
14225 (vfmlal_lane_high_u32): Ditto.
14226 (vfmlsl_lane_high_u32): Ditto.
14227 (vfmlal_laneq_high_u32): Ditto.
14228 (vfmlsl_laneq_high_u32): Ditto.
14229 (vfmlalq_lane_high_u32): Ditto.
14230 (vfmlslq_lane_high_u32): Ditto.
14231 (vfmlalq_laneq_high_u32): Ditto.
14232 (vfmlslq_laneq_high_u32): Ditto.
14233 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14234 (AARCH64_FL_FOR_ARCH8_4): New.
14235 (AARCH64_ISA_F16FML): New ISA flag.
14236 (TARGET_F16FML): New feature flag for fp16fml.
14237 (doc/invoke.texi): Document new fp16fml option.
14239 2018-01-10 Michael Collison <michael.collison@arm.com>
14241 * config/aarch64/aarch64-builtins.c:
14242 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14243 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14244 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14245 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14246 (AARCH64_ISA_SHA3): New ISA flag.
14247 (TARGET_SHA3): New feature flag for sha3.
14248 * config/aarch64/iterators.md (sha512_op): New int attribute.
14249 (CRYPTO_SHA512): New int iterator.
14250 (UNSPEC_SHA512H): New unspec.
14251 (UNSPEC_SHA512H2): Ditto.
14252 (UNSPEC_SHA512SU0): Ditto.
14253 (UNSPEC_SHA512SU1): Ditto.
14254 * config/aarch64/aarch64-simd-builtins.def
14255 (aarch64_crypto_sha512hqv2di): New builtin.
14256 (aarch64_crypto_sha512h2qv2di): Ditto.
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/aarch64-simd.md:
14264 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14265 (aarch64_crypto_sha512su0qv2di): Ditto.
14266 (aarch64_crypto_sha512su1qv2di): Ditto.
14267 (aarch64_eor3qv8hi): Ditto.
14268 (aarch64_rax1qv2di): Ditto.
14269 (aarch64_xarqv2di): Ditto.
14270 (aarch64_bcaxqv8hi): Ditto.
14271 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14272 (vsha512h2q_u64): Ditto.
14273 (vsha512su0q_u64): Ditto.
14274 (vsha512su1q_u64): Ditto.
14275 (veor3q_u16): Ditto.
14276 (vrax1q_u64): Ditto.
14277 (vxarq_u64): Ditto.
14278 (vbcaxq_u16): Ditto.
14279 * config/arm/types.md (crypto_sha512): New type attribute.
14280 (crypto_sha3): Ditto.
14281 (doc/invoke.texi): Document new sha3 option.
14283 2018-01-10 Michael Collison <michael.collison@arm.com>
14285 * config/aarch64/aarch64-builtins.c:
14286 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14287 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14288 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14289 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14290 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14291 (AARCH64_ISA_SM4): New ISA flag.
14292 (TARGET_SM4): New feature flag for sm4.
14293 * config/aarch64/aarch64-simd-builtins.def
14294 (aarch64_sm3ss1qv4si): Ditto.
14295 (aarch64_sm3tt1aq4si): Ditto.
14296 (aarch64_sm3tt1bq4si): Ditto.
14297 (aarch64_sm3tt2aq4si): Ditto.
14298 (aarch64_sm3tt2bq4si): Ditto.
14299 (aarch64_sm3partw1qv4si): Ditto.
14300 (aarch64_sm3partw2qv4si): Ditto.
14301 (aarch64_sm4eqv4si): Ditto.
14302 (aarch64_sm4ekeyqv4si): Ditto.
14303 * config/aarch64/aarch64-simd.md:
14304 (aarch64_sm3ss1qv4si): Ditto.
14305 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14306 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14307 (aarch64_sm4eqv4si): Ditto.
14308 (aarch64_sm4ekeyqv4si): Ditto.
14309 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14310 (sm3part_op): Ditto.
14311 (CRYPTO_SM3TT): Ditto.
14312 (CRYPTO_SM3PART): Ditto.
14313 (UNSPEC_SM3SS1): New unspec.
14314 (UNSPEC_SM3TT1A): Ditto.
14315 (UNSPEC_SM3TT1B): Ditto.
14316 (UNSPEC_SM3TT2A): Ditto.
14317 (UNSPEC_SM3TT2B): Ditto.
14318 (UNSPEC_SM3PARTW1): Ditto.
14319 (UNSPEC_SM3PARTW2): Ditto.
14320 (UNSPEC_SM4E): Ditto.
14321 (UNSPEC_SM4EKEY): Ditto.
14322 * config/aarch64/constraints.md (Ui2): New constraint.
14323 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14324 * config/arm/types.md (crypto_sm3): New type attribute.
14325 (crypto_sm4): Ditto.
14326 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14327 (vsm3tt1aq_u32): Ditto.
14328 (vsm3tt1bq_u32): Ditto.
14329 (vsm3tt2aq_u32): Ditto.
14330 (vsm3tt2bq_u32): Ditto.
14331 (vsm3partw1q_u32): Ditto.
14332 (vsm3partw2q_u32): Ditto.
14333 (vsm4eq_u32): Ditto.
14334 (vsm4ekeyq_u32): Ditto.
14335 (doc/invoke.texi): Document new sm4 option.
14337 2018-01-10 Michael Collison <michael.collison@arm.com>
14339 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14340 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14341 (AARCH64_FL_FOR_ARCH8_4): New.
14342 (AARCH64_FL_V8_4): New flag.
14343 (doc/invoke.texi): Document new armv8.4-a option.
14345 2018-01-10 Michael Collison <michael.collison@arm.com>
14347 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14348 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14349 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14350 * config/aarch64/aarch64-option-extension.def: Add
14351 AARCH64_OPT_EXTENSION of 'sha2'.
14352 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14353 (crypto): Disable sha2 and aes if crypto disabled.
14354 (crypto): Enable aes and sha2 if enabled.
14355 (simd): Disable sha2 and aes if simd disabled.
14356 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14358 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14359 (TARGET_SHA2): New feature flag for sha2.
14360 (TARGET_AES): New feature flag for aes.
14361 * config/aarch64/aarch64-simd.md:
14362 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14363 conditional on TARGET_AES.
14364 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14365 (aarch64_crypto_sha1hsi): Make pattern conditional
14367 (aarch64_crypto_sha1hv4si): Ditto.
14368 (aarch64_be_crypto_sha1hv4si): Ditto.
14369 (aarch64_crypto_sha1su1v4si): Ditto.
14370 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14371 (aarch64_crypto_sha1su0v4si): Ditto.
14372 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14373 (aarch64_crypto_sha256su0v4si): Ditto.
14374 (aarch64_crypto_sha256su1v4si): Ditto.
14375 (doc/invoke.texi): Document new aes and sha2 options.
14377 2018-01-10 Martin Sebor <msebor@redhat.com>
14379 PR tree-optimization/83781
14380 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14383 2018-01-11 Martin Sebor <msebor@gmail.com>
14384 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14386 PR tree-optimization/83501
14387 PR tree-optimization/81703
14389 * tree-ssa-strlen.c (get_string_cst): Rename...
14390 (get_string_len): ...to this. Handle global constants.
14391 (handle_char_store): Adjust.
14393 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14394 Jim Wilson <jimw@sifive.com>
14396 * config/riscv/riscv-protos.h (riscv_output_return): New.
14397 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14398 (riscv_attribute_table, riscv_output_return),
14399 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14400 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14401 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14403 (riscv_expand_prologue): Add early return for naked function.
14404 (riscv_expand_epilogue): Likewise.
14405 (riscv_function_ok_for_sibcall): Return false for naked function.
14406 (riscv_set_current_function): New.
14407 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14408 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14409 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14410 * doc/extend.texi (RISC-V Function Attributes): New.
14412 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14414 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14415 check for 128-bit long double before checking TCmode.
14416 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14417 128-bit long doubles before checking TFmode or TCmode.
14418 (FLOAT128_IBM_P): Likewise.
14420 2018-01-10 Martin Sebor <msebor@redhat.com>
14422 PR tree-optimization/83671
14423 * builtins.c (c_strlen): Unconditionally return zero for the empty
14425 Use -Warray-bounds for warnings.
14426 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14427 for non-constant array indices with COMPONENT_REF, arrays of
14428 arrays, and pointers to arrays.
14429 (gimple_fold_builtin_strlen): Determine and set length range for
14430 non-constant character arrays.
14432 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14434 PR middle-end/81897
14435 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14438 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14440 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14442 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14445 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14446 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14447 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14448 indexed_or_indirect_operand predicate.
14449 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14450 (*vsx_le_perm_load_v8hi): Likewise.
14451 (*vsx_le_perm_load_v16qi): Likewise.
14452 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14453 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14454 (*vsx_le_perm_store_v8hi): Likewise.
14455 (*vsx_le_perm_store_v16qi): Likewise.
14456 (eight unnamed splitters): Likewise.
14458 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14460 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14461 * config/rs6000/emmintrin.h: Likewise.
14462 * config/rs6000/mmintrin.h: Likewise.
14463 * config/rs6000/xmmintrin.h: Likewise.
14465 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14468 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14470 * tree.c (tree_nop_conversion): Return true for location wrapper
14472 (maybe_wrap_with_location): New function.
14473 (selftest::check_strip_nops): New function.
14474 (selftest::test_location_wrappers): New function.
14475 (selftest::tree_c_tests): Call it.
14476 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14477 (maybe_wrap_with_location): New decl.
14478 (EXPR_LOCATION_WRAPPER_P): New macro.
14479 (location_wrapper_p): New inline function.
14480 (tree_strip_any_location_wrapper): New inline function.
14482 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14485 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14486 stack_realign_offset for the largest alignment of stack slot
14488 (ix86_find_max_used_stack_alignment): New function.
14489 (ix86_finalize_stack_frame_flags): Use it. Set
14490 max_used_stack_alignment if we don't realign stack.
14491 * config/i386/i386.h (machine_function): Add
14492 max_used_stack_alignment.
14494 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14496 * config/arm/arm.opt (-mbranch-cost): New option.
14497 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14500 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14503 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14504 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14506 2018-01-10 Richard Biener <rguenther@suse.de>
14509 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14510 early out so it also covers the case where we have a non-NULL
14513 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14515 PR tree-optimization/83753
14516 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14517 for non-strided grouped accesses if the number of elements is 1.
14519 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14522 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14523 * i386.h (TARGET_USE_GATHER): Define.
14524 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14526 2018-01-10 Martin Liska <mliska@suse.cz>
14529 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14530 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14532 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14533 CLEANUP_NO_PARTITIONING is not set.
14535 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14537 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14538 for vectors, as a partial revert of r254296.
14539 * rtl.h (const_vec_p): Delete.
14540 (const_vec_duplicate_p): Don't test for vector CONSTs.
14541 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14542 * expmed.c (make_tree): Likewise.
14545 * common.md (E, F): Use CONSTANT_P instead of checking for
14547 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14548 checking for CONST_VECTOR.
14550 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14552 PR middle-end/83575
14553 * predict.c (force_edge_cold): Handle in more sane way edges
14554 with no prediction.
14556 2018-01-09 Carl Love <cel@us.ibm.com>
14558 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14560 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14561 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14562 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14563 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14564 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14565 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14566 * config/rs6000/rs6000-protos.h: Add extern defition for
14567 rs6000_generate_float2_double_code.
14568 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14570 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14571 (float2_v2df): Add define_expand.
14573 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14576 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14577 op_mode in the force_to_mode call.
14579 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14581 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14582 instead of checking each element individually.
14583 (aarch64_evpc_uzp): Likewise.
14584 (aarch64_evpc_zip): Likewise.
14585 (aarch64_evpc_ext): Likewise.
14586 (aarch64_evpc_rev): Likewise.
14587 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14588 instead of checking each element individually. Return true without
14590 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14591 whether all selected elements come from the same input, instead of
14592 checking each element individually. Remove calls to gen_rtx_REG,
14593 start_sequence and end_sequence and instead assert that no rtl is
14596 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14598 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14599 order of HIGH and CONST checks.
14601 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14603 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14604 if the destination isn't an SSA_NAME.
14606 2018-01-09 Richard Biener <rguenther@suse.de>
14608 PR tree-optimization/83668
14609 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14611 (canonicalize_loop_form): ... here, renamed from ...
14612 (canonicalize_loop_closed_ssa_form): ... this and amended to
14613 swap successor edges for loop exit blocks to make us use
14614 the RPO order we need for initial schedule generation.
14616 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14618 PR tree-optimization/64811
14619 * match.pd: When optimizing comparisons with Inf, avoid
14620 introducing or losing exceptions from comparisons with NaN.
14622 2018-01-09 Martin Liska <mliska@suse.cz>
14625 * asan.c (shadow_mem_size): Add gcc_assert.
14627 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14629 Don't save registers in main().
14632 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14633 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14634 * config/avr/avr.c (avr_set_current_function): Don't error if
14635 naked, OS_task or OS_main are specified at the same time.
14636 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14638 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14640 * common/config/avr/avr-common.c (avr_option_optimization_table):
14641 Switch on -mmain-is-OS_task for optimizing compilations.
14643 2018-01-09 Richard Biener <rguenther@suse.de>
14645 PR tree-optimization/83572
14646 * graphite.c: Include cfganal.h.
14647 (graphite_transform_loops): Connect infinite loops to exit
14648 and remove fake edges at the end.
14650 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14652 * ipa-inline.c (edge_badness): Revert accidental checkin.
14654 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14657 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14658 symbols; not inline clones.
14660 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14663 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14664 hard registers. Formatting fixes.
14666 PR preprocessor/83722
14667 * gcc.c (try_generate_repro): Pass
14668 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14669 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14672 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14673 Kito Cheng <kito.cheng@gmail.com>
14675 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14676 (riscv_leaf_function_p): Delete.
14677 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14679 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14681 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14683 (do_ifelse): New function.
14684 (do_isel): New function.
14685 (do_sub3): New function.
14686 (do_add3): New function.
14687 (do_load_mask_compare): New function.
14688 (do_overlap_load_compare): New function.
14689 (expand_compare_loop): New function.
14690 (expand_block_compare): Call expand_compare_loop() when appropriate.
14691 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14692 option description.
14693 (-mblock-compare-inline-loop-limit): New option.
14695 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14698 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14699 Reverse order of second and third operands in first alternative.
14700 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14701 of first and second elements in UNSPEC_VPERMR vector.
14702 (altivec_expand_vec_perm_le): Likewise.
14704 2018-01-08 Jeff Law <law@redhat.com>
14706 PR rtl-optimizatin/81308
14707 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14708 (process_switch): If group_case_labels makes a change, then set
14710 (pass_convert_switch::execute): If a switch is converted, then
14711 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14713 PR rtl-optimization/81308
14714 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14717 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14719 PR target/83663 - Revert r255946
14720 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14721 generation for cases where splatting a value is not useful.
14722 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14723 across a vec_duplicate and a paradoxical subreg forming a vector
14724 mode to a vec_concat.
14726 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14728 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14729 -march=armv8.3-a variants.
14730 * config/arm/t-multilib: Likewise.
14731 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14733 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14735 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14737 (cceq_ior_compare_complement): Give it a name so I can use it, and
14738 change boolean_or_operator predicate to boolean_operator so it can
14739 be used to generate a crand.
14740 (eqne): New code iterator.
14741 (bd/bd_neg): New code_attrs.
14742 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14743 a single define_insn.
14744 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14745 decrement (bdnzt/bdnzf/bdzt/bdzf).
14746 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14747 with the new names of the branch decrement patterns, and added the
14748 names of the branch decrement conditional patterns.
14750 2018-01-08 Richard Biener <rguenther@suse.de>
14752 PR tree-optimization/83563
14753 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14756 2018-01-08 Richard Biener <rguenther@suse.de>
14758 PR middle-end/83713
14759 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14761 2018-01-08 Richard Biener <rguenther@suse.de>
14763 PR tree-optimization/83685
14764 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14765 references to abnormals.
14767 2018-01-08 Richard Biener <rguenther@suse.de>
14770 * dwarf2out.c (output_indirect_strings): Handle empty
14771 skeleton_debug_str_hash.
14772 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14774 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14776 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14777 (emit_store_direct): Likewise.
14778 (arc_trampoline_adjust_address): Likewise.
14779 (arc_asm_trampoline_template): New function.
14780 (arc_initialize_trampoline): Use asm_trampoline_template.
14781 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14782 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14783 * config/arc/arc.md (flush_icache): Delete pattern.
14785 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14787 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14788 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14791 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14794 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14795 by not USED_FOR_TARGET.
14796 (make_pass_resolve_sw_modes): Likewise.
14798 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14800 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14803 2018-01-08 Richard Biener <rguenther@suse.de>
14805 PR middle-end/83580
14806 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14808 2018-01-08 Richard Biener <rguenther@suse.de>
14810 PR middle-end/83517
14811 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14813 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14815 PR middle-end/81897
14816 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14817 basic blocks with a small number of successors.
14818 (convert_control_dep_chain_into_preds): Improve handling of
14820 (dump_predicates): Split apart into...
14821 (dump_pred_chain): ...here...
14822 (dump_pred_info): ...and here.
14823 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14824 (can_chain_union_be_invalidated_p): Improve check for invalidation
14826 (uninit_uses_cannot_happen): Avoid unnecessary if
14827 convert_control_dep_chain_into_preds yielded nothing.
14829 2018-01-06 Martin Sebor <msebor@redhat.com>
14831 PR tree-optimization/83640
14832 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14833 subtracting negative offset from size.
14834 (builtin_access::overlap): Adjust offset bounds of the access to fall
14835 within the size of the object if possible.
14837 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14839 PR rtl-optimization/83699
14840 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14841 extract_bit_field_as_subreg to cases in which the extracted
14842 value is also a vector.
14844 * lra-constraints.c (process_alt_operands): Test for the equivalence
14845 substitutions when detecting a possible reload cycle.
14847 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14850 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14851 by default if flag_selective_schedling{,2}. Formatting fixes.
14853 PR rtl-optimization/83682
14854 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14855 if it has non-VECTOR_MODE element mode.
14856 (vec_duplicate_p): Likewise.
14858 PR middle-end/83694
14859 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14860 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14862 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14865 * config/i386/i386-builtin.def
14866 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14867 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14868 Require also OPTION_MASK_ISA_AVX512F in addition to
14869 OPTION_MASK_ISA_GFNI.
14870 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14871 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14872 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14873 to OPTION_MASK_ISA_GFNI.
14874 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14875 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14876 OPTION_MASK_ISA_AVX512BW.
14877 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14878 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14879 addition to OPTION_MASK_ISA_GFNI.
14880 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14881 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14882 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14883 to OPTION_MASK_ISA_GFNI.
14884 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14885 a requirement for all ISAs rather than any of them with a few
14887 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14889 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14890 bitmasks to be enabled with 3 exceptions, instead of requiring any
14891 enabled ISA with lots of exceptions.
14892 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14893 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14894 Change avx512bw in isa attribute to avx512f.
14895 * config/i386/sgxintrin.h: Add license boilerplate.
14896 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14897 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14898 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14899 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14901 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14902 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14903 temporarily sse2 rather than sse if not enabled already.
14906 * config/i386/sse.md (VI248_VLBW): Rename to ...
14907 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14908 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14909 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14910 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14911 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14912 mode iterator instead of VI248_VLBW.
14914 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14916 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14917 (record_modified): Skip clobbers; add debug output.
14918 (param_change_prob): Use sreal frequencies.
14920 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14922 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
14923 punt for user-aligned variables.
14925 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
14927 * tree-chrec.c (chrec_contains_symbols): Return true for
14930 2018-01-05 Sudakshina Das <sudi.das@arm.com>
14933 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
14934 of (x|y) == x for BICS pattern.
14936 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14938 PR tree-optimization/83605
14939 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
14940 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
14943 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
14945 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
14946 * config/epiphany/rtems.h: New file.
14948 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14949 Uros Bizjak <ubizjak@gmail.com>
14952 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
14953 QIreg_operand instead of register_operand predicate.
14954 * config/i386/i386.c (ix86_rop_should_change_byte_p,
14955 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
14956 comments instead of -fmitigate[-_]rop.
14958 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14961 * cgraphunit.c (symbol_table::compile): Switch to text_section
14962 before calling assembly_start debug hook.
14963 * run-rtl-passes.c (run_rtl_passes): Likewise.
14966 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14968 * tree-vrp.c (extract_range_from_binary_expr_1): Check
14969 range_int_cst_p rather than !symbolic_range_p before calling
14970 extract_range_from_multiplicative_op_1.
14972 2018-01-04 Jeff Law <law@redhat.com>
14974 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
14975 redundant test in assertion.
14977 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14979 * doc/rtl.texi: Document machine_mode wrapper classes.
14981 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14983 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
14984 using tree_to_uhwi.
14986 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
14988 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
14989 the VEC_PERM_EXPR fold to fail.
14991 2018-01-04 Jakub Jelinek <jakub@redhat.com>
14994 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
14995 to switched_sections.
14997 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15000 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15001 test for d.testing.
15003 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15006 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15007 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15009 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15012 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15013 is BLKmode and bitpos not zero or mode change is needed.
15015 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15018 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15021 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15024 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15025 instead of MULT rtx. Update all corresponding splitters.
15026 (*saddl_se): Ditto.
15027 (*ssub<modesuffix>): Ditto.
15028 (*ssubl_se): Ditto.
15029 (*cmp_sadd_di): Update split patterns.
15030 (*cmp_sadd_si): Ditto.
15031 (*cmp_sadd_sidi): Ditto.
15032 (*cmp_ssub_di): Ditto.
15033 (*cmp_ssub_si): Ditto.
15034 (*cmp_ssub_sidi): Ditto.
15035 * config/alpha/predicates.md (const23_operand): New predicate.
15036 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15037 Look for ASHIFT, not MULT inner operand.
15038 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15040 2018-01-04 Martin Liska <mliska@suse.cz>
15042 PR gcov-profile/83669
15043 * gcov.c (output_intermediate_file): Add version to intermediate
15045 * doc/gcov.texi: Document new field 'version' in intermediate
15046 file format. Fix location of '-k' option of gcov command.
15048 2018-01-04 Martin Liska <mliska@suse.cz>
15051 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15053 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15055 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15057 2018-01-03 Martin Sebor <msebor@redhat.com>
15059 PR tree-optimization/83655
15060 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15061 checking calls with invalid arguments.
15063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15065 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15066 (vectorizable_mask_load_store): Delete.
15067 (vectorizable_call): Return false for masked loads and stores.
15068 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15069 instead of gimple_assign_rhs1.
15070 (vectorizable_load): Handle IFN_MASK_LOAD.
15071 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15075 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15077 (vectorizable_mask_load_store): ...here.
15078 (vectorizable_load): ...and here.
15080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15082 * tree-vect-stmts.c (vect_build_all_ones_mask)
15083 (vect_build_zero_merge_argument): New functions, split out from...
15084 (vectorizable_load): ...here.
15086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15088 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15090 (vectorizable_mask_load_store): ...here.
15091 (vectorizable_store): ...and here.
15093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15095 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15097 (vectorizable_mask_load_store): ...here.
15099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15101 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15102 (vect_model_store_cost): Take a vec_load_store_type instead of a
15104 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15105 (vect_model_store_cost): Take a vec_load_store_type instead of a
15107 (vectorizable_mask_load_store): Update accordingly.
15108 (vectorizable_store): Likewise.
15109 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15113 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15114 IFN_MASK_LOAD calls here rather than...
15115 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
15122 fall back to extract_bit_field_as_subreg if vec_extract
15125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15126 Alan Hayward <alan.hayward@arm.com>
15127 David Sherwood <david.sherwood@arm.com>
15129 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15130 they are variable or constant sized.
15131 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15132 slots for constant-sized data.
15134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15135 Alan Hayward <alan.hayward@arm.com>
15136 David Sherwood <david.sherwood@arm.com>
15138 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15139 handling COND_EXPRs with boolean comparisons, try to find a better
15140 basis for the mask type than the boolean itself.
15142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15144 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15145 is calculated and how it can be overridden.
15146 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15147 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15149 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15153 Alan Hayward <alan.hayward@arm.com>
15154 David Sherwood <david.sherwood@arm.com>
15156 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15157 Remove the mode argument.
15158 (aarch64_simd_valid_immediate): Remove the mode and inverse
15160 * config/aarch64/iterators.md (bitsize): New iterator.
15161 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15162 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15163 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15164 aarch64_simd_valid_immediate.
15165 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15166 (aarch64_reg_or_bic_imm): Likewise.
15167 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15168 with an insn_type enum and msl with a modifier_type enum.
15169 Replace element_width with a scalar_mode. Change the shift
15170 to unsigned int. Add constructors for scalar_float_mode and
15171 scalar_int_mode elements.
15172 (aarch64_vect_float_const_representable_p): Delete.
15173 (aarch64_can_const_movi_rtx_p)
15174 (aarch64_simd_scalar_immediate_valid_for_move)
15175 (aarch64_simd_make_constant): Update call to
15176 aarch64_simd_valid_immediate.
15177 (aarch64_advsimd_valid_immediate_hs): New function.
15178 (aarch64_advsimd_valid_immediate): Likewise.
15179 (aarch64_simd_valid_immediate): Remove mode and inverse
15180 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15181 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15182 and aarch64_float_const_representable_p on the result.
15183 (aarch64_output_simd_mov_immediate): Remove mode argument.
15184 Update call to aarch64_simd_valid_immediate and use of
15185 simd_immediate_info.
15186 (aarch64_output_scalar_simd_mov_immediate): Update call
15189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15190 Alan Hayward <alan.hayward@arm.com>
15191 David Sherwood <david.sherwood@arm.com>
15193 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15194 (mode_nunits): Likewise CONST_MODE_NUNITS.
15195 * machmode.def (ADJUST_NUNITS): Document.
15196 * genmodes.c (mode_data::need_nunits_adj): New field.
15197 (blank_mode): Update accordingly.
15198 (adj_nunits): New variable.
15199 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15201 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15202 listed in adj_nunits.
15203 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15204 listed in adj_nunits. Don't emit case statements for such modes.
15205 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15206 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15207 nothing if adj_nunits is nonnull.
15208 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15209 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15210 (emit_mode_fbit): Update use of print_maybe_const_decl.
15211 (emit_move_size): Likewise. Treat the array as non-const
15213 (emit_mode_adjustments): Handle adj_nunits.
15215 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15217 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15218 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15219 (VECTOR_MODES): Use it.
15220 (make_vector_modes): Take the prefix as an argument.
15222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15223 Alan Hayward <alan.hayward@arm.com>
15224 David Sherwood <david.sherwood@arm.com>
15226 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15227 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15228 for MODE_VECTOR_BOOL.
15229 * machmode.def (VECTOR_BOOL_MODE): Document.
15230 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15231 (make_vector_bool_mode): New function.
15232 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15234 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15235 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15237 * stor-layout.c (int_mode_for_mode): Likewise.
15238 * tree.c (build_vector_type_for_mode): Likewise.
15239 * varasm.c (output_constant_pool_2): Likewise.
15240 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15241 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15242 for MODE_VECTOR_BOOL.
15243 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15244 of mode class checks.
15245 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15246 instead of a list of mode class checks.
15247 (expand_vector_scalar_condition): Likewise.
15248 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15251 Alan Hayward <alan.hayward@arm.com>
15252 David Sherwood <david.sherwood@arm.com>
15254 * machmode.h (mode_size): Change from unsigned short to
15256 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15257 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15258 or if measurement_type is not polynomial.
15259 (fixed_size_mode::includes_p): Check for constant-sized modes.
15260 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15261 return a poly_uint16 rather than an unsigned short.
15262 (emit_mode_size): Change the type of mode_size from unsigned short
15263 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15264 (emit_mode_adjustments): Cope with polynomial vector sizes.
15265 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15267 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15269 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15270 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15271 * caller-save.c (setup_save_areas): Likewise.
15272 (replace_reg_with_saved_mem): Likewise.
15273 * calls.c (emit_library_call_value_1): Likewise.
15274 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15275 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15276 (gen_lowpart_for_combine): Likewise.
15277 * convert.c (convert_to_integer_1): Likewise.
15278 * cse.c (equiv_constant, cse_insn): Likewise.
15279 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15280 (cselib_subst_to_values): Likewise.
15281 * dce.c (word_dce_process_block): Likewise.
15282 * df-problems.c (df_word_lr_mark_ref): Likewise.
15283 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15284 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15285 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15286 (rtl_for_decl_location): Likewise.
15287 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15288 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15289 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15290 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15291 (expand_expr_real_1): Likewise.
15292 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15293 (pad_below): Likewise.
15294 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15295 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15296 * ira.c (get_subreg_tracking_sizes): Likewise.
15297 * ira-build.c (ira_create_allocno_objects): Likewise.
15298 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15299 (ira_sort_regnos_for_alter_reg): Likewise.
15300 * ira-costs.c (record_operand_costs): Likewise.
15301 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15302 (resolve_simple_move): Likewise.
15303 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15304 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15305 (lra_constraints): Likewise.
15306 (CONST_POOL_OK_P): Reject variable-sized modes.
15307 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15308 (add_pseudo_to_slot, lra_spill): Likewise.
15309 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15310 * optabs-query.c (get_best_extraction_insn): Likewise.
15311 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15312 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15313 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15314 * recog.c (offsettable_address_addr_space_p): Likewise.
15315 * regcprop.c (maybe_mode_change): Likewise.
15316 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15317 * regrename.c (build_def_use): Likewise.
15318 * regstat.c (dump_reg_info): Likewise.
15319 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15320 (find_reloads, find_reloads_subreg_address): Likewise.
15321 * reload1.c (eliminate_regs_1): Likewise.
15322 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15323 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15324 (simplify_binary_operation_1, simplify_subreg): Likewise.
15325 * targhooks.c (default_function_arg_padding): Likewise.
15326 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15327 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15328 (verify_gimple_assign_ternary): Likewise.
15329 * tree-inline.c (estimate_move_cost): Likewise.
15330 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15331 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15332 (get_address_cost_ainc): Likewise.
15333 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15334 (vect_supportable_dr_alignment): Likewise.
15335 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15336 (vectorizable_reduction): Likewise.
15337 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15338 (vectorizable_operation, vectorizable_load): Likewise.
15339 * tree.c (build_same_sized_truth_vector_type): Likewise.
15340 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15341 * var-tracking.c (emit_note_insn_var_location): Likewise.
15342 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15343 (ADDR_VEC_ALIGN): Likewise.
15345 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15346 Alan Hayward <alan.hayward@arm.com>
15347 David Sherwood <david.sherwood@arm.com>
15349 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15351 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15352 or if measurement_type is polynomial.
15353 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15354 * combine.c (make_extraction): Likewise.
15355 * dse.c (find_shift_sequence): Likewise.
15356 * dwarf2out.c (mem_loc_descriptor): Likewise.
15357 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15358 (extract_bit_field, extract_low_bits): Likewise.
15359 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15360 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15361 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15362 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15363 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15364 * reload.c (find_reloads): Likewise.
15365 * reload1.c (alter_reg): Likewise.
15366 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15367 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15368 * tree-if-conv.c (predicate_mem_writes): Likewise.
15369 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15370 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15371 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15372 * valtrack.c (dead_debug_insert_temp): Likewise.
15373 * varasm.c (mergeable_constant_section): Likewise.
15374 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15377 Alan Hayward <alan.hayward@arm.com>
15378 David Sherwood <david.sherwood@arm.com>
15380 * expr.c (expand_assignment): Cope with polynomial mode sizes
15381 when assigning to a CONCAT.
15383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15384 Alan Hayward <alan.hayward@arm.com>
15385 David Sherwood <david.sherwood@arm.com>
15387 * machmode.h (mode_precision): Change from unsigned short to
15389 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15391 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15392 or if measurement_type is not polynomial.
15393 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15394 in which the mode is already known to be a scalar_int_mode.
15395 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15396 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15398 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15399 for GET_MODE_PRECISION.
15400 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15401 for GET_MODE_PRECISION.
15402 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15404 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15405 (expand_field_assignment, make_extraction): Likewise.
15406 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15407 (get_last_value): Likewise.
15408 * convert.c (convert_to_integer_1): Likewise.
15409 * cse.c (cse_insn): Likewise.
15410 * expr.c (expand_expr_real_1): Likewise.
15411 * lra-constraints.c (simplify_operand_subreg): Likewise.
15412 * optabs-query.c (can_atomic_load_p): Likewise.
15413 * optabs.c (expand_atomic_load): Likewise.
15414 (expand_atomic_store): Likewise.
15415 * ree.c (combine_reaching_defs): Likewise.
15416 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15417 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15418 * tree.h (type_has_mode_precision_p): Likewise.
15419 * ubsan.c (instrument_si_overflow): Likewise.
15421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15422 Alan Hayward <alan.hayward@arm.com>
15423 David Sherwood <david.sherwood@arm.com>
15425 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15426 polynomial numbers of units.
15427 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15428 (valid_vector_subparts_p): New function.
15429 (build_vector_type): Remove temporary shim and take the number
15430 of units as a poly_uint64 rather than an int.
15431 (build_opaque_vector_type): Take the number of units as a
15432 poly_uint64 rather than an int.
15433 * tree.c (build_vector_from_ctor): Handle polynomial
15434 TYPE_VECTOR_SUBPARTS.
15435 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15436 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15437 (build_vector_from_val): If the number of units is variable,
15438 use build_vec_duplicate_cst for constant operands and
15439 VEC_DUPLICATE_EXPR otherwise.
15440 (make_vector_type): Remove temporary is_constant ().
15441 (build_vector_type, build_opaque_vector_type): Take the number of
15442 units as a poly_uint64 rather than an int.
15443 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15445 * cfgexpand.c (expand_debug_expr): Likewise.
15446 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15447 (store_constructor, expand_expr_real_1): Likewise.
15448 (const_scalar_mask_from_tree): Likewise.
15449 * fold-const-call.c (fold_const_reduction): Likewise.
15450 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15451 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15452 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15453 (fold_relational_const): Likewise.
15454 (native_interpret_vector): Likewise. Change the size from an
15455 int to an unsigned int.
15456 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15457 TYPE_VECTOR_SUBPARTS.
15458 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15459 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15460 duplicating a non-constant operand into a variable-length vector.
15461 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15462 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15463 * ipa-icf.c (sem_variable::equals): Likewise.
15464 * match.pd: Likewise.
15465 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15466 * print-tree.c (print_node): Likewise.
15467 * stor-layout.c (layout_type): Likewise.
15468 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15469 * tree-cfg.c (verify_gimple_comparison): Likewise.
15470 (verify_gimple_assign_binary): Likewise.
15471 (verify_gimple_assign_ternary): Likewise.
15472 (verify_gimple_assign_single): Likewise.
15473 * tree-pretty-print.c (dump_generic_node): Likewise.
15474 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15475 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15476 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15477 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15478 (vect_shift_permute_load_chain): Likewise.
15479 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15480 (expand_vector_condition, optimize_vector_constructor): Likewise.
15481 (lower_vec_perm, get_compute_type): Likewise.
15482 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15483 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15484 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15485 (vect_recog_mask_conversion_pattern): Likewise.
15486 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15487 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15488 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15489 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15490 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15491 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15492 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15493 (supportable_widening_operation): Likewise.
15494 (supportable_narrowing_operation): Likewise.
15495 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15497 * varasm.c (output_constant): Likewise.
15499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15500 Alan Hayward <alan.hayward@arm.com>
15501 David Sherwood <david.sherwood@arm.com>
15503 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15504 so that both the length == 3 and length != 3 cases set up their
15505 own permute vectors. Add comments explaining why we know the
15506 number of elements is constant.
15507 (vect_permute_load_chain): Likewise.
15509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15510 Alan Hayward <alan.hayward@arm.com>
15511 David Sherwood <david.sherwood@arm.com>
15513 * machmode.h (mode_nunits): Change from unsigned char to
15515 (ONLY_FIXED_SIZE_MODES): New macro.
15516 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15517 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15518 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15520 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15521 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15522 or if measurement_type is not polynomial.
15523 * genmodes.c (ZERO_COEFFS): New macro.
15524 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15526 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15527 Use ZERO_COEFFS when emitting initializers.
15528 * data-streamer.h (bp_pack_poly_value): New function.
15529 (bp_unpack_poly_value): Likewise.
15530 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15531 for GET_MODE_NUNITS.
15532 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15533 for GET_MODE_NUNITS.
15534 * tree.c (make_vector_type): Remove temporary shim and make
15535 the real function take the number of units as a poly_uint64
15536 rather than an int.
15537 (build_vector_type_for_mode): Handle polynomial nunits.
15538 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15539 * emit-rtl.c (const_vec_series_p_1): Likewise.
15540 (gen_rtx_CONST_VECTOR): Likewise.
15541 * fold-const.c (test_vec_duplicate_folding): Likewise.
15542 * genrecog.c (validate_pattern): Likewise.
15543 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15544 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15545 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15546 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15547 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15548 * rtlanal.c (subreg_get_info): Likewise.
15549 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15550 (vect_grouped_load_supported): Likewise.
15551 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15552 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15553 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15554 (simplify_const_unary_operation, simplify_binary_operation_1)
15555 (simplify_const_binary_operation, simplify_ternary_operation)
15556 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15557 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15558 instead of CONST_VECTOR_NUNITS.
15559 * varasm.c (output_constant_pool_2): Likewise.
15560 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15561 explicit-encoded elements in the XVEC for variable-length vectors.
15563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15565 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15568 Alan Hayward <alan.hayward@arm.com>
15569 David Sherwood <david.sherwood@arm.com>
15571 * coretypes.h (fixed_size_mode): Declare.
15572 (fixed_size_mode_pod): New typedef.
15573 * builtins.h (target_builtins::x_apply_args_mode)
15574 (target_builtins::x_apply_result_mode): Change type to
15575 fixed_size_mode_pod.
15576 * builtins.c (apply_args_size, apply_result_size, result_vector)
15577 (expand_builtin_apply_args_1, expand_builtin_apply)
15578 (expand_builtin_return): Update accordingly.
15580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15582 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15583 * cselib.c (cselib_hash_rtx): Likewise.
15584 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15585 CONST_VECTOR encoding.
15587 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15588 Jeff Law <law@redhat.com>
15591 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15592 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15593 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15594 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15597 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15598 explicitly probe *sp in a noreturn function if there were any callee
15599 register saves or frame pointer is needed.
15601 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15604 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15605 BLKmode for ternary, binary or unary expressions.
15608 * var-tracking.c (delete_vta_debug_insn): New inline function.
15609 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15610 insns from get_insns () to NULL instead of each bb separately.
15611 Use delete_vta_debug_insn. No longer static.
15612 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15613 delete_vta_debug_insns callers.
15614 * rtl.h (delete_vta_debug_insns): Declare.
15615 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15616 instead of variable_tracking_main.
15618 2018-01-03 Martin Sebor <msebor@redhat.com>
15620 PR tree-optimization/83603
15621 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15622 arguments past the endof the argument list in functions declared
15623 without a prototype.
15624 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15625 Avoid checking when arguments are null.
15627 2018-01-03 Martin Sebor <msebor@redhat.com>
15630 * doc/extend.texi (attribute const): Fix a typo.
15631 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15632 issuing -Wsuggest-attribute for void functions.
15634 2018-01-03 Martin Sebor <msebor@redhat.com>
15636 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15637 offset_int::from instead of wide_int::to_shwi.
15638 (maybe_diag_overlap): Remove assertion.
15639 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15640 * gimple-ssa-sprintf.c (format_directive): Same.
15641 (parse_directive): Same.
15642 (sprintf_dom_walker::compute_format_length): Same.
15643 (try_substitute_return_value): Same.
15645 2018-01-03 Jeff Law <law@redhat.com>
15647 PR middle-end/83654
15648 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15649 non-constant residual for zero at runtime and avoid probing in
15650 that case. Reorganize code for trailing problem to mirror handling
15653 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15655 PR tree-optimization/83501
15656 * tree-ssa-strlen.c (get_string_cst): New.
15657 (handle_char_store): Call get_string_cst.
15659 2018-01-03 Martin Liska <mliska@suse.cz>
15661 PR tree-optimization/83593
15662 * tree-ssa-strlen.c: Include tree-cfg.h.
15663 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15664 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15665 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15667 (strlen_dom_walker::before_dom_children): Call
15668 gimple_purge_dead_eh_edges. Dump tranformation with details
15670 (strlen_dom_walker::before_dom_children): Update call by adding
15671 new argument cleanup_eh.
15672 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15674 2018-01-03 Martin Liska <mliska@suse.cz>
15677 * cif-code.def (VARIADIC_THUNK): New enum value.
15678 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15681 2018-01-03 Jan Beulich <jbeulich@suse.com>
15683 * sse.md (mov<mode>_internal): Tighten condition for when to use
15684 vmovdqu<ssescalarsize> for TI and OI modes.
15686 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15688 Update copyright years.
15690 2018-01-03 Martin Liska <mliska@suse.cz>
15693 * ipa-visibility.c (function_and_variable_visibility): Skip
15694 functions with noipa attribure.
15696 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15698 * gcc.c (process_command): Update copyright notice dates.
15699 * gcov-dump.c (print_version): Ditto.
15700 * gcov.c (print_version): Ditto.
15701 * gcov-tool.c (print_version): Ditto.
15702 * gengtype.c (create_file): Ditto.
15703 * doc/cpp.texi: Bump @copying's copyright year.
15704 * doc/cppinternals.texi: Ditto.
15705 * doc/gcc.texi: Ditto.
15706 * doc/gccint.texi: Ditto.
15707 * doc/gcov.texi: Ditto.
15708 * doc/install.texi: Ditto.
15709 * doc/invoke.texi: Ditto.
15711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15713 * vector-builder.h (vector_builder::m_full_nelts): Change from
15714 unsigned int to poly_uint64.
15715 (vector_builder::full_nelts): Update prototype accordingly.
15716 (vector_builder::new_vector): Likewise.
15717 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15718 (vector_builder::operator ==): Likewise.
15719 (vector_builder::finalize): Likewise.
15720 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15721 Take the number of elements as a poly_uint64 rather than an
15723 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15724 from unsigned int to poly_uint64.
15725 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15726 (vec_perm_indices::new_vector): Likewise.
15727 (vec_perm_indices::length): Likewise.
15728 (vec_perm_indices::nelts_per_input): Likewise.
15729 (vec_perm_indices::input_nelts): Likewise.
15730 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15731 number of elements per input as a poly_uint64 rather than an
15732 unsigned int. Use the original encoding for variable-length
15733 vectors, rather than clamping each individual element.
15734 For the second and subsequent elements in each pattern,
15735 clamp the step and base before clamping their sum.
15736 (vec_perm_indices::series_p): Handle polynomial element counts.
15737 (vec_perm_indices::all_in_range_p): Likewise.
15738 (vec_perm_indices_to_tree): Likewise.
15739 (vec_perm_indices_to_rtx): Likewise.
15740 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15741 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15742 (tree_vector_builder::new_binary_operation): Handle polynomial
15743 element counts. Return false if we need to know the number
15744 of elements at compile time.
15745 * fold-const.c (fold_vec_perm): Punt if the number of elements
15746 isn't known at compile time.
15748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15750 * vec-perm-indices.h (vec_perm_builder): Change element type
15751 from HOST_WIDE_INT to poly_int64.
15752 (vec_perm_indices::element_type): Update accordingly.
15753 (vec_perm_indices::clamp): Handle polynomial element_types.
15754 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15755 (vec_perm_indices::all_in_range_p): Likewise.
15756 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15758 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15759 polynomial vec_perm_indices element types.
15760 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15761 * fold-const.c (fold_vec_perm): Likewise.
15762 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15763 * tree-vect-generic.c (lower_vec_perm): Likewise.
15764 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15765 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15766 element type to HOST_WIDE_INT.
15768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15769 Alan Hayward <alan.hayward@arm.com>
15770 David Sherwood <david.sherwood@arm.com>
15772 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15773 rather than an int. Use plus_constant.
15774 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15775 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15778 Alan Hayward <alan.hayward@arm.com>
15779 David Sherwood <david.sherwood@arm.com>
15781 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15782 a HOST_WIDE_INT to a poly_int64.
15784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15785 Alan Hayward <alan.hayward@arm.com>
15786 David Sherwood <david.sherwood@arm.com>
15788 * calls.c (load_register_parameters): Cope with polynomial
15789 mode sizes. Require a constant size for BLKmode parameters
15790 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15791 forces a parameter to be padded at the lsb end in order to
15792 fill a complete number of words, require the parameter size
15793 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
15800 from unsigned int to poly_uint64_pod.
15801 (alter_reg): Treat mode sizes as polynomial.
15803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15804 Alan Hayward <alan.hayward@arm.com>
15805 David Sherwood <david.sherwood@arm.com>
15807 * reload.c (complex_word_subreg_p): New function.
15808 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
15815 operands whose sizes aren't ordered.
15816 (match_reload): Refer to this check here.
15818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15819 Alan Hayward <alan.hayward@arm.com>
15820 David Sherwood <david.sherwood@arm.com>
15822 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15823 that the mode size is in the set {1, 2, 4, 8, 16}.
15825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15826 Alan Hayward <alan.hayward@arm.com>
15827 David Sherwood <david.sherwood@arm.com>
15829 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15830 Use plus_constant instead of gen_rtx_PLUS.
15832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15833 Alan Hayward <alan.hayward@arm.com>
15834 David Sherwood <david.sherwood@arm.com>
15836 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15837 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15838 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15839 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15840 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15841 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15842 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15843 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15844 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15845 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15847 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15848 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15849 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15850 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15851 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15852 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15853 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15854 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15855 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15856 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15858 * expr.c (emit_move_resolve_push): Treat the input and result
15859 of PUSH_ROUNDING as a poly_int64.
15860 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15861 (emit_push_insn): Likewise.
15862 * lra-eliminations.c (mark_not_eliminable): Likewise.
15863 * recog.c (push_operand): Likewise.
15864 * reload1.c (elimination_effects): Likewise.
15865 * rtlanal.c (nonzero_bits1): Likewise.
15866 * calls.c (store_one_arg): Likewise. Require the padding to be
15867 known at compile time.
15869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15870 Alan Hayward <alan.hayward@arm.com>
15871 David Sherwood <david.sherwood@arm.com>
15873 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15874 Use plus_constant instead of gen_rtx_PLUS.
15876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15877 Alan Hayward <alan.hayward@arm.com>
15878 David Sherwood <david.sherwood@arm.com>
15880 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15881 rather than an int.
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_1): Use tree_to_poly_uint64
15888 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15889 via stack temporaries. Treat the mode size as polynomial too.
15891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15892 Alan Hayward <alan.hayward@arm.com>
15893 David Sherwood <david.sherwood@arm.com>
15895 * expr.c (expand_expr_real_2): When handling conversions involving
15896 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15897 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15898 as a poly_uint64 too.
15900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15901 Alan Hayward <alan.hayward@arm.com>
15902 David Sherwood <david.sherwood@arm.com>
15904 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15907 Alan Hayward <alan.hayward@arm.com>
15908 David Sherwood <david.sherwood@arm.com>
15910 * combine.c (can_change_dest_mode): Handle polynomial
15911 REGMODE_NATURAL_SIZE.
15912 * expmed.c (store_bit_field_1): Likewise.
15913 * expr.c (store_constructor): Likewise.
15914 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15915 and polynomial REGMODE_NATURAL_SIZE.
15916 (gen_lowpart_common): Likewise.
15917 * reginfo.c (record_subregs_of_mode): Likewise.
15918 * rtlanal.c (read_modify_subreg_p): Likewise.
15920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15921 Alan Hayward <alan.hayward@arm.com>
15922 David Sherwood <david.sherwood@arm.com>
15924 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
15925 numbers of elements.
15927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15928 Alan Hayward <alan.hayward@arm.com>
15929 David Sherwood <david.sherwood@arm.com>
15931 * match.pd: Cope with polynomial numbers of vector elements.
15933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15934 Alan Hayward <alan.hayward@arm.com>
15935 David Sherwood <david.sherwood@arm.com>
15937 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
15938 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
15945 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
15946 (ipa_simd_modify_function_body): Likewise.
15948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15949 Alan Hayward <alan.hayward@arm.com>
15950 David Sherwood <david.sherwood@arm.com>
15952 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
15953 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
15954 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
15955 (expand_vector_condition, vector_element): Likewise.
15956 (subparts_gt): New function.
15957 (get_compute_type): Use subparts_gt.
15958 (count_type_subparts): Delete.
15959 (expand_vector_operations_1): Use subparts_gt instead of
15960 count_type_subparts.
15962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15963 Alan Hayward <alan.hayward@arm.com>
15964 David Sherwood <david.sherwood@arm.com>
15966 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
15967 (vect_compile_time_alias): ...this new function. Do the calculation
15968 on poly_ints rather than trees.
15969 (vect_prune_runtime_alias_test_list): Update call accordingly.
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_build_slp_tree_1): Handle polynomial
15977 (vect_schedule_slp_instance): Likewise.
15979 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15980 Alan Hayward <alan.hayward@arm.com>
15981 David Sherwood <david.sherwood@arm.com>
15983 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
15984 constant and extern definitions for variable-length vectors.
15985 (vect_get_constant_vectors): Note that the number of units
15986 is known to be constant.
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 (vectorizable_conversion): Treat the number
15993 of units as polynomial. Choose between WIDE and NARROW based
15996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15997 Alan Hayward <alan.hayward@arm.com>
15998 David Sherwood <david.sherwood@arm.com>
16000 * tree-vect-stmts.c (simd_clone_subparts): New function.
16001 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 (vectorizable_call): Treat the number of
16008 vectors as polynomial. Use build_index_vector for
16009 IFN_GOMP_SIMD_LANE.
16011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16012 Alan Hayward <alan.hayward@arm.com>
16013 David Sherwood <david.sherwood@arm.com>
16015 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16016 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16017 for variable-length vectors.
16018 (vectorizable_mask_load_store): Treat the number of units as
16019 polynomial, asserting that it is constant if the condition has
16020 already been enforced.
16021 (vectorizable_store, vectorizable_load): Likewise.
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_live_operation): Treat the number
16028 of units as polynomial. Punt if we can't tell at compile time
16029 which vector contains the final result.
16031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16032 Alan Hayward <alan.hayward@arm.com>
16033 David Sherwood <david.sherwood@arm.com>
16035 * tree-vect-loop.c (vectorizable_induction): Treat the number
16036 of units as polynomial. Punt on SLP inductions. Use an integer
16037 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16038 cast of such a series for variable-length floating-point
16041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16042 Alan Hayward <alan.hayward@arm.com>
16043 David Sherwood <david.sherwood@arm.com>
16045 * tree.h (build_index_vector): Declare.
16046 * tree.c (build_index_vector): New function.
16047 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16048 of units as polynomial, forcibly converting it to a constant if
16049 vectorizable_reduction has already enforced the condition.
16050 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16051 to create a {1,2,3,...} vector.
16052 (vectorizable_reduction): Treat the number of units as polynomial.
16053 Choose vectype_in based on the largest scalar element size rather
16054 than the smallest number of units. Enforce the restrictions
16057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16058 Alan Hayward <alan.hayward@arm.com>
16059 David Sherwood <david.sherwood@arm.com>
16061 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16062 number of units as polynomial.
16064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16065 Alan Hayward <alan.hayward@arm.com>
16066 David Sherwood <david.sherwood@arm.com>
16068 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16069 * target.def (autovectorize_vector_sizes): Return the vector sizes
16070 by pointer, using vector_sizes rather than a bitmask.
16071 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16072 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16073 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16075 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16076 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16077 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16078 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16079 * omp-general.c (omp_max_vf): Likewise.
16080 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16081 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16082 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16083 * tree-vect-slp.c (vect_slp_bb): Likewise.
16084 * doc/tm.texi: Regenerate.
16085 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16087 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16088 the vector size as a poly_uint64 rather than an unsigned int.
16089 (current_vector_size): Change from an unsigned int to a poly_uint64.
16090 (get_vectype_for_scalar_type): Update accordingly.
16091 * tree.h (build_truth_vector_type): Take the size and number of
16092 units as a poly_uint64 rather than an unsigned int.
16093 (build_vector_type): Add a temporary overload that takes
16094 the number of units as a poly_uint64 rather than an unsigned int.
16095 * tree.c (make_vector_type): Likewise.
16096 (build_truth_vector_type): Take the number of units as a poly_uint64
16097 rather than an unsigned int.
16099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16100 Alan Hayward <alan.hayward@arm.com>
16101 David Sherwood <david.sherwood@arm.com>
16103 * target.def (get_mask_mode): Take the number of units and length
16104 as poly_uint64s rather than unsigned ints.
16105 * targhooks.h (default_get_mask_mode): Update accordingly.
16106 * targhooks.c (default_get_mask_mode): Likewise.
16107 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16108 * doc/tm.texi: Regenerate.
16110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16111 Alan Hayward <alan.hayward@arm.com>
16112 David Sherwood <david.sherwood@arm.com>
16114 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16115 * omp-general.c (omp_max_vf): Likewise.
16116 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16117 (expand_omp_simd): Handle polynomial safelen.
16118 * omp-low.c (omplow_simd_context): Add a default constructor.
16119 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16120 (lower_rec_simd_input_clauses): Update accordingly.
16121 (lower_rec_input_clauses): Likewise.
16123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16124 Alan Hayward <alan.hayward@arm.com>
16125 David Sherwood <david.sherwood@arm.com>
16127 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16128 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16129 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16130 (vect_analyze_slp_cost): Likewise.
16131 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16132 (vect_model_load_cost): Likewise.
16134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16135 Alan Hayward <alan.hayward@arm.com>
16136 David Sherwood <david.sherwood@arm.com>
16138 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16139 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16140 from an unsigned int * to a poly_uint64_pod *.
16141 (calculate_unrolling_factor): New function.
16142 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16145 Alan Hayward <alan.hayward@arm.com>
16146 David Sherwood <david.sherwood@arm.com>
16148 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16149 from an unsigned int to a poly_uint64.
16150 (_loop_vec_info::slp_unrolling_factor): Likewise.
16151 (_loop_vec_info::vectorization_factor): Change from an int
16153 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16154 (vect_get_num_vectors): New function.
16155 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16156 (vect_get_num_copies): Use vect_get_num_vectors.
16157 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16158 to an unsigned int *.
16159 (vect_analyze_data_refs): Change min_vf from an int * to a
16161 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16162 than an unsigned HOST_WIDE_INT.
16163 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16164 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16165 to an unsigned int *.
16166 (vect_analyze_data_ref_dependences): Likewise.
16167 (vect_compute_data_ref_alignment): Handle polynomial vf.
16168 (vect_enhance_data_refs_alignment): Likewise.
16169 (vect_prune_runtime_alias_test_list): Likewise.
16170 (vect_shift_permute_load_chain): Likewise.
16171 (vect_supportable_dr_alignment): Likewise.
16172 (dependence_distance_ge_vf): Take the vectorization factor as a
16173 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16174 (vect_analyze_data_refs): Change min_vf from an int * to a
16176 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16177 vfm1 as a poly_uint64 rather than an int. Make the same change
16178 for the returned bound_scalar.
16179 (vect_gen_vector_loop_niters): Handle polynomial vf.
16180 (vect_do_peeling): Likewise. Update call to
16181 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16182 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16184 * tree-vect-loop.c (vect_determine_vectorization_factor)
16185 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16186 (vect_get_known_peeling_cost): Likewise.
16187 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16188 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16189 (vect_transform_loop): Likewise. Use the lowest possible VF when
16190 updating the upper bounds of the loop.
16191 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16192 rather than an int.
16193 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16194 polynomial unroll factors.
16195 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16196 (vect_make_slp_decision): Likewise.
16197 (vect_supported_load_permutation_p): Likewise, and polynomial
16199 (vect_analyze_slp_cost): Handle polynomial vf.
16200 (vect_slp_analyze_node_operations): Likewise.
16201 (vect_slp_analyze_bb_1): Likewise.
16202 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16203 than an unsigned HOST_WIDE_INT.
16204 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16205 (vectorizable_load): Handle polynomial vf.
16206 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16208 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16211 Alan Hayward <alan.hayward@arm.com>
16212 David Sherwood <david.sherwood@arm.com>
16214 * match.pd: Handle bit operations involving three constants
16215 and try to fold one pair.
16217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16219 * tree-vect-loop-manip.c: Include gimple-fold.h.
16220 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16221 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16222 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16223 Add a path that uses a step of VF instead of 1, but disable it
16225 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16226 and niters_no_overflow parameters. Update calls to
16227 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16228 Create a new SSA name if the latter choses to use a ste other
16229 than zero, and return it via niters_vector_mult_vf_var.
16230 * tree-vect-loop.c (vect_transform_loop): Update calls to
16231 vect_do_peeling, vect_gen_vector_loop_niters and
16232 slpeel_make_loop_iterate_ntimes.
16233 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16234 (vect_gen_vector_loop_niters): Update declarations after above changes.
16236 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16238 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16239 128-bit round to integer instructions.
16240 (ceil<mode>2): Likewise.
16241 (btrunc<mode>2): Likewise.
16242 (round<mode>2): Likewise.
16244 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16246 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16247 unaligned VSX load/store on P8/P9.
16248 (expand_block_clear): Allow the use of unaligned VSX
16249 load/store on P8/P9.
16251 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16253 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16255 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16256 swap associated with both a load and a store.
16258 2018-01-02 Andrew Waterman <andrew@sifive.com>
16260 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16261 * config/riscv/riscv.md (clear_cache): Use it.
16263 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16265 * web.c: Remove out-of-date comment.
16267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16269 * expr.c (fixup_args_size_notes): Check that any existing
16270 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16271 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16272 (emit_single_push_insn): ...here.
16274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16276 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16277 (const_vector_encoded_nelts): New function.
16278 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16279 (const_vector_int_elt, const_vector_elt): Declare.
16280 * emit-rtl.c (const_vector_int_elt_1): New function.
16281 (const_vector_elt): Likewise.
16282 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16283 of CONST_VECTOR_ELT.
16285 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16287 * expr.c: Include rtx-vector-builder.h.
16288 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16289 directly on the tree encoding.
16290 (const_vector_from_tree): Likewise.
16291 * optabs.c: Include rtx-vector-builder.h.
16292 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16293 sequence of "u" values.
16294 * vec-perm-indices.c: Include rtx-vector-builder.h.
16295 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16296 directly on the vec_perm_indices encoding.
16298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16300 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16301 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16302 * rtx-vector-builder.h: New file.
16303 * rtx-vector-builder.c: Likewise.
16304 * rtl.h (rtx_def::u2): Add a const_vector field.
16305 (CONST_VECTOR_NPATTERNS): New macro.
16306 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16307 (CONST_VECTOR_DUPLICATE_P): Likewise.
16308 (CONST_VECTOR_STEPPED_P): Likewise.
16309 (CONST_VECTOR_ENCODED_ELT): Likewise.
16310 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16311 (unwrap_const_vec_duplicate): Likewise.
16312 (const_vec_series_p): Check for a non-duplicated vector encoding.
16313 Say that the function only returns true for integer vectors.
16314 * emit-rtl.c: Include rtx-vector-builder.h.
16315 (gen_const_vec_duplicate_1): Delete.
16316 (gen_const_vector): Call gen_const_vec_duplicate instead of
16317 gen_const_vec_duplicate_1.
16318 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16319 (gen_const_vec_duplicate): Use rtx_vector_builder.
16320 (gen_const_vec_series): Likewise.
16321 (gen_rtx_CONST_VECTOR): Likewise.
16322 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16323 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16324 Build a new vector rather than modifying a CONST_VECTOR in-place.
16325 (handle_special_swappables): Update call accordingly.
16326 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16327 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16328 Build a new vector rather than modifying a CONST_VECTOR in-place.
16329 (handle_special_swappables): Update call accordingly.
16331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16333 * simplify-rtx.c (simplify_const_binary_operation): Use
16334 CONST_VECTOR_ELT instead of XVECEXP.
16336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16338 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16339 the selector elements to be different from the data elements
16340 if the selector is a VECTOR_CST.
16341 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16342 ssizetype for the selector.
16344 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16346 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16347 before testing each element individually.
16348 * tree-vect-generic.c (lower_vec_perm): Likewise.
16350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16352 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16353 * selftest-run-tests.c (selftest::run_tests): Call it.
16354 * vector-builder.h (vector_builder::operator ==): New function.
16355 (vector_builder::operator !=): Likewise.
16356 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16357 (vec_perm_indices::all_from_input_p): New function.
16358 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16359 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16360 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16361 instead of reading the VECTOR_CST directly. Detect whether both
16362 vector inputs are the same before constructing the vec_perm_indices,
16363 and update the number of inputs argument accordingly. Use the
16364 utility functions added above. Only construct sel2 if we need to.
16366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16368 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16369 the broadcast of the low byte.
16370 (expand_mult_highpart): Use an explicit encoding for the permutes.
16371 * optabs-query.c (can_mult_highpart_p): Likewise.
16372 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16373 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16374 (vectorizable_bswap): Likewise.
16375 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16376 explicit encoding for the power-of-2 permutes.
16377 (vect_permute_store_chain): Likewise.
16378 (vect_grouped_load_supported): Likewise.
16379 (vect_permute_load_chain): Likewise.
16381 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16383 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16384 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16385 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16386 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16387 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16388 (vect_gen_perm_mask_any): Likewise.
16390 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16392 * int-vector-builder.h: New file.
16393 * vec-perm-indices.h: Include int-vector-builder.h.
16394 (vec_perm_indices): Redefine as an int_vector_builder.
16395 (auto_vec_perm_indices): Delete.
16396 (vec_perm_builder): Redefine as a stand-alone class.
16397 (vec_perm_indices::vec_perm_indices): New function.
16398 (vec_perm_indices::clamp): Likewise.
16399 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16400 (vec_perm_indices::new_vector): New function.
16401 (vec_perm_indices::new_expanded_vector): Update for new
16402 vec_perm_indices class.
16403 (vec_perm_indices::rotate_inputs): New function.
16404 (vec_perm_indices::all_in_range_p): Operate directly on the
16405 encoded form, without computing elided elements.
16406 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16407 encoding. Update for new vec_perm_indices class.
16408 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16409 the given vec_perm_builder.
16410 (expand_vec_perm_var): Update vec_perm_builder constructor.
16411 (expand_mult_highpart): Use vec_perm_builder instead of
16412 auto_vec_perm_indices.
16413 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16414 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16415 or double series encoding as appropriate.
16416 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16417 vec_perm_indices instead of auto_vec_perm_indices.
16418 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16419 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16420 (vect_permute_store_chain): Likewise.
16421 (vect_grouped_load_supported): Likewise.
16422 (vect_permute_load_chain): Likewise.
16423 (vect_shift_permute_load_chain): Likewise.
16424 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16425 (vect_transform_slp_perm_load): Likewise.
16426 (vect_schedule_slp_instance): Likewise.
16427 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16428 (vectorizable_mask_load_store): Likewise.
16429 (vectorizable_bswap): Likewise.
16430 (vectorizable_store): Likewise.
16431 (vectorizable_load): Likewise.
16432 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16433 vec_perm_indices instead of auto_vec_perm_indices. Use
16434 tree_to_vec_perm_builder to read the vector from a tree.
16435 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16436 vec_perm_builder instead of a vec_perm_indices.
16437 (have_whole_vector_shift): Use vec_perm_builder and
16438 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16439 truncation to calc_vec_perm_mask_for_shift.
16440 (vect_create_epilog_for_reduction): Likewise.
16441 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16442 from auto_vec_perm_indices to vec_perm_indices.
16443 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16444 instead of changing individual elements.
16445 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16446 the vector in d.perm.
16447 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16448 from auto_vec_perm_indices to vec_perm_indices.
16449 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16450 instead of changing individual elements.
16451 (arm_vectorize_vec_perm_const): Use new_vector to install
16452 the vector in d.perm.
16453 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16454 Update vec_perm_builder constructor.
16455 (rs6000_expand_interleave): Likewise.
16456 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16457 (rs6000_expand_interleave): Likewise.
16459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16461 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16462 to qimode could truncate the indices.
16463 * optabs.c (expand_vec_perm_var): Likewise.
16465 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16467 * Makefile.in (OBJS): Add vec-perm-indices.o.
16468 * vec-perm-indices.h: New file.
16469 * vec-perm-indices.c: Likewise.
16470 * target.h (vec_perm_indices): Replace with a forward class
16472 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16473 * optabs.h: Include vec-perm-indices.h.
16474 (expand_vec_perm): Delete.
16475 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16476 (expand_vec_perm_const): Declare.
16477 * target.def (vec_perm_const_ok): Replace with...
16478 (vec_perm_const): ...this new hook.
16479 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16480 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16481 * doc/tm.texi: Regenerate.
16482 * optabs.def (vec_perm_const): Delete.
16483 * doc/md.texi (vec_perm_const): Likewise.
16484 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16485 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16486 expand_vec_perm for constant permutation vectors. Assert that
16487 the mode of variable permutation vectors is the integer equivalent
16488 of the mode that is being permuted.
16489 * optabs-query.h (selector_fits_mode_p): Declare.
16490 * optabs-query.c: Include vec-perm-indices.h.
16491 (selector_fits_mode_p): New function.
16492 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16493 is defined, instead of checking whether the vec_perm_const_optab
16494 exists. Use targetm.vectorize.vec_perm_const instead of
16495 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16496 fit in the vector mode before using a variable permute.
16497 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16498 vec_perm_indices instead of an rtx.
16499 (expand_vec_perm): Replace with...
16500 (expand_vec_perm_const): ...this new function. Take the selector
16501 as a vec_perm_indices rather than an rtx. Also take the mode of
16502 the selector. Update call to shift_amt_for_vec_perm_mask.
16503 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16504 Use vec_perm_indices::new_expanded_vector to expand the original
16505 selector into bytes. Check whether the indices fit in the vector
16506 mode before using a variable permute.
16507 (expand_vec_perm_var): Make global.
16508 (expand_mult_highpart): Use expand_vec_perm_const.
16509 * fold-const.c: Includes vec-perm-indices.h.
16510 * tree-ssa-forwprop.c: Likewise.
16511 * tree-vect-data-refs.c: Likewise.
16512 * tree-vect-generic.c: Likewise.
16513 * tree-vect-loop.c: Likewise.
16514 * tree-vect-slp.c: Likewise.
16515 * tree-vect-stmts.c: Likewise.
16516 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16518 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16519 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16520 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16521 (aarch64_vectorize_vec_perm_const): ...this new function.
16522 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16523 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16524 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16525 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16526 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16527 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16528 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16530 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16531 check for NEON modes.
16532 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16533 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16534 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16535 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16537 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16538 the old VEC_PERM_CONST conditions.
16539 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16540 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16541 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16542 (ia64_vectorize_vec_perm_const_ok): Merge into...
16543 (ia64_vectorize_vec_perm_const): ...this new function.
16544 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16545 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16546 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16547 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16548 * config/mips/mips.c (mips_expand_vec_perm_const)
16549 (mips_vectorize_vec_perm_const_ok): Merge into...
16550 (mips_vectorize_vec_perm_const): ...this new function.
16551 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16552 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16553 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16554 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16555 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16556 (rs6000_expand_vec_perm_const): Delete.
16557 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16559 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16560 (altivec_expand_vec_perm_const_le): Take each operand individually.
16561 Operate on constant selectors rather than rtxes.
16562 (altivec_expand_vec_perm_const): Likewise. Update call to
16563 altivec_expand_vec_perm_const_le.
16564 (rs6000_expand_vec_perm_const): Delete.
16565 (rs6000_vectorize_vec_perm_const_ok): Delete.
16566 (rs6000_vectorize_vec_perm_const): New function.
16567 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16568 an element count and rtx array.
16569 (rs6000_expand_extract_even): Update call accordingly.
16570 (rs6000_expand_interleave): Likewise.
16571 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16572 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16573 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16574 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16575 (rs6000_expand_vec_perm_const): Delete.
16576 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16577 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16578 (altivec_expand_vec_perm_const_le): Take each operand individually.
16579 Operate on constant selectors rather than rtxes.
16580 (altivec_expand_vec_perm_const): Likewise. Update call to
16581 altivec_expand_vec_perm_const_le.
16582 (rs6000_expand_vec_perm_const): Delete.
16583 (rs6000_vectorize_vec_perm_const_ok): Delete.
16584 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16585 reference to the SPE evmerge intructions.
16586 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16587 an element count and rtx array.
16588 (rs6000_expand_extract_even): Update call accordingly.
16589 (rs6000_expand_interleave): Likewise.
16590 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16591 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16593 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16595 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16597 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16598 vector mode and that that mode matches the mode of the data
16600 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16601 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16602 directly using expand_vec_perm_1 when forcing selectors into
16604 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16608 * optabs-query.h (can_vec_perm_p): Delete.
16609 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16610 * optabs-query.c (can_vec_perm_p): Split into...
16611 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16612 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16613 particular selector is valid.
16614 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16615 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16616 (vect_grouped_load_supported): Likewise.
16617 (vect_shift_permute_load_chain): Likewise.
16618 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16619 (vect_transform_slp_perm_load): Likewise.
16620 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16621 (vectorizable_bswap): Likewise.
16622 (vect_gen_perm_mask_checked): Likewise.
16623 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16624 implementations of variable permutation vectors into account
16625 when deciding which selector to use.
16626 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16627 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16628 with a false third argument.
16629 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16630 to test whether the constant selector is valid and can_vec_perm_var_p
16631 to test whether a variable selector is valid.
16633 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16635 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16636 * optabs-query.c (can_vec_perm_p): Likewise.
16637 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16638 instead of vec_perm_indices.
16639 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16640 (vect_gen_perm_mask_checked): Likewise,
16641 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16642 (vect_gen_perm_mask_checked): Likewise,
16644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16646 * optabs-query.h (qimode_for_vec_perm): Declare.
16647 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16648 (qimode_for_vec_perm): ...this new function.
16649 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16651 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16653 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16654 does not have a conditional at the top.
16656 2018-01-02 Richard Biener <rguenther@suse.de>
16658 * ipa-inline.c (big_speedup_p): Fix expression.
16660 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16663 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16666 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16670 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16671 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16672 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16673 cond_taken_branch_cost 3->4.
16675 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16677 PR tree-optimization/83581
16678 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16679 TODO_cleanup_cfg if any changes have been made.
16681 PR middle-end/83608
16682 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16683 convert_modes if target mode has the right side, but different mode
16686 PR middle-end/83609
16687 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16688 last argument when extracting from CONCAT. If either from_real or
16689 from_imag is NULL, use expansion through memory. If result is not
16690 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16691 the parts directly to inner mode, if even that fails, use expansion
16694 PR middle-end/83623
16695 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16696 check for bswap in mode rather than HImode and use that in expand_unop
16699 Copyright (C) 2018 Free Software Foundation, Inc.
16701 Copying and distribution of this file, with or without modification,
16702 are permitted in any medium without royalty provided the copyright
16703 notice and this notice are preserved.